Session Date/Time: 16 Apr 2026 13:15
WEBVTT
1 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:00:00.000 --> 00:00:01.750 Hashington.
2 "TRN6-29-BANFF/speaker_5" (890799616_5) 00:00:03.750 --> 00:00:06.750 Sebastian, whatever you say is being recorded.
3 "TRN6-29-BANFF/speaker_5" (890799616_5) 00:00:07.250 --> 00:00:08.750 What are you telling Nate?
4 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:00:10.750 --> 00:00:12.250 So you can get seeded as director.
5 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:00:13.250 --> 00:00:37.399 Let's go ahead and get started folks. So once again, this is the IETF note well. You should be familiar with it by now, but if not, please do take some time to do so by using favorite internet search engine. These are the policies under which we participate here. Also, as a reminder, we do have a lot of remote participants and audio being the web.
6 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:00:37.399 --> 00:00:51.659 Say that it is, we're gonna ask that folks sustain from having side conversations or making lots of extraneeous noise and when you're speaking, please do so distinctly and loudly, even uncomfortably loudly for some people.
7 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:00:53.860 --> 00:01:01.610 So today, we have half of a day, we're scheduled up until 1pmI believe.
8 "TRN6-29-BANFF/speaker_8" (890799616_8) 00:01:01.610 --> 00:01:02.110 We.
9 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:01:02.110 --> 00:01:12.970 We may use all of that time, we may not. We've gone through everything on the issues list that we identified in the agenda for the break. Congratulations. Thank you all very much.
10 "TRN6-29-BANFF/speaker_6" (890799616_6) 00:01:13.470 --> 00:01:14.470 That's.
11 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:01:14.470 --> 00:01:25.350 Is that good enough for you? Oh, that **** it. So what we thought we might do is to reprise a few of the key areas we've discussed.
12 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:01:25.350 --> 00:01:41.700 And I think 1st we're gonna take a quick look back at search where we ended up yesterday to see if anybody had had any reflections or suggestions popped into their minds between the end of the meeting yesterday and this morning. Then we'll do likewise for the training term.
13 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:01:41.700 --> 00:02:01.700 I will briefly go back to section three two, which I think Martin had made a few adjustments to the PR we can take a look at and if other people have any reflections. We're not gonna spend a tremendous amount of time on either of those three. We just want to see if we can get a little bit of extra juice out of the discussion while we're all here in the room or synchronously on.
14 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:02:01.700 --> 00:02:22.880 Online. And then we can spend a little time going back to the use discussion, seeing if people have had any thoughts about requirements or or proposals or or trying to to get a little more context on, on what the problems and what the possibilities are there so that people can go away and come up.
15 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:02:22.880 --> 00:02:48.470 Get more fully formed proposals. If if we'll see how that discussion goes, it might be that we break early and allow a smaller group of folks to go off and have a, a more unstructured casual discussion and if that's the right thing to do, we'll do that. But I I don't expect we're gonna come to any significant progress with, with use today, but if we can have little preliminary discussion to, to help look those further efforts I think that would be a really good use of our time here if that makes sense.
16 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:02:48.970 --> 00:02:57.970 And, and we'll also I we talked briefly yesterday about our future arrangements. I don't know if there's anything to add to that at this point, so I think we don't need to do that, do we.
17 "TRN6-29-BANFF/speaker_5" (890799616_5) 00:02:57.970 --> 00:03:04.970 I think like we did simply talk about like upcoming meeting stuff, like maybe we can kind of wrap on it for people who are not herious.
18 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:03:04.970 --> 00:03:22.880 Today sure 5 min summary of that, yeah. But otherwise, you know, we'll of course talk about next steps based upon where we're at the end of the day, but I I think we all kind of roughly know what that looks like in terms of we need more proposals, we need to refine the text that we do have, take home all the things.
19 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:03:22.880 --> 00:03:50.360 We've talked about this week and and and you know reflect on it, see if we can fine tune it and confirm that it's the right way to go. Oh, thank you, and yes, we have notes being taken. Thank you very much. If folks could collaborate and contribute to that, that would be really helpful. Ok, so let's get the notesball off the screen or do this. Why is my browser not responding.
20 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:03:50.360 --> 00:03:51.990 There we go.
21 "TRN6-29-BANFF/speaker_5" (890799616_5) 00:03:51.990 --> 00:03:56.000 Are you not using Firefox?
22 "TRN6-29-BANFF/speaker_4" (890799616_4) 00:03:56.000 --> 00:03:59.179 He's using a real browser.
23 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:03:59.179 --> 00:04:01.929 That's good. Sure. Safari.
24 "TRN6-29-BANFF/speaker_6" (890799616_6) 00:04:01.929 --> 00:04:03.929 Oh NO, that's not real.
25 "TRN6-29-BANFF/speaker_2" (890799616_2) 00:04:03.929 --> 00:04:04.429 Awesome.
26 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:04:13.119 --> 00:04:16.140 So how is the network going to look better?
27 "TRN6-29-BANFF/speaker_6" (890799616_6) 00:04:16.140 --> 00:04:17.390 I should probably throw in the body.
28 "TRN6-29-BANFF/speaker_2" (890799616_2) 00:04:17.390 --> 00:04:17.890 Okay.
29 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:04:17.890 --> 00:04:38.019 So we had cert and from memory that was we selected two oh one, correct? Yep.
30 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:04:39.799 --> 00:04:56.720 So this is the language where we ended up yesterday, make it a little bit bigger for folks if I can. Did anybody have any thoughts about this or or Martin, this was stable after the meeting, wasn't it?
31 "TRN6-29-BANFF/speaker_6" (890799616_6) 00:04:56.720 --> 00:05:06.619 Yeah, yeah, that hasn't changed since we last.
32 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:05:06.619 --> 00:05:13.619 Happy to leave this where it is. Just wanted to give folks an opportunity to make any further comments about it.
33 "TRN6-29-BANFF/speaker_6" (890799616_6) 00:05:13.619 --> 00:05:17.869 Again that's the existing sentence that's just been tweaked a tiny bit. Sure.
34 "TRN6-29-BANFF/speaker_2" (890799616_2) 00:05:18.369 --> 00:05:28.510 Are there any thoughts about substantive, non substantive?
35 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:05:30.010 --> 00:05:34.510 That was that was the sticking point before the substances of non stop substance.
36 "TRN6-29-BANFF/speaker_5" (890799616_5) 00:05:34.510 --> 00:05:35.510 Subsentimented.
37 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:05:35.510 --> 00:05:36.010 Yeah, but.
38 "TRN6-29-BANFF/speaker_6" (890799616_6) 00:05:36.010 --> 00:05:37.010 I think there's been an issue.
39 "TRN6-29-BANFF/speaker_2" (890799616_2) 00:05:37.010 --> 00:05:37.510 Well for the.
40 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:05:37.510 --> 00:05:39.260 Good.
41 "TRN6-29-BANFF/speaker_6" (890799616_6) 00:05:39.260 --> 00:05:40.010 That's it?
42 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:05:40.010 --> 00:05:54.040 Ice.
43 "TRN6-29-BANFF/speaker_5" (890799616_5) 00:05:54.040 --> 00:05:56.059 Is raise hand.
44 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:05:56.059 --> 00:05:57.309 There is a race hand Chris.
45 "TRN6-29-BANFF/speaker_6" (890799616_6) 00:05:57.309 --> 00:05:59.380 Sure.
46 "TRN6-29-BANFF/speaker_1" (890799616_1) 00:05:59.380 --> 00:06:00.130 That's two.
47 "TRN6-29-BANFF/speaker_6" (890799616_6) 00:06:01.130 --> 00:06:09.630 So two oh two is what we discussed yesterday. I think the question was about the choice of word more than.
48 "TRN6-29-BANFF/speaker_2" (890799616_2) 00:06:09.630 --> 00:06:10.130 Oh yeah.
49 "TRN6-29-BANFF/speaker_6" (890799616_6) 00:06:10.130 --> 00:06:27.599 The overall thing, whether this paragraph that talks about changes to the presentation includes the term unsubstantive. And I heard from from Nate very strongly.
50 "TRN6-29-BANFF/speaker_6" (890799616_6) 00:06:27.599 --> 00:06:45.790 Prefers to retain that word and I heard some concerns about the inclusion of the word, but I don't think this is a like a major point of contention behalfs, but obviously if we can do something better then need to do that.
51 "TRN6-29-BANFF/speaker_5" (890799616_5) 00:06:45.790 --> 00:07:00.790 So I I I think the concern that Andrew brought up yesterday, he's not here today, right? Like, but like I think what he brought up was like, you know, non substantive is like how do you measure that was like really how do you differentiate that from a substantive change was the question. I think we can solve it, like you said, it's probably more.
52 "TRN6-29-BANFF/speaker_6" (890799616_6) 00:07:00.790 --> 00:07:06.290 So made the point these are preferences, so maybe we don't need to have this be absolutely ironclated.
53 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:07:06.290 --> 00:07:09.500 Right, right.
54 "TRN6-29-BANFF/speaker_6" (890799616_6) 00:07:09.500 --> 00:07:18.750 So yeah, Brad's on the screen there saying that open to words that could improve clarity and that's where we're at. Sure.
55 "TRN6-29-BANFF/speaker_5" (890799616_5) 00:07:18.750 --> 00:07:26.420 And Chris, you load your hand. Was your point made or?
56 "TRN6-29-BANFF/speaker_6" (890799616_6) 00:07:26.420 --> 00:07:29.920 Yeah.
57 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:07:29.920 --> 00:07:46.459 So, just from a mechanical standpoint, it strikes me that, you know, if, if we're gonna discuss this for a while and have issues against it, it's gonna be a little bit awkward to evaluate it as a pull request. And I'm wondering if, if folks would be concerned if.
58 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:07:46.459 --> 00:08:16.309 We, go ahead and and merged this PR into the editor's draft, and within agreement that if if it's not gonna be in a either two things will have one of one of two things will happen. Either we'll get consensus on it before we publish the next working group draft or we flag it as not yet having consensus if if we do publish a new draft. People, is that comfortable for folks? Just that I think for a lot of people especially for folks who aren't familiar with github.
59 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:08:16.309 --> 00:08:20.640 It's difficult to find and evaluate this text.
60 "TRN6-29-BANFF/speaker_2" (890799616_2) 00:08:22.119 --> 00:08:25.369 Let me merge it with a banner that says that or.
61 "TRN6-29-BANFF/speaker_5" (890799616_5) 00:08:25.369 --> 00:08:28.869 It's been merged into the editor's copy before publication.
62 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:08:28.869 --> 00:08:36.559 Right. I mean, to be clear, we can put that better on the entire draft, nothing in the draft currently has consensus. Yeah.
63 "TRN6-29-BANFF/speaker_2" (890799616_2) 00:08:36.559 --> 00:08:40.780 I don't want to. You know how you wanted to handle those.
64 "TRN6-29-BANFF/speaker_5" (890799616_5) 00:08:40.780 --> 00:08:42.598 Chris, go ahead 1st.
65 "Chris Needham" (1992016384) 00:08:42.598 --> 00:09:13.679 Thank you. Yes, sorry I couldn't reply earlier. I was juggling too many open windows. So I so I noticed, I think somebody in GitHub Tyler commented about verbatim and made or asked the question and I think in doing so sort of suggests whether the, like the provisions that we've included around translation should be sort of qualified as like verbatim translations.
66 "Chris Needham" (1992016384) 00:09:13.679 --> 00:09:30.479 And I think this perhaps points to perhaps needing a bit more clarity around what we mean by non substantive as well. So that comment I think came in on oh I lost track of it.
67 "Chris Needham" (1992016384) 00:09:30.479 --> 00:09:42.080 On the pull request to remove verbatim, which was, I forget which number. Yeah.
68 "TRN6-29-BANFF/speaker_5" (890799616_5) 00:09:42.080 --> 00:09:52.830 Yeah, I, I think we just did a quick poll yesterday Chris, to see if you want to keep it or not, but like personally to me like you know translation doesn't make sense. Like, you know, it's kind of contradictory to.
69 "TRN6-29-BANFF/speaker_6" (890799616_6) 00:09:52.830 --> 00:09:54.900 Say that.
70 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:09:54.900 --> 00:09:56.400 But.
71 "TRN6-29-BANFF/speaker_5" (890799616_5) 00:09:56.400 --> 00:10:02.364 But if you want to raise an issue, Chris, like please go ahead and put it in and we can talk over it, but.
72 "Chris Needham" (1992016384) 00:10:02.364 --> 00:10:08.158 Yeah, so I'm just I'm just relaying Tyler really at this point.
73 "TRN6-29-BANFF/speaker_5" (890799616_5) 00:10:08.158 --> 00:10:09.908 Thank you, Lauren.
74 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:10:10.408 --> 00:10:15.408 Well yes yesterday I think it was who asked for a definition of small.
75 "TRN6-29-BANFF/speaker_5" (890799616_5) 00:10:15.408 --> 00:10:15.908 Can you speak.
76 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:10:15.908 --> 00:10:22.958 Okay, Aaron was I think asking for a definition of summaries and when we.
77 "Nick Sullivan" (652825600) 00:10:22.958 --> 00:10:30.518 Yeah, yeah.
78 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:10:30.518 --> 00:10:57.690 Okay, when we read there is a one sentence paragraph about summaries, NO definition and NO relationship with non substantive excepts etc. So there will be something to maybe not to define explicitly, but maybe to highlight the relationship between what is an except, what is a substantive change and what is a summary. Is that clear?
79 "TRN6-29-BANFF/speaker_2" (890799616_2) 00:10:57.690 --> 00:10:58.190 Yep.
80 "TRN6-29-BANFF/speaker_2" (890799616_2) 00:10:58.690 --> 00:10:59.190 Yeah.
81 "TRN6-29-BANFF/speaker_5" (890799616_5) 00:10:59.190 --> 00:11:00.690 Lennon.
82 "TRN6-29-BANFF/speaker_2" (890799616_2) 00:11:00.690 --> 00:11:25.690 That's what I wanted to do, I wanted to get back to the logistics. I wonder if it might be better given that this is still really slightly controversial rather than merging it into the editor's draft, would it, could we do it as a specific because we don't really have a PR of just this against a current draft. That might be easier to add it as well, so.
83 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:11:25.690 --> 00:11:29.190 That's what this is, right? This is a PR against the current draft.
84 "TRN6-29-BANFF/speaker_2" (890799616_2) 00:11:29.190 --> 00:11:31.190 Of the current draft? Yes.
85 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:11:31.190 --> 00:11:43.190 Yeah. I, I you know thinking through it I'm wondering if maybe we should just proactively go and label every section is that this section does not have consensus yet, just because there's so many sensitivities around this.
86 "TRN6-29-BANFF/speaker_6" (890799616_6) 00:11:43.190 --> 00:11:46.190 On the top of the document that says this entire document. Oh.
87 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:11:46.190 --> 00:11:50.400 We'll we'll take that to discuss with the editors to try and clarify that.
88 "TRN6-29-BANFF/speaker_5" (890799616_5) 00:11:50.400 --> 00:11:52.780 Thanks, Mike.
89 "TRN6-29-BANFF/speaker_1" (890799616_1) 00:11:52.780 --> 00:11:53.530 So.
90 "TRN6-29-BANFF/speaker_2" (890799616_2) 00:11:53.530 --> 00:12:07.619 At the risk of blowing anything up, this explicitly excludes the use of summaries. Do we want content owners to be able to express that they are willing to be included in summaries?
91 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:12:07.619 --> 00:12:17.039 That, go ahead and propose a new vocabulary term if you would like to do that.
92 "TRN6-29-BANFF/speaker_1" (890799616_1) 00:12:17.039 --> 00:12:18.789 Yeah, can I ask a clarifying question?
93 "TRN6-29-BANFF/speaker_6" (890799616_6) 00:12:18.789 --> 00:12:19.289 Sure.
94 "TRN6-29-BANFF/speaker_1" (890799616_1) 00:12:19.789 --> 00:12:27.789 So you're saying where someone wants to opt out of training but opt into kind of full AI search?
95 "TRN6-29-BANFF/speaker_2" (890799616_2) 00:12:27.789 --> 00:12:32.440 Yeah. Okay.
96 "TRN6-29-BANFF/speaker_5" (890799616_5) 00:12:32.440 --> 00:12:34.190 Yeah.
97 "TRN6-29-BANFF/speaker_2" (890799616_2) 00:12:34.190 --> 00:12:35.190 I I.
98 "TRN6-29-BANFF/speaker_6" (890799616_6) 00:12:35.190 --> 00:12:35.690 Agree.
99 "TRN6-29-BANFF/speaker_1" (890799616_1) 00:12:35.690 --> 00:12:47.690 Mike, I think there, there's probably a really strong need for for this sort of preference of not training but still being included in search results and.
100 "TRN6-29-BANFF/speaker_1" (890799616_1) 00:12:48.190 --> 00:12:51.690 Things like brag. The other point I wanted to make.
101 "TRN6-29-BANFF/speaker_2" (890799616_2) 00:12:51.690 --> 00:12:52.190 Because that.
102 "TRN6-29-BANFF/speaker_1" (890799616_1) 00:12:52.190 --> 00:13:07.190 Verbatim is pretty challenging here to define because there could be things like encoding changes. Is that verbatim if someone does escaping of characters and whatnot, I think it's just a challenging word.
103 "TRN6-29-BANFF/speaker_6" (890799616_6) 00:13:07.190 --> 00:13:07.690 So I have.
104 "TRN6-29-BANFF/speaker_1" (890799616_1) 00:13:07.690 --> 00:13:09.690 Happy removing.
105 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:13:09.690 --> 00:13:13.436 Right?
106 "Brad Silver" (352249088) 00:13:13.436 --> 00:13:31.069 Thanks. I I I I think I I I raised this yesterday, but it was the end of the day and I left and I I think we were all running out of steam. The last sentence, and the the condition that it's ok to use the.
107 "Brad Silver" (352249088) 00:13:31.069 --> 00:13:51.069 The the content for training AI models as long as the outputs are only used in accordance with the above the conditions. In other words, you can train a whole model as long as the use of the content in relation to that model is somehow restricted in terms of the way that the output works. That, that, that feels like it goes much further than it needs.
108 "Brad Silver" (352249088) 00:13:51.069 --> 00:14:14.119 Needs to go. I I feel like what we were aiming for here is if you want to be in search, then you gotta be ok with your content being used to train things or be processed in accordance or with the extent to the extent that is necessary to provide the product that you are saying you're ok with me.
109 "Brad Silver" (352249088) 00:14:14.119 --> 00:14:44.539 Being a part of. This goes further than that. This says that the, the, that a model can be trained using your content, but it just might kind of have a guardrail on it in relation to that specific content. And that's, that's I think a material you know the step away from just saying I'm gonna allow my content to be used solely to the extent necessary to provide this service that I'm, I'm opting into.
110 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:14:44.539 --> 00:15:10.746 So, so Brad, I playing back the discussion we had in, in my mind, I think the place we arrived at was the the expectation is is that the content can be used to train a model, but that model has constraints on its use. It can only be used for that search service. And then I I would agree that this, this sentence could say that in a better way, but it sounds like you're going in a different direction.
111 "Brad Silver" (352249088) 00:15:10.746 --> 00:15:27.229 Yeah, well maybe if I was presented with saying it in a different way, I might be like, ok, that's that's that that could be fine. But here's the just stepping backwards for a second, right? Let's just so there's, there's publishers and search.
112 "Brad Silver" (352249088) 00:15:27.229 --> 00:15:47.229 Obviously have a, sort of like a symbiotic relationship here. You know, publishers need search for referral traffic. There's a strong incentive/pressure to say ok to search because you need that. Companies that are providing search are also building models. They.
113 "Brad Silver" (352249088) 00:15:47.229 --> 00:16:07.229 They get access to content for building those models by virtue of the fact that they are search providers. This gives them a tremendous amount of power to build models that maybe other companies that are building models don't have because they don't have the ability to say, we need, you know, to have access to that, that level of, of.
114 "Brad Silver" (352249088) 00:16:07.229 --> 00:16:27.709 Of of of of content. I think we need to just kind of have that balance in mind that when we provided content for search, we only wanted to be used for search and we only want it to be used for AI processing in relation to the provision of that search service period.
115 "Brad Silver" (352249088) 00:16:27.709 --> 00:16:51.802 I don't think that it's fair or right, that we are sort of giving up or saying yes to more anything more than that. And this sentence implies that there is something more than that, as long as there is some safeguard built into the output side of whatever model they are building using the content that they wouldn't otherwise have had access to if they weren't a search provider.
116 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:16:51.802 --> 00:17:07.052 So I don't think you do diverge where we went yesterday. I hear a lot of commonality between what you're saying and what Nate was saying yesterday and I I I think that this is the intent here. It's just not perfectly captured yet.
117 "TRN6-29-BANFF/speaker_5" (890799616_5) 00:17:07.052 --> 00:17:12.552 Just skip ahead and like talk like let Justin go next, but if you have some thoughts, just talk about it.
118 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:17:13.052 --> 00:17:15.300 Five.
119 "TRN6-29-BANFF/speaker_6" (890799616_6) 00:17:15.300 --> 00:17:31.859 Gone, I I recognize Brad's concern and the intent of this was to do exactly what Brad was requesting now and I think that was the agreement in the room. I've put a, a a small tweak to the language here that's now on screen.
120 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:17:31.859 --> 00:17:32.609 Does that.
121 "TRN6-29-BANFF/speaker_6" (890799616_6) 00:17:32.609 --> 00:17:44.601 Help from Brad.
122 "Brad Silver" (352249088) 00:17:44.601 --> 00:18:01.881 It's better. I think maybe we said training of AI models required for the provision of s of of of the search service that is used subject. So I just sort of the AI models. I want to make sure that this is an AI model.
123 "TRN6-29-BANFF/speaker_6" (890799616_6) 00:18:01.881 --> 00:18:05.042 I want to get to the substance of this though. Yes.
124 "Brad Silver" (352249088) 00:18:05.042 --> 00:18:14.097 But yes, I think you've heard what I'm saying, if you say that that's what you want to convey then, then yes, I don't think we have to like find the exact language right now.
125 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:18:14.097 --> 00:18:36.679 Right. And I think there are a lot of ways Martin's done one way to express it, but I think you wanna make sure that the model that is trained with with this data if it is if it's trained under this permission, the permission only applies if or or the preference only applies if that model is is exclusively used to provide that search service. Yeah.
126 "Brad Silver" (352249088) 00:18:36.679 --> 00:18:46.845 Yeah, I like words that's you know say things like to the extent and yes, I but yes, I I that's that sounds like you're hearing what I'm saying.
127 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:18:46.845 --> 00:18:49.095 Okay.
128 "TRN6-29-BANFF/speaker_6" (890799616_6) 00:18:49.095 --> 00:18:54.929 My understanding is that was the substance of the agreement that was read for today. It's just that we need to be.
129 "TRN6-29-BANFF/speaker_6" (890799616_6) 00:18:54.929 --> 00:19:14.929 Better about spelling it out precisely. And, and when I said outputs, my thinking was outputs was both the training of the model as an output and the outputs of that model in its use, but that was insufficiently precise, I agree, and so we'll iterate on on the exact wording, but as long as we.
130 "TRN6-29-BANFF/speaker_6" (890799616_6) 00:19:14.929 --> 00:19:23.180 Agree on the substance, I think we can still do the editorial work to make it clearer.
131 "TRN6-29-BANFF/speaker_5" (890799616_5) 00:19:23.180 --> 00:19:35.430 So, just before you go Justin, right? So maybe like can we keep track of that as a separate issue? Sure, like, instead of like keeping the spending and like keeping all this like.
132 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:19:36.430 --> 00:19:43.430 That's the intent is merge this and then we it's easier to create individual issues against the text in the draft.
133 "TRN6-29-BANFF/speaker_5" (890799616_5) 00:19:43.430 --> 00:19:45.780 Yeah, thanks, Justin.
134 "TRN6-29-BANFF/speaker_1" (890799616_1) 00:19:45.780 --> 00:19:53.030 I like the use of not substantive now, so thank you for that change. I think we're getting closer to items.
135 "TRN6-29-BANFF/speaker_6" (890799616_6) 00:19:53.030 --> 00:19:53.530 It's a.
136 "TRN6-29-BANFF/speaker_1" (890799616_1) 00:19:53.530 --> 00:19:55.530 It's important that non substantive.
137 "TRN6-29-BANFF/speaker_2" (890799616_2) 00:19:55.530 --> 00:19:56.030 The.
138 "TRN6-29-BANFF/speaker_1" (890799616_1) 00:19:56.030 --> 00:19:57.530 Means that the material.
139 "TRN6-29-BANFF/speaker_2" (890799616_2) 00:19:57.530 --> 00:19:58.030 Meaning of your.
140 "TRN6-29-BANFF/speaker_1" (890799616_1) 00:19:58.030 --> 00:20:05.030 I don't think this covers that, hence the word excert.
141 "TRN6-29-BANFF/speaker_2" (890799616_2) 00:20:05.030 --> 00:20:05.530 How about it.
142 "TRN6-29-BANFF/speaker_1" (890799616_1) 00:20:05.530 --> 00:20:06.530 So I.
143 "TRN6-29-BANFF/speaker_2" (890799616_2) 00:20:06.530 --> 00:20:07.530 I think that's important that I.
144 "TRN6-29-BANFF/speaker_1" (890799616_1) 00:20:07.530 --> 00:20:09.530 I don't feel that it's covering it now.
145 "TRN6-29-BANFF/speaker_1" (890799616_1) 00:20:10.030 --> 00:20:11.030 I'm hearing a.
146 "TRN6-29-BANFF/speaker_2" (890799616_2) 00:20:11.030 --> 00:20:12.530 A lot of, we've had a lot of debate about job.
147 "TRN6-29-BANFF/speaker_1" (890799616_1) 00:20:12.530 --> 00:20:14.030 Sort of summary enough.
148 "TRN6-29-BANFF/speaker_2" (890799616_2) 00:20:14.030 --> 00:20:17.530 So that I mean that allows involvement in general.
149 "TRN6-29-BANFF/speaker_1" (890799616_1) 00:20:17.530 --> 00:20:21.030 Sort of search, but not generative summary.
150 "TRN6-29-BANFF/speaker_2" (890799616_2) 00:20:21.030 --> 00:20:22.030 It maybe inappropriate.
151 "TRN6-29-BANFF/speaker_1" (890799616_1) 00:20:22.030 --> 00:20:26.030 Preference that may or may not be part of the conversation of just being testing.
152 "TRN6-29-BANFF/speaker_2" (890799616_2) 00:20:26.030 --> 00:20:34.030 But related to the combaton search, I I think it's important from a content owner perspective that content owners are not full.
153 "TRN6-29-BANFF/speaker_1" (890799616_1) 00:20:34.030 --> 00:20:35.530 Forced and I'm not saying we're doing this.
154 "TRN6-29-BANFF/speaker_2" (890799616_2) 00:20:35.530 --> 00:20:36.530 Are not forced in.
155 "TRN6-29-BANFF/speaker_1" (890799616_1) 00:20:36.530 --> 00:20:37.530 Some way.
156 "TRN6-29-BANFF/speaker_2" (890799616_2) 00:20:37.530 --> 00:20:39.530 You either i'll.
157 "TRN6-29-BANFF/speaker_1" (890799616_1) 00:20:39.530 --> 00:20:45.530 Allow us to AI or you're out of search in some way. And I don't think we're doing that. I just think that that's an important thing to.
158 "TRN6-29-BANFF/speaker_2" (890799616_2) 00:20:45.530 --> 00:20:46.530 Sect because one.
159 "TRN6-29-BANFF/speaker_1" (890799616_1) 00:20:46.530 --> 00:20:48.530 An organization holds a bit of a gatekeeper.
160 "TRN6-29-BANFF/speaker_2" (890799616_2) 00:20:48.530 --> 00:20:49.030 Role.
161 "TRN6-29-BANFF/speaker_1" (890799616_1) 00:20:49.030 --> 00:20:50.530 To the content creator.
162 "TRN6-29-BANFF/speaker_2" (890799616_2) 00:20:50.530 --> 00:20:51.530 That's why I'm basically mindful.
163 "TRN6-29-BANFF/speaker_1" (890799616_1) 00:20:51.530 --> 00:20:52.959 Those creators.
164 "TRN6-29-BANFF/speaker_5" (890799616_5) 00:20:52.959 --> 00:21:02.209 Yeah and if you look at it specifically, Justin the 1st sentence, underlying sentence, this category does not include the used summary so specifically these are the generated summaries.
165 "TRN6-29-BANFF/speaker_1" (890799616_1) 00:21:02.209 --> 00:21:04.209 So yeah and I think that's important.
166 "TRN6-29-BANFF/speaker_6" (890799616_6) 00:21:04.209 --> 00:21:08.540 So a distinction, ok.
167 "TRN6-29-BANFF/speaker_5" (890799616_5) 00:21:08.540 --> 00:21:11.438 Chris, you had a hand up. Did you wanna say something?
168 "Chris Needham" (1992016384) 00:21:11.438 --> 00:21:28.579 I'm not sure sorry I've kind of been hovering with my hand up and down. So agree very much with what with what Brad said. I'm I'm sort of reassured to to hear that that's so that you know you you're seeing that as the general consensus in the room as well.
169 "Chris Needham" (1992016384) 00:21:28.579 --> 00:21:48.239 Yeah, I think I think, yeah, let's follow up on the on the on the wording. I I do think it needs a bit of tweaking just to make sure that the, you know, the the use, the training and the application of the AI models is, is coped appropriately, but, if we, if the approach is to, to sort of.
170 "Chris Needham" (1992016384) 00:21:48.239 --> 00:21:57.043 Follow that up through, through GitHub issues then I'm I'm fine with that, and that's good.
171 "TRN6-29-BANFF/speaker_5" (890799616_5) 00:21:59.459 --> 00:22:00.209 Laura.
172 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:22:00.209 --> 00:22:05.709 Yes, I I would like to be sure about the scope of, the summaries that are there. Are there.
173 "TRN6-29-BANFF/speaker_1" (890799616_1) 00:22:05.709 --> 00:22:06.709 Or you speak up?
174 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:22:06.709 --> 00:22:14.209 Yeah, is it the summary of one of the results of the summary of the set of results you are?
175 "TRN6-29-BANFF/speaker_6" (890799616_6) 00:22:14.209 --> 00:22:19.540 Looking for it. Either one of them. In this case.
176 "TRN6-29-BANFF/speaker_2" (890799616_2) 00:22:19.540 --> 00:22:22.219 I think.
177 "TRN6-29-BANFF/speaker_6" (890799616_6) 00:22:22.219 --> 00:22:25.969 So, the question came up in chat about about.
178 "TRN6-29-BANFF/speaker_1" (890799616_1) 00:22:25.969 --> 00:22:26.469 And.
179 "TRN6-29-BANFF/speaker_6" (890799616_6) 00:22:26.469 --> 00:22:41.469 So the non substantive changes thing. If you imagine your except is the entire document and your your changes to summarize the entire document, that would be a non substantive change provided that the summary was accurate.
180 "TRN6-29-BANFF/speaker_6" (890799616_6) 00:22:42.509 --> 00:23:02.509 But I think the, the way we address that is the last sentence of the non substandard changes paragraph, which is to say that constraints on, on what can be accepted, what what can be taken and the size of the thing that can be taken constrain what can be extracted from the document and then that further cons.
181 "TRN6-29-BANFF/speaker_6" (890799616_6) 00:23:02.509 --> 00:23:18.140 Strains what, what, what the modifications can be. If, if you can only take one sentence, then you can't summarize anything more than that sentence in in presenting it to someone for accessibility purposes.
182 "TRN6-29-BANFF/speaker_5" (890799616_5) 00:23:18.140 --> 00:23:22.390 Yeah, and and specifically like doing that before any of this applies, right, is the.
183 "TRN6-29-BANFF/speaker_6" (890799616_6) 00:23:22.390 --> 00:23:23.390 Right, right.
184 "TRN6-29-BANFF/speaker_5" (890799616_5) 00:23:23.890 --> 00:23:35.390 Krishna, like, you talked about like I think like February's talked about this like your at least the big search controls that talks about the lens and everything. Is that something you can share publicly on this list on like.
185 "TRN6-29-BANFF/speaker_6" (890799616_6) 00:23:35.390 --> 00:23:37.390 It's pretty well known for.
186 "TRN6-29-BANFF/speaker_5" (890799616_5) 00:23:37.390 --> 00:23:42.390 But if she can just like put it on the Webex chat, like, you know, a URL for that like so Lauren can look at it.
187 "TRN6-29-BANFF/speaker_6" (890799616_6) 00:23:42.890 --> 00:23:45.390 Max snippets and NO structure.
188 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:23:48.760 --> 00:23:57.155 Okay, anything else here? Of course, there's more here, but for today? Chris, Aaron has your hand.
189 "Chris Needham" (1992016384) 00:23:57.155 --> 00:24:05.399 Thank you. Yeah, just just on that point about existing controls. So I, I, you know, I understand from this, you know, what you're.
190 "Chris Needham" (1992016384) 00:24:05.399 --> 00:24:24.449 Sort of referring to like the, you know, the, the NO snippet and max snippet and so on. And, and I, I think in that context it maybe helpful to try to standardize those things. I I think that those are.
191 "Chris Needham" (1992016384) 00:24:24.449 --> 00:24:39.509 Perhaps you know conventions that that exist today, but not necessarily standards document. And so I'm, you know, while I don't I don't particularly see Krishna's draft as.
192 "Chris Needham" (1992016384) 00:24:39.509 --> 00:24:59.509 Yeah, helpful in terms of AI sort of output and AI sort of usage. I I do see it as valuable as a, as a standardized way to to communicate between sites and search engines. And so you know I think from from that point of view, it might be helpful to, you know, to be looking at Krishna's draft as well.
193 "Chris Needham" (1992016384) 00:24:59.509 --> 00:25:14.339 And then you've got something that this could refer to rather than infer, you know, referring implicitly to those things, you know, we would have something that, you know, that's more, you know, more of a standardized thing that this could point to.
194 "TRN6-29-BANFF/speaker_5" (890799616_5) 00:25:14.339 --> 00:25:25.097 That's good, thanks. Chris, just to make sure like the stuff I was talking about, the search engine controls, not necessarily Krishna's draft because there's some correlation to it, but it's something that's pre existing.
195 "Chris Needham" (1992016384) 00:25:25.097 --> 00:25:28.625 Oh, you see or you saying that there is an existing thing for search?
196 "TRN6-29-BANFF/speaker_5" (890799616_5) 00:25:29.375 --> 00:25:35.875 The stuff we use to instruct like the search engines like you know what kind of snippets and stuff like that, how long it can be and things like that.
197 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:25:35.875 --> 00:25:40.375 Yeah, so and I think we're gonna see an issue about that, so.
198 "TRN6-29-BANFF/speaker_3" (890799616_3) 00:25:40.375 --> 00:25:55.209 Yeah, forgive me if we're supposed to only be focusing on the non substantive paragraph, but the bullet point that says only asset, only excerpts from the asset are displayed, would prohibit the display of anything that is not an excerpt from the asset.
199 "TRN6-29-BANFF/speaker_3" (890799616_3) 00:25:55.709 --> 00:26:11.269 Like a color, a link to another asset, you know, words not drawn from the asset in any way. I really feel like these, these ways of phrasing things that are so restrictive that just say only X is allowed.
200 "TRN6-29-BANFF/speaker_3" (890799616_3) 00:26:11.269 --> 00:26:17.670 Are too narrow and inflexible given the way this needs to work.
201 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:26:17.670 --> 00:26:32.670 So, so one reading of the intent here is, is when experts are of the asset are displayed, they are used only to assist you in evaluating the relevance the result, and it sounds like that might address your concern. It wouldn't yeah. I guess the question is, is that the intent and is everybody comfortable with that?
202 "TRN6-29-BANFF/speaker_6" (890799616_6) 00:26:32.670 --> 00:26:40.560 Happy to change.
203 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:26:40.560 --> 00:26:43.099 This English thing is tricky.
204 "TRN6-29-BANFF/speaker_1" (890799616_1) 00:26:43.099 --> 00:26:53.540 Think.
205 "TRN6-29-BANFF/speaker_5" (890799616_5) 00:26:53.540 --> 00:26:56.790 Kevin is your on the same point or like you wanna wait till.
206 "TRN6-29-BANFF/speaker_1" (890799616_1) 00:26:56.790 --> 00:26:58.290 It's a bigger it's a.
207 "TRN6-29-BANFF/speaker_2" (890799616_2) 00:26:58.290 --> 00:26:58.790 Okay.
208 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:26:58.790 --> 00:26:59.290 We'll just.
209 "TRN6-29-BANFF/speaker_5" (890799616_5) 00:26:59.290 --> 00:27:13.500 We'll wait for like market to make some hotfix and then.
210 "TRN6-29-BANFF/speaker_3" (890799616_3) 00:27:15.400 --> 00:27:21.900 I have another but let me know if I should just be making edits in the GitHub.
211 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:27:21.900 --> 00:27:31.390 We're gonna look at this real quick and try and resolve that one you just raised real quick 1st, I think. If I can find it.
212 "TRN6-29-BANFF/speaker_6" (890799616_6) 00:27:31.390 --> 00:27:36.750 You might not yeah ok You might get better.
213 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:27:36.750 --> 00:27:37.250 Okay.
214 "TRN6-29-BANFF/speaker_6" (890799616_6) 00:27:37.250 --> 00:27:42.750 They are only used to assist users in evaluating the relevant.
215 "TRN6-29-BANFF/speaker_1" (890799616_1) 00:27:42.750 --> 00:27:44.040 Result.
216 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:27:44.040 --> 00:27:46.500 That's what they said, not sure.
217 "TRN6-29-BANFF/speaker_6" (890799616_6) 00:27:46.500 --> 00:27:50.750 Does that, is that what you just said, Mark?
218 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:27:51.250 --> 00:27:56.319 Roughly, yeah. Is that closer Aaron?
219 "TRN6-29-BANFF/speaker_3" (890799616_3) 00:27:56.319 --> 00:28:17.069 I still think only is a problem like when are displayed, they are used to assist or they are intended to assist, therefore the purpose of assisting but like NO operator of a platform can control how a user actually uses something.
220 "TRN6-29-BANFF/speaker_6" (890799616_6) 00:28:17.069 --> 00:28:19.239 Yeah. And.
221 "TRN6-29-BANFF/speaker_3" (890799616_3) 00:28:19.239 --> 00:28:29.489 It like making, putting language in there that implies that responsibility is just a doorway to clients and disputes and.
222 "TRN6-29-BANFF/speaker_3" (890799616_3) 00:28:29.989 --> 00:28:31.989 Think unhelpful.
223 "TRN6-29-BANFF/speaker_6" (890799616_6) 00:28:31.989 --> 00:28:34.979 Five.
224 "TRN6-29-BANFF/speaker_3" (890799616_3) 00:28:34.979 --> 00:28:41.479 Excepts from the asset maybe displayed to assist users.
225 "TRN6-29-BANFF/speaker_2" (890799616_2) 00:28:41.479 --> 00:28:43.229 May is a little.
226 "TRN6-29-BANFF/speaker_3" (890799616_3) 00:28:43.229 --> 00:28:46.729 It is permissive, it's saying you don't have to have an excert, but you can.
227 "TRN6-29-BANFF/speaker_2" (890799616_2) 00:28:46.729 --> 00:28:48.229 I agree about all.
228 "TRN6-29-BANFF/speaker_6" (890799616_6) 00:28:48.229 --> 00:28:51.760 What needs to be a condition?
229 "TRN6-29-BANFF/speaker_3" (890799616_3) 00:28:51.760 --> 00:28:53.619 Why?
230 "TRN6-29-BANFF/speaker_6" (890799616_6) 00:28:53.619 --> 00:28:55.869 The structure of the thing if we.
231 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:28:55.869 --> 00:28:59.369 This category of use only applies under the following conditions.
232 "TRN6-29-BANFF/speaker_3" (890799616_3) 00:28:59.369 --> 00:29:22.060 So the only and condition is already there. Saying here is a permission is not saying anything about a condition in which the permission doesn't apply. So you may do x, doesn't say you may also do y, you may also do Z.
233 "TRN6-29-BANFF/speaker_2" (890799616_2) 00:29:22.060 --> 00:29:24.810 It doesn't restrict you from doing why you say.
234 "TRN6-29-BANFF/speaker_3" (890799616_3) 00:29:24.810 --> 00:29:26.310 Right, but.
235 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:29:26.310 --> 00:29:26.810 Yeah.
236 "TRN6-29-BANFF/speaker_3" (890799616_3) 00:29:26.810 --> 00:29:41.959 This says this category views only applies under the following conditions you are doing why. It's already baked in.
237 "TRN6-29-BANFF/speaker_2" (890799616_2) 00:29:41.959 --> 00:29:46.339 You're suggesting get rid of only.
238 "TRN6-29-BANFF/speaker_3" (890799616_3) 00:29:46.339 --> 00:29:47.089 Yeah.
239 "TRN6-29-BANFF/speaker_2" (890799616_2) 00:29:48.089 --> 00:29:51.089 And that seems perfectly reasonable. That's fine. Yeah.
240 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:29:51.089 --> 00:29:52.589 Sorry, Leonard.
241 "TRN6-29-BANFF/speaker_6" (890799616_6) 00:29:52.589 --> 00:29:53.089 I'm.
242 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:29:53.089 --> 00:29:53.589 Speak.
243 "TRN6-29-BANFF/speaker_2" (890799616_2) 00:29:53.589 --> 00:30:10.050 I'm sorry I'm sorry. I was asking if she only wanted to get rid of the word only. She implied that that was ok. And so we were just then sort of heading and hawing about that, but I think that would be a recommendation would be to just remove the word only.
244 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:30:10.050 --> 00:30:10.550 From the.
245 "TRN6-29-BANFF/speaker_3" (890799616_3) 00:30:10.550 --> 00:30:11.050 The the.
246 "TRN6-29-BANFF/speaker_2" (890799616_2) 00:30:11.050 --> 00:30:12.050 Hot from the same.
247 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:30:12.050 --> 00:30:12.550 Underlying.
248 "TRN6-29-BANFF/speaker_2" (890799616_2) 00:30:12.550 --> 00:30:15.050 Does the category of use only applies?
249 "TRN6-29-BANFF/speaker_3" (890799616_3) 00:30:15.050 --> 00:30:17.550 About 2nd bullet? Yes, 2nd bullet.
250 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:30:17.550 --> 00:30:22.219 Very problem. Yeah.
251 "TRN6-29-BANFF/speaker_6" (890799616_6) 00:30:22.219 --> 00:30:25.969 I could remove only from the top as well, I think that would be fine.
252 "TRN6-29-BANFF/speaker_2" (890799616_2) 00:30:25.969 --> 00:30:28.469 Has to remove it from both, sorry.
253 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:30:28.469 --> 00:30:30.500 U.
254 "TRN6-29-BANFF/speaker_2" (890799616_2) 00:30:30.500 --> 00:30:33.276 If you're removing it from 2nd one, then the top one as well.
255 "Chris Needham" (1992016384) 00:30:33.276 --> 00:30:39.420 I, I would object removing it from the top one, I'm sorry. Okay.
256 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:30:39.420 --> 00:30:40.916 If I would.
257 "Chris Needham" (1992016384) 00:30:40.916 --> 00:30:43.645 I don't know if we're doing queuing or or what, but.
258 "TRN6-29-BANFF/speaker_5" (890799616_5) 00:30:43.645 --> 00:30:45.895 Yeah just yeah so.
259 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:30:45.895 --> 00:30:47.079 Kevin.
260 "TRN6-29-BANFF/speaker_5" (890799616_5) 00:30:47.079 --> 00:30:48.829 Did you wanna talk or you wanna wait?
261 "TRN6-29-BANFF/speaker_1" (890799616_1) 00:30:48.829 --> 00:30:50.329 Well, actually now I.
262 "TRN6-29-BANFF/speaker_2" (890799616_2) 00:30:50.329 --> 00:30:51.329 I'm not.
263 "TRN6-29-BANFF/speaker_1" (890799616_1) 00:30:51.329 --> 00:30:56.829 Not sure this makes sense. When excerpts from the asset are displayed, they are only used assignment.
264 "TRN6-29-BANFF/speaker_2" (890799616_2) 00:30:56.829 --> 00:30:59.329 As Aaron knows.
265 "TRN6-29-BANFF/speaker_1" (890799616_1) 00:30:59.329 --> 00:31:06.829 They can't control what the user is using it for. Yeah. And so wouldn't the only better fit when excepts from the app set.
266 "TRN6-29-BANFF/speaker_2" (890799616_2) 00:31:06.829 --> 00:31:07.329 Our.
267 "TRN6-29-BANFF/speaker_1" (890799616_1) 00:31:07.329 --> 00:31:17.939 Displayed only to assist users. So the purpose of displaying the excerpt is only to assist the user.
268 "TRN6-29-BANFF/speaker_5" (890799616_5) 00:31:17.939 --> 00:31:34.130 I think the proposal on the table is remove the word only because there's a word yeah because the word only appears in the overarching condition on the top. So it's just like they are used it's like I think Martin is making a change.
269 "TRN6-29-BANFF/speaker_1" (890799616_1) 00:31:34.630 --> 00:31:37.130 Yeah I'm not sure what that sentence means then, but.
270 "TRN6-29-BANFF/speaker_6" (890799616_6) 00:31:37.130 --> 00:31:39.130 So yeah I'm struggling a little bit as well.
271 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:31:39.130 --> 00:31:43.130 Yeah, I don't wanna get too far into the weeds here.
272 "TRN6-29-BANFF/speaker_5" (890799616_5) 00:31:43.130 --> 00:31:45.420 Chris, go ahead and say your thing.
273 "Chris Needham" (1992016384) 00:31:45.420 --> 00:32:10.418 Thanks. I I just wanted to ask Aaron a sort of clarifying question. Is, is, is your concern that the only, is, is that we can't, we can't put requirements essentially on how users sort of, make use of the output, you know, that that's been displayed or is it that you you're wanting to enable some other kind of.
274 "TRN6-29-BANFF/speaker_4" (890799616_4) 00:32:10.418 --> 00:32:11.168 Usage.
275 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:32:11.168 --> 00:32:13.525 Processing that.
276 "Chris Needham" (1992016384) 00:32:13.525 --> 00:32:39.296 We should really be we that really we should be enumerating as well here. It's the former. Right. Because that cause that part I, I I agree with. My concern is, you know, that, that removing restrictions opens the door to other kinds of usages that, you know, we would need to talk about, but clarifying that, you know, for that former purpose seems, seems perfectly fine to me.
277 "TRN6-29-BANFF/speaker_2" (890799616_2) 00:32:39.296 --> 00:32:41.046 And x serve service.
278 "TRN6-29-BANFF/speaker_1" (890799616_1) 00:32:41.046 --> 00:32:43.619 Displayed only to assist users.
279 "TRN6-29-BANFF/speaker_2" (890799616_2) 00:32:43.619 --> 00:32:45.542 Wouldn't that, wouldn't that work?
280 "Chris Needham" (1992016384) 00:32:45.542 --> 00:32:49.520 It would only in that case Kevin.
281 "TRN6-29-BANFF/speaker_6" (890799616_6) 00:32:49.520 --> 00:32:50.770 You could drop the word on you again.
282 "TRN6-29-BANFF/speaker_1" (890799616_1) 00:32:52.260 --> 00:32:53.510 If that's acceptable.
283 "TRN6-29-BANFF/speaker_1" (890799616_1) 00:32:54.010 --> 00:32:54.510 But.
284 "TRN6-29-BANFF/speaker_3" (890799616_3) 00:32:54.510 --> 00:32:59.010 You don't, you don't want to condition the category on having to fill an excel.
285 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:32:59.010 --> 00:33:01.219 Right.
286 "TRN6-29-BANFF/speaker_3" (890799616_3) 00:33:06.540 --> 00:33:11.290 You always have to have a snippet. Somebody knows snippets, that means they are NO longer in this program.
287 "TRN6-29-BANFF/speaker_6" (890799616_6) 00:33:11.290 --> 00:33:11.790 The 1st.
288 "TRN6-29-BANFF/speaker_2" (890799616_2) 00:33:11.790 --> 00:33:13.790 The semantic thing, right? So.
289 "TRN6-29-BANFF/speaker_6" (890799616_6) 00:33:13.790 --> 00:33:14.290 Is that if.
290 "TRN6-29-BANFF/speaker_2" (890799616_2) 00:33:14.290 --> 00:33:20.790 The Assets are only displayed because assets are displayed only.
291 "TRN6-29-BANFF/speaker_6" (890799616_6) 00:33:20.790 --> 00:33:21.290 To.
292 "TRN6-29-BANFF/speaker_2" (890799616_2) 00:33:21.290 --> 00:33:22.290 Assist the.
293 "TRN6-29-BANFF/speaker_6" (890799616_6) 00:33:22.290 --> 00:33:22.790 So.
294 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:33:22.790 --> 00:33:35.790 I, I think maybe we need to go away and think about this a bit more because, you know, even in the current spelling here without the word only, this category of use only applies under the phone conditions. Excepts from the asset are displayed.
295 "TRN6-29-BANFF/speaker_3" (890799616_3) 00:33:37.000 --> 00:33:37.750 Right.
296 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:33:37.750 --> 00:33:38.750 Okay.
297 "TRN6-29-BANFF/speaker_5" (890799616_5) 00:33:38.750 --> 00:33:40.250 I Leonard if you.
298 "TRN6-29-BANFF/speaker_2" (890799616_2) 00:33:40.250 --> 00:33:47.119 So I put a proposal in both chat and also against the PR.
299 "TRN6-29-BANFF/speaker_6" (890799616_6) 00:33:47.119 --> 00:33:49.369 I don't mind one is one.
300 "TRN6-29-BANFF/speaker_2" (890799616_2) 00:33:49.369 --> 00:33:49.869 Right.
301 "TRN6-29-BANFF/speaker_6" (890799616_6) 00:33:49.869 --> 00:33:53.050 If someone want to read it out?
302 "TRN6-29-BANFF/speaker_2" (890799616_2) 00:33:53.050 --> 00:34:03.050 Yeah, so I wrote I changed it to when excerts from the asset are displayed, they serve to assist users in evaluating the relevance of the results.
303 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:34:03.050 --> 00:34:09.550 That that's more the along the lines of what I had originally was talking about, yeah. Right. When is the important part there?
304 "TRN6-29-BANFF/speaker_5" (890799616_5) 00:34:09.550 --> 00:34:18.050 I do have a different suggestion. I was thinking except from the asset can be displayed to assist users. Kind of makes it.
305 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:34:18.050 --> 00:34:18.550 Right.
306 "TRN6-29-BANFF/speaker_6" (890799616_6) 00:34:18.550 --> 00:34:19.550 Give us another condition, right?
307 "TRN6-29-BANFF/speaker_5" (890799616_5) 00:34:19.550 --> 00:34:20.550 What?
308 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:34:20.550 --> 00:34:44.280 Right, we're getting towards smithing again, so Martin, why don't you incorporate Leonard's suggestion there and we'll let this sit for a little while and see how people percolate with it.
309 "TRN6-29-BANFF/speaker_6" (890799616_6) 00:34:44.280 --> 00:34:54.790 This, you're getting good at this.
310 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:34:54.790 --> 00:34:57.790 I have, Sorry that's that's an old hand, sorry.
311 "TRN6-29-BANFF/speaker_1" (890799616_1) 00:34:57.790 --> 00:34:59.520 Aaron?
312 "TRN6-29-BANFF/speaker_3" (890799616_3) 00:34:59.520 --> 00:35:26.549 Yeah, I think saying that a trained model is an output of a use is gonna have ramifications across the vocabulary that are bad and that we should not like there's a way to scope in limitations on model usage without defining them as an output because that I think is more clearly understood to mean just like the inference time.
313 "TRN6-29-BANFF/speaker_3" (890799616_3) 00:35:26.549 --> 00:35:28.889 Consequence output.
314 "TRN6-29-BANFF/speaker_6" (890799616_6) 00:35:30.300 --> 00:35:36.550 Can you suggest text, alright yeah because we we discussed this, we, we understand the intent, but we.
315 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:35:36.550 --> 00:35:47.050 We, we, we left this in a place where we, we know that this, this sentence needs work, it's just and I think we were in the tent, but the way to spell it is is definitely something we need to go away and think about.
316 "TRN6-29-BANFF/speaker_3" (890799616_3) 00:35:47.050 --> 00:36:01.870 Yeah, I might have to do that too cause regarding an imply could be difficult. I also have concerns about implies of preference, like, can that implication be broken? So I think that whole paragraph needs reworking and I can.
317 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:36:01.870 --> 00:36:13.479 And i'm hoping what we can do is have a a relatively lightweight process where a lot of this is editorial suggestions.
318 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:36:14.229 --> 00:36:20.729 And we'll rely on the editors to try and reflect what they've heard as the agreements in the and the the tenner of the discussion we've had so far.
319 "TRN6-29-BANFF/speaker_6" (890799616_6) 00:36:20.729 --> 00:36:23.729 And we'll give people a chance to review any exactly.
320 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:36:23.729 --> 00:36:36.239 And we'll go through a couple of iterations and hopefully it will be in a place that then we can go and do a consensus call.
321 "TRN6-29-BANFF/speaker_5" (890799616_5) 00:36:36.239 --> 00:36:41.043 Brad is your concern the same as you raised before or some response to what Aaron said?
322 "Brad Silver" (352249088) 00:36:41.043 --> 00:36:50.360 No, it's it's the, it's the same.
323 "TRN6-29-BANFF/speaker_1" (890799616_1) 00:36:50.360 --> 00:36:52.719 Okay.
324 "TRN6-29-BANFF/speaker_2" (890799616_2) 00:36:58.100 --> 00:37:00.719 Yeah.
325 "Chris Needham" (1992016384) 00:37:00.719 --> 00:37:30.619 So, so at the risk of blowing this whole thing up entirely, like there's there's a, there's a I think a very valid point being made in in the chat here, which is we don't actually have a kind of a good definition of what search is, like if you look at to the very top of the definition, we just assume that there's common knowledge of like this is what a search application is and and Kevin's comment talks about, well, do we have a way of sort of meaningfully.
326 "Chris Needham" (1992016384) 00:37:30.619 --> 00:37:41.857 Distinguishing this from other kinds of applications, you know, such as, you know, the various examples of sort of chatbot type interfaces that he gives.
327 "TRN6-29-BANFF/speaker_6" (890799616_6) 00:37:41.857 --> 00:37:48.780 I mean, the, the top, the 1st sentence is an attempt to address exactly that point.
328 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:37:48.780 --> 00:37:50.030 They're actually is also.
329 "TRN6-29-BANFF/speaker_6" (890799616_6) 00:37:50.030 --> 00:38:03.600 Included exactly that language I'm not sure why you're saying that we don't have a definition. I think that's a definition.
330 "TRN6-29-BANFF/speaker_1" (890799616_1) 00:38:03.600 --> 00:38:04.850 Which is the definition?
331 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:38:05.350 --> 00:38:05.850 The search.
332 "TRN6-29-BANFF/speaker_6" (890799616_6) 00:38:05.850 --> 00:38:12.179 Category definition you're pointing out. The application that directs.
333 "TRN6-29-BANFF/speaker_2" (890799616_2) 00:38:12.179 --> 00:38:13.340 Yeah.
334 "TRN6-29-BANFF/speaker_1" (890799616_1) 00:38:13.340 --> 00:38:14.590 But it's.
335 "TRN6-29-BANFF/speaker_6" (890799616_6) 00:38:14.590 --> 00:38:16.590 Sort of I think what people are trying to get.
336 "TRN6-29-BANFF/speaker_4" (890799616_4) 00:38:16.590 --> 00:38:18.590 That's a little bit circular. Your definition of the.
337 "TRN6-29-BANFF/speaker_6" (890799616_6) 00:38:18.590 --> 00:38:20.590 Which category depends on a definition.
338 "TRN6-29-BANFF/speaker_4" (890799616_4) 00:38:20.590 --> 00:38:21.090 Of.
339 "TRN6-29-BANFF/speaker_1" (890799616_1) 00:38:21.090 --> 00:38:22.090 The application that is.
340 "TRN6-29-BANFF/speaker_6" (890799616_6) 00:38:22.090 --> 00:38:22.590 No.
341 "TRN6-29-BANFF/speaker_1" (890799616_1) 00:38:22.590 --> 00:38:23.090 Well it is.
342 "TRN6-29-BANFF/speaker_4" (890799616_4) 00:38:23.090 --> 00:38:24.090 There is one.
343 "TRN6-29-BANFF/speaker_3" (890799616_3) 00:38:24.090 --> 00:38:26.590 Relying original draft.
344 "TRN6-29-BANFF/speaker_6" (890799616_6) 00:38:26.590 --> 00:38:30.929 Directs users to the location of those assets.
345 "TRN6-29-BANFF/speaker_2" (890799616_2) 00:38:30.929 --> 00:38:32.429 So have you dropped.
346 "TRN6-29-BANFF/speaker_4" (890799616_4) 00:38:32.429 --> 00:38:33.929 Forward to search there and just say.
347 "TRN6-29-BANFF/speaker_2" (890799616_2) 00:38:34.429 --> 00:38:36.929 That's that a the search category covers the use of.
348 "TRN6-29-BANFF/speaker_4" (890799616_4) 00:38:36.929 --> 00:38:37.929 Assets in an application.
349 "TRN6-29-BANFF/speaker_2" (890799616_2) 00:38:37.929 --> 00:38:39.429 That directs users.
350 "TRN6-29-BANFF/speaker_6" (890799616_6) 00:38:39.429 --> 00:38:39.929 I would also.
351 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:38:39.929 --> 00:38:41.429 Okay drop search there. Okay.
352 "TRN6-29-BANFF/speaker_1" (890799616_1) 00:38:41.429 --> 00:38:44.929 But but regular chat directs me to things all the time.
353 "TRN6-29-BANFF/speaker_6" (890799616_6) 00:38:45.429 --> 00:38:49.939 Yeah, absolutely.
354 "TRN6-29-BANFF/speaker_1" (890799616_1) 00:38:49.939 --> 00:38:56.459 So how is that a distinction that defines a search application from.
355 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:38:56.459 --> 00:39:02.709 I think it it we probably need the wordsmith along with something like an application whose primary purpose is.
356 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:39:03.209 --> 00:39:03.709 Well.
357 "TRN6-29-BANFF/speaker_4" (890799616_4) 00:39:03.709 --> 00:39:05.709 I do have a definition of search application.
358 "TRN6-29-BANFF/speaker_1" (890799616_1) 00:39:05.709 --> 00:39:06.709 Yeah, we do.
359 "TRN6-29-BANFF/speaker_4" (890799616_4) 00:39:06.709 --> 00:39:07.209 Yes.
360 "TRN6-29-BANFF/speaker_3" (890799616_3) 00:39:07.209 --> 00:39:10.209 But I think the question is, do we want to make that.
361 "TRN6-29-BANFF/speaker_1" (890799616_1) 00:39:10.209 --> 00:39:11.709 I'm just asking the question. I I I.
362 "TRN6-29-BANFF/speaker_3" (890799616_3) 00:39:11.709 --> 00:39:23.419 The question is, do we want to make that distinction or do we want to say that anything that is behaving this way is within the preference regardless of what the like broader kind of category of the application?
363 "TRN6-29-BANFF/speaker_1" (890799616_1) 00:39:23.419 --> 00:39:24.669 Then what, then we need to define.
364 "TRN6-29-BANFF/speaker_2" (890799616_2) 00:39:24.669 --> 00:39:25.169 Find the behavior.
365 "TRN6-29-BANFF/speaker_1" (890799616_1) 00:39:25.169 --> 00:39:26.169 Think you're very clearly.
366 "TRN6-29-BANFF/speaker_3" (890799616_3) 00:39:26.169 --> 00:39:29.219 Yeah.
367 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:39:29.219 --> 00:39:30.969 Right, we do have a queue.
368 "TRN6-29-BANFF/speaker_5" (890799616_5) 00:39:30.969 --> 00:39:36.969 Yeah, we do and in section two Kevin, there's like a definition, if you don't like it, make a pro.
369 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:39:36.969 --> 00:39:37.469 Okay.
370 "TRN6-29-BANFF/speaker_5" (890799616_5) 00:39:37.469 --> 00:39:40.969 Thanks. Our next is Nate.
371 "TRN6-29-BANFF/speaker_4" (890799616_4) 00:39:40.969 --> 00:39:59.120 Yeah, I to be clear like in the conventions and definitions, there is a definition of search application is not part of the search category here and that one there is still the issue that that's it's so broad that it basically because it's just like anything that enables you to load something which is like literally anything on the web basically.
372 "TRN6-29-BANFF/speaker_4" (890799616_4) 00:39:59.120 --> 00:40:01.969 Right, but like there is that definition.
373 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:40:01.969 --> 00:40:02.469 Yeah.
374 "TRN6-29-BANFF/speaker_5" (890799616_5) 00:40:02.469 --> 00:40:05.969 And that's what I meant by the section too, right? Like it's there and if you don't like it.
375 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:40:05.969 --> 00:40:08.469 Yeah.
376 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:40:11.379 --> 00:40:12.129 From.
377 "TRN6-29-BANFF/speaker_1" (890799616_1) 00:40:12.129 --> 00:40:34.800 I guess Tim and robot, I just I'm I'm wondering if it matters whether it's like whether whatever application it is, people decide it is a search or it is not a search if the behavior that this defines is being met, like that, I don't know, it seems.
378 "TRN6-29-BANFF/speaker_1" (890799616_1) 00:40:34.800 --> 00:40:51.930 Yeah seems like the the thing that's trying to be avoided is the, like, either not sending people to the origin or to the asset or generative content that.
379 "TRN6-29-BANFF/speaker_1" (890799616_1) 00:40:51.930 --> 00:41:08.250 Changes the meaning of the asset or misrepresents the meaning of the asset. And this like tackles that, solves that, and so I don't, I don't know that it being classified as search or not.
380 "TRN6-29-BANFF/speaker_1" (890799616_1) 00:41:08.250 --> 00:41:13.560 I'm I'm wondering why that matters.
381 "TRN6-29-BANFF/speaker_5" (890799616_5) 00:41:13.560 --> 00:41:14.310 Kevin.
382 "TRN6-29-BANFF/speaker_1" (890799616_1) 00:41:14.310 --> 00:41:19.310 So this goes back to the point that Mike was making earlier and that I was making yesterday, which is whether.
383 "TRN6-29-BANFF/speaker_2" (890799616_2) 00:41:19.310 --> 00:41:20.810 They're not publishers want to be a.
384 "TRN6-29-BANFF/speaker_1" (890799616_1) 00:41:20.810 --> 00:41:22.310 Able to enable rack.
385 "TRN6-29-BANFF/speaker_2" (890799616_2) 00:41:22.310 --> 00:41:22.810 That follows.
386 "TRN6-29-BANFF/speaker_1" (890799616_1) 00:41:22.810 --> 00:41:24.310 Those conditions like these.
387 "TRN6-29-BANFF/speaker_2" (890799616_2) 00:41:24.310 --> 00:41:25.310 And so.
388 "TRN6-29-BANFF/speaker_1" (890799616_1) 00:41:25.310 --> 00:41:26.310 So to reframe that.
389 "TRN6-29-BANFF/speaker_2" (890799616_2) 00:41:26.310 --> 00:41:26.810 Question.
390 "TRN6-29-BANFF/speaker_1" (890799616_1) 00:41:26.810 --> 00:41:30.310 Do we want this? Do we want an exception like this?
391 "TRN6-29-BANFF/speaker_2" (890799616_2) 00:41:30.310 --> 00:41:31.310 U.
392 "TRN6-29-BANFF/speaker_1" (890799616_1) 00:41:31.310 --> 00:41:33.310 That conditions use of the asset on.
393 "TRN6-29-BANFF/speaker_2" (890799616_2) 00:41:33.310 --> 00:41:35.310 And particular requirements around.
394 "TRN6-29-BANFF/speaker_1" (890799616_1) 00:41:35.310 --> 00:41:39.370 How it is displayed to be an exception for.
395 "TRN6-29-BANFF/speaker_1" (890799616_1) 00:41:39.870 --> 00:41:40.870 Training.
396 "TRN6-29-BANFF/speaker_2" (890799616_2) 00:41:40.870 --> 00:41:42.870 And rack.
397 "TRN6-29-BANFF/speaker_1" (890799616_1) 00:41:42.870 --> 00:41:48.370 Do you want the option to be able to say, I don't mind rag if you're doing this?
398 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:41:48.370 --> 00:41:59.780 But but but that's not this discussion. We're, we, that's the kind of the use. We've we've called that inference, we've called it use, we've called it rag. We need proposals in that space.
399 "TRN6-29-BANFF/speaker_1" (890799616_1) 00:41:59.780 --> 00:42:00.530 No, I'm saying.
400 "TRN6-29-BANFF/speaker_2" (890799616_2) 00:42:00.530 --> 00:42:01.030 I'm sorry.
401 "TRN6-29-BANFF/speaker_1" (890799616_1) 00:42:01.030 --> 00:42:01.530 Oh.
402 "TRN6-29-BANFF/speaker_6" (890799616_6) 00:42:01.530 --> 00:42:02.030 I'm saying.
403 "TRN6-29-BANFF/speaker_1" (890799616_1) 00:42:02.030 --> 00:42:09.530 This could use this, go make this the search exception, make this the exception for use of assets that display stuff like this.
404 "TRN6-29-BANFF/speaker_1" (890799616_1) 00:42:10.030 --> 00:42:12.100 And then that can apply.
405 "TRN6-29-BANFF/speaker_2" (890799616_2) 00:42:12.100 --> 00:42:15.350 To either or both the rag or train.
406 "TRN6-29-BANFF/speaker_5" (890799616_5) 00:42:15.350 --> 00:42:31.350 I I think the intent is like there's people who want one or the other, so I don't think we can combine them. I think that's why we had to leave it to later like to figure it out because I I think from the people who want to express the preferences at least it's a clear indication people would say one yes to one and not not not to other.
407 "TRN6-29-BANFF/speaker_2" (890799616_2) 00:42:31.350 --> 00:42:33.850 Coming on them to be able to define the differences.
408 "TRN6-29-BANFF/speaker_5" (890799616_5) 00:42:33.850 --> 00:42:45.350 Correct like and that's why the like people are talking about aius, it got pulled out because we didn't get consents on it but we do need to spend time on it. But I it's certainly not the same thing is the point, yeah. So.
409 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:42:45.350 --> 00:42:46.850 So I.
410 "TRN6-29-BANFF/speaker_4" (890799616_4) 00:42:46.850 --> 00:42:56.090 To Timid Robot's .1 way, and I'm not wedded at this, and maybe maybe I haven't even thought it through fully, but yeah, the, the.
411 "TRN6-29-BANFF/speaker_1" (890799616_1) 00:42:56.090 --> 00:42:57.090 Deb.
412 "TRN6-29-BANFF/speaker_4" (890799616_4) 00:42:57.090 --> 00:43:19.340 Defining search applications, you could just say use of one, one or more assets, two direct users to the location where those assets were retrieved, and because search application is so broad, like what, what's the what what work is search application doing there? Maybe I'm missing something, but it's so broad that it's basically just any application.
413 "TRN6-29-BANFF/speaker_5" (890799616_5) 00:43:19.340 --> 00:43:22.090 You're talking about the definition in section two at this point?
414 "TRN6-29-BANFF/speaker_4" (890799616_4) 00:43:22.090 --> 00:43:42.090 I'm talking about the 1st line that's kind of above where it says this category only applies? Like if you go up before the red, is that, is that yellow part still in there? I think it is, right? Yeah. So that's, that's what I'm saying. You could modify that line to just say the use of one or more assets to direct users to the location.
415 "TRN6-29-BANFF/speaker_4" (890799616_4) 00:43:42.590 --> 00:43:44.590 Of those assets or whatever.
416 "TRN6-29-BANFF/speaker_6" (890799616_6) 00:43:44.590 --> 00:43:45.090 With the.
417 "TRN6-29-BANFF/speaker_4" (890799616_4) 00:43:45.590 --> 00:43:47.090 Can't quite read how the red line.
418 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:43:47.090 --> 00:44:07.080 But I I think that that what I heard was that that could be interpreted to significantly broaden the application to this. So e.g., if you had a chat application on your, you know, laptop, that, you know, when it displays a link gave you the option to summarize that link, would this be a control that applies to that now?
419 "TRN6-29-BANFF/speaker_4" (890799616_4) 00:44:07.080 --> 00:44:08.330 I mean what.
420 "TRN6-29-BANFF/speaker_6" (890799616_6) 00:44:08.330 --> 00:44:08.830 One.
421 "TRN6-29-BANFF/speaker_4" (890799616_4) 00:44:08.830 --> 00:44:11.830 What work is the search application doing to not.
422 "TRN6-29-BANFF/speaker_2" (890799616_2) 00:44:11.830 --> 00:44:12.330 Like that.
423 "TRN6-29-BANFF/speaker_4" (890799616_4) 00:44:12.330 --> 00:44:14.330 As a practical matter right now.
424 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:44:14.330 --> 00:44:25.830 In our current definition? Yeah well from all of our current discussions I think our intent is to make this about internet search engines. We can do a better job of reflecting that I think.
425 "TRN6-29-BANFF/speaker_4" (890799616_4) 00:44:25.830 --> 00:44:28.830 Maybe maybe the search application definition that needs refined.
426 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:44:28.830 --> 00:44:31.330 I I personally I tend to agree with.
427 "TRN6-29-BANFF/speaker_6" (890799616_6) 00:44:31.330 --> 00:44:36.830 Do we need to include a mention of queries?
428 "TRN6-29-BANFF/speaker_2" (890799616_2) 00:44:36.830 --> 00:44:37.330 Well.
429 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:44:37.330 --> 00:44:45.830 Well, let's go brainstorm about it. My suggestion would be something like product purpose because that's a good distinguished between this another secondary purposes.
430 "TRN6-29-BANFF/speaker_2" (890799616_2) 00:44:45.830 --> 00:44:46.330 Sorry mark.
431 "TRN6-29-BANFF/speaker_1" (890799616_1) 00:44:46.330 --> 00:44:47.330 I didn't really care either.
432 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:44:47.330 --> 00:44:48.830 Sorry? You repeated.
433 "TRN6-29-BANFF/speaker_4" (890799616_4) 00:44:48.830 --> 00:44:49.330 I really hear you.
434 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:44:49.330 --> 00:45:07.090 My my personal suggestion would be to concentrate on, on a definition that includes includes a phrase like primary purpose because that's a good way to distinguish between other applications that happen to follow links. Get it, I think.
435 "TRN6-29-BANFF/speaker_1" (890799616_1) 00:45:07.090 --> 00:45:09.590 One thing, one piece of work that.
436 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:45:09.590 --> 00:45:10.590 The search.
437 "TRN6-29-BANFF/speaker_1" (890799616_1) 00:45:10.590 --> 00:45:11.590 The application is doing and.
438 "TRN6-29-BANFF/speaker_1" (890799616_1) 00:45:12.590 --> 00:45:14.590 Is making space for the ranking.
439 "TRN6-29-BANFF/speaker_2" (890799616_2) 00:45:14.590 --> 00:45:18.360 Did.
440 "TRN6-29-BANFF/speaker_1" (890799616_1) 00:45:18.360 --> 00:45:22.360 Right, you.
441 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:45:22.360 --> 00:45:23.610 Yeah went into.
442 "TRN6-29-BANFF/speaker_1" (890799616_1) 00:45:23.610 --> 00:45:25.610 You consider and discard assets.
443 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:45:25.610 --> 00:45:26.610 In the search application.
444 "TRN6-29-BANFF/speaker_1" (890799616_1) 00:45:27.610 --> 00:45:29.110 You know or if you rank them such that.
445 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:45:29.110 --> 00:45:30.110 But.
446 "TRN6-29-BANFF/speaker_1" (890799616_1) 00:45:30.110 --> 00:45:32.110 Which means not linking to that.
447 "TRN6-29-BANFF/speaker_6" (890799616_6) 00:45:32.110 --> 00:45:33.110 And so just, just be.
448 "TRN6-29-BANFF/speaker_1" (890799616_1) 00:45:33.110 --> 00:45:35.110 Making this bit say.
449 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:45:35.110 --> 00:45:37.610 It has to direct it to the asset to qualify for the.
450 "TRN6-29-BANFF/speaker_1" (890799616_1) 00:45:37.610 --> 00:45:39.760 Exception leaves all of that out.
451 "TRN6-29-BANFF/speaker_6" (890799616_6) 00:45:48.479 --> 00:46:08.850 So I'm just gonna point out that it's very hard as an editor to manage this because we've jumped from one issue to another without making any resolution on the other items and this sort of jumping around makes it very difficult for them for me to do anything.
452 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:46:08.850 --> 00:46:23.010 I I I think I would like us to get into a state where this text is in the editor's draft and we even perhaps refer to the open issues in different parts of the editor's draft. I know that that's something we don't love doing.
453 "TRN6-29-BANFF/speaker_1" (890799616_1) 00:46:23.010 --> 00:46:23.510 But.
454 "TRN6-29-BANFF/speaker_6" (890799616_6) 00:46:23.510 --> 00:46:24.510 Really don't like doing that.
455 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:46:24.510 --> 00:46:30.500 It might, it might help people.
456 "TRN6-29-BANFF/speaker_6" (890799616_6) 00:46:30.500 --> 00:46:32.750 I think would help more than that would help.
457 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:46:32.750 --> 00:46:34.300 You'll see.
458 "TRN6-29-BANFF/speaker_5" (890799616_5) 00:46:34.300 --> 00:46:36.479 Give me robot?
459 "TRN6-29-BANFF/speaker_1" (890799616_1) 00:46:40.159 --> 00:47:06.960 To my earlier point about kind of this doesn't matter if it's a primary search application or not, the the last paragraph restricts any kind of AI building or training or whatnot to search. And so like like someone wouldn't be able to.
460 "TRN6-29-BANFF/speaker_1" (890799616_1) 00:47:06.960 --> 00:47:23.370 Like that excludes like a generic kind of like chat interface that provides results, right? Like it's, you know, kind of it to to address people's kind of concerns about rag.
461 "TRN6-29-BANFF/speaker_1" (890799616_1) 00:47:23.370 --> 00:47:49.409 Like the rags that are, you know, kind of, updating models or creating like hybrid models et cetera, from the, the, the source assets, they would only qualify for this preference if their primary thing was, you know, if they were only doing search because of that last paragraph.
462 "TRN6-29-BANFF/speaker_5" (890799616_5) 00:47:49.909 --> 00:47:56.909 Right, I think it was Mark's proposal as well like used to, but I think Mark, you intended to do it in section two right in the definition of the search application.
463 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:47:56.909 --> 00:47:58.409 So, yeah.
464 "TRN6-29-BANFF/speaker_6" (890799616_6) 00:47:58.409 --> 00:48:01.280 Yeah.
465 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:48:01.280 --> 00:48:04.530 I, I feel like we're getting to diminishing returns here.
466 "TRN6-29-BANFF/speaker_5" (890799616_5) 00:48:06.030 --> 00:48:09.959 Kevin and Aaron, if you want to make a quick one.
467 "TRN6-29-BANFF/speaker_3" (890799616_3) 00:48:09.959 --> 00:48:13.209 Well, I was gonna compound the problem by pointing out something else, though.
468 "TRN6-29-BANFF/speaker_5" (890799616_5) 00:48:13.209 --> 00:48:14.209 Well yeah.
469 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:48:14.209 --> 00:48:21.709 If it's, if it's substantive, we we it'd be good to talk about it in this room. If it's editorial, you know, or or how we're worrying things. Yeah I.
470 "TRN6-29-BANFF/speaker_3" (890799616_3) 00:48:21.709 --> 00:48:32.709 I don't think it's it's just a bit like language I don't think we could solve it with language structure, but can you scroll up to defines what this.
471 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:48:32.709 --> 00:48:35.709 Sorry I didn't hear a whole bit of what you said cause there was a lot of coughing.
472 "TRN6-29-BANFF/speaker_3" (890799616_3) 00:48:35.709 --> 00:48:40.209 Can you scroll up to the sentence that says what this category is before we then get to the conditions on.
473 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:48:40.209 --> 00:48:44.340 This one right here? Yeah.
474 "TRN6-29-BANFF/speaker_3" (890799616_3) 00:48:46.360 --> 00:49:16.469 There we go. Yeah. So, what this is saying before we get to any other condition or statement about the category is that the search application must direct users to the location of the asset, which means that if the asset never ranks high enough to be displayed to a user, then it cannot be indexed at all. And you cannot index something unless you know already that you will be showing it, which fundamentally defeats the whole ability to use assets.
475 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:49:16.469 --> 00:49:20.469 So, so perhaps an application that might direct users to the location or.
476 "TRN6-29-BANFF/speaker_6" (890799616_6) 00:49:20.469 --> 00:49:28.969 I have text in the chat if anyone wants to read that that I think addresses these concerns rather than just come trying to come up with adversarial readings.
477 "TRN6-29-BANFF/speaker_3" (890799616_3) 00:49:28.969 --> 00:49:34.469 I think, is that language, and I'm sorry I'm trying to follow both the talking in the chat. Yeah.
478 "TRN6-29-BANFF/speaker_6" (890799616_6) 00:49:34.469 --> 00:49:35.469 I know it's difficult.
479 "TRN6-29-BANFF/speaker_3" (890799616_3) 00:49:35.469 --> 00:49:43.969 But is, is that language the 1st introduction sentence that defines the overall category or is it in that later paragraphs?
480 "TRN6-29-BANFF/speaker_6" (890799616_6) 00:49:43.969 --> 00:50:02.469 No, NO. So I put it, the very last message in the chat at this moment is a suggestion for an alternative replacement for that 1st sentence that I believe the concerns that you have, which is use of an asset in applications, the primary purpose of that application is to take you.
481 "TRN6-29-BANFF/speaker_1" (890799616_1) 00:50:02.469 --> 00:50:02.969 It's queries.
482 "TRN6-29-BANFF/speaker_6" (890799616_6) 00:50:02.969 --> 00:50:10.000 Select assets appropriate to those queries and direct users to the location of those assets.
483 "TRN6-29-BANFF/speaker_3" (890799616_3) 00:50:10.000 --> 00:50:34.129 I think this is an improvement, but does every search application have to start with a user query? Can the user say, like, you know me, you know when I come to this, I want links to interesting articles and I'm gonna do that five times a day without each time saying, give me a link to another.
484 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:50:34.129 --> 00:50:34.629 And I think.
485 "TRN6-29-BANFF/speaker_6" (890799616_6) 00:50:34.629 --> 00:50:49.629 Think, I think you can, you can frame that as a search query. I think that's, that's taking a pretty fine breeding of this one. I think that's a, if you if you want to look for, if you want to look for problems in this language. You're always gonna find them.
486 "TRN6-29-BANFF/speaker_3" (890799616_3) 00:50:49.629 --> 00:50:50.129 We're here for it.
487 "TRN6-29-BANFF/speaker_6" (890799616_6) 00:50:50.129 --> 00:50:52.629 Right? But that seems like stretching.
488 "TRN6-29-BANFF/speaker_1" (890799616_1) 00:50:52.629 --> 00:50:53.129 I.
489 "TRN6-29-BANFF/speaker_3" (890799616_3) 00:50:53.129 --> 00:51:11.199 We don't because we we distinguish between things that are query based and query less and without focusing too much on the products of any one particular company, like this would mean that things can't be in Google News. Because the homepage of Google News is not a query, it is, here's what's in the news.
490 "TRN6-29-BANFF/speaker_1" (890799616_1) 00:51:11.199 --> 00:51:11.949 So.
491 "TRN6-29-BANFF/speaker_6" (890799616_6) 00:51:12.449 --> 00:51:18.449 You, you cannot possibly frame something like Google News as having an implicit query.
492 "TRN6-29-BANFF/speaker_3" (890799616_3) 00:51:18.449 --> 00:51:25.949 I mean if we wanna, if we wanna say that there are the implicit queries where the user has not actually like done anything.
493 "TRN6-29-BANFF/speaker_3" (890799616_3) 00:51:26.949 --> 00:51:34.370 I feel like we're kind of lowing like then why why have a reference to query at all?
494 "TRN6-29-BANFF/speaker_2" (890799616_2) 00:51:34.370 --> 00:51:36.870 What if i wish they got the word user?
495 "TRN6-29-BANFF/speaker_2" (890799616_2) 00:51:37.370 --> 00:51:41.870 I mean to me that your issue was around users, not around queries, it sounds.
496 "TRN6-29-BANFF/speaker_3" (890799616_3) 00:51:41.870 --> 00:51:51.540 No, I think these applications have users pretty comfortable with the limitation that the application should be serving like just human beings rather than.
497 "TRN6-29-BANFF/speaker_3" (890799616_3) 00:51:51.540 --> 00:52:11.540 Well, and and the intermediate processing is taken care of by like a later explanation, but I I do think that like a query based as the defining characteristic is not entirely the point. And I I also feel like the goal here is not to say this is what a search engine has been for however many years and.
498 "TRN6-29-BANFF/speaker_3" (890799616_3) 00:52:11.540 --> 00:52:35.870 So we will only permit the things that look exactly like that. I think the goal is to say that when you are using the asset in a way that helps people find the asset and doesn't do other stuff with it that we think is inappropriate, that you get to do that, whether or not you're a chat application or a search engine or the next thing that comes along that like we aren't predicting yet.
499 "TRN6-29-BANFF/speaker_6" (890799616_6) 00:52:35.870 --> 00:52:40.669 Yeah, ok. I see your point. Give me, give me something I can use that.
500 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:52:40.669 --> 00:52:41.169 I just.
501 "TRN6-29-BANFF/speaker_3" (890799616_3) 00:52:44.699 --> 00:52:57.110 I mean I like, I I think there's a way to do this, which is the draft I submitted. And I'm, I don't know and like another way but I can think, I can try to come up with another one.
502 "TRN6-29-BANFF/speaker_6" (890799616_6) 00:52:57.110 --> 00:53:03.760 Yeah, I I think you have a host of other problems with that text. Yeah. Yeah.
503 "TRN6-29-BANFF/speaker_5" (890799616_5) 00:53:03.760 --> 00:53:17.340 Martin, did you want to talk about a different point? Like I think we should just go through and hear from everybody and then we come back to it. Glenn.
504 "TRN6-29-BANFF/speaker_2" (890799616_2) 00:53:17.340 --> 00:53:45.199 Then? Oh, sorry. I thought it was far lower in the queue but I wanted to come into this conversation. What I pull that query and we change this effectively any content that is dynamically generated for presentation would fall into this category now. That would include electronic program guides when you go to a streaming service and it says, here's movies you can watch.
505 "TRN6-29-BANFF/speaker_2" (890799616_2) 00:53:46.379 --> 00:54:05.960 Are we making it so expansive that it becomes, you know, it's almost the category of it done with the computer? Because it seems like I I think where we were with search was we, we had a classic use case we were addressing. And I think your point is well taken that there are future use cases that we.
506 "TRN6-29-BANFF/speaker_2" (890799616_2) 00:54:05.960 --> 00:54:39.139 We don't understand that may may, we may want to support, but I'm wondering if the path isn't to have classic search that we sort of understand today defined, and then we have a separate category for those other things that are not as well defined because by their very nature of being undefined, we don't know their boundaries and conditions that we want to talk about, but for search classic search that is, I think we do. Does that help?
507 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:54:39.139 --> 00:54:41.389 I think.
508 "TRN6-29-BANFF/speaker_3" (890799616_3) 00:54:41.389 --> 00:54:57.889 I don't think there's a problem with being over expansive if all we're talking about are things like a program guide that has the title of a show and then like tells you where and when to watch it. I don't think that's what anybody's concerned about trying to exclude from.
509 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:54:57.889 --> 00:55:01.250 This category. So we've got a queue, so let's.
510 "TRN6-29-BANFF/speaker_2" (890799616_2) 00:55:01.250 --> 00:55:24.330 So I I do worry that it is, it's a feedback on that. I worry that EPGs today class EPGs don't have things, but expanded EPGs could. And so I I I would be more comfortable keeping classic search in its own category because I I understand what that is. Expanded use case search, not comfortable pulling in like this.
511 "TRN6-29-BANFF/speaker_3" (890799616_3) 00:55:24.330 --> 00:55:34.959 But isn't isn't the like expanded, sorry I didn't catch your acronym, Expanded information captured by the other conditions in the category?
512 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:55:34.959 --> 00:55:35.709 Yeah.
513 "TRN6-29-BANFF/speaker_2" (890799616_2) 00:55:35.709 --> 00:55:44.670 The answer is I don't know because I haven't evaluated all the boundary conditions with that expanded state. Again, with classic search, it's a well.
514 "TRN6-29-BANFF/speaker_2" (890799616_2) 00:55:44.670 --> 00:56:00.709 Well discussed topic space, and so I know the, the, the, the concerns and the issues that have to be dealt with, and so I can evaluate this against that with the other expanded skills, I can't yet without extensive thinkings for all the cases. That's why i'm uncomfortable with it.
515 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:56:00.709 --> 00:56:04.790 So so but but that sounds like you've got homework now. Go and think through.
516 "TRN6-29-BANFF/speaker_2" (890799616_2) 00:56:04.790 --> 00:56:29.780 I know I disagree. I I think we all have we all suffer from that problem that the we know what classic search looks and feels like and partly drew here. If we expand this, then we are opening up a can of rooms essentially, like I said the extreme version of this is done with a computer and and that opens up a lot of cases which we have never considered or discussed as a community or as a room.
517 "TRN6-29-BANFF/speaker_5" (890799616_5) 00:56:29.780 --> 00:56:32.030 Okay, thank you, Nate.
518 "TRN6-29-BANFF/speaker_4" (890799616_4) 00:56:32.530 --> 00:56:34.530 I'd put a suggestion in the chat.
519 "TRN6-29-BANFF/speaker_4" (890799616_4) 00:56:35.030 --> 00:56:37.530 That just basically took out the query part.
520 "TRN6-29-BANFF/speaker_4" (890799616_4) 00:56:38.030 --> 00:56:47.030 I I did that quickly. I wonder if anyone has thoughts on that, including any publishers or anything because I do take the point about like Google discover, for instance.
521 "TRN6-29-BANFF/speaker_2" (890799616_2) 00:56:47.030 --> 00:56:47.530 But.
522 "TRN6-29-BANFF/speaker_4" (890799616_4) 00:56:47.530 --> 00:56:49.030 So I do take that point.
523 "TRN6-29-BANFF/speaker_6" (890799616_6) 00:56:49.030 --> 00:56:55.530 Yeah, and news and a bunch of other things that implicitly do these sorts of things. I mean even like the Facebook newsfeed.
524 "TRN6-29-BANFF/speaker_2" (890799616_2) 00:56:55.530 --> 00:56:56.939 No?
525 "TRN6-29-BANFF/speaker_6" (890799616_6) 00:56:56.939 --> 00:56:58.879 Right.
526 "TRN6-29-BANFF/speaker_5" (890799616_5) 00:56:58.879 --> 00:57:00.129 Okay, once.
527 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:57:00.129 --> 00:57:03.629 Yeah, I mean Glenn triggered me out something. So.
528 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:57:04.629 --> 00:57:09.129 You trigger me all sorts of things buddy, but when I go to, e.g., Netflix.
529 "TRN6-29-BANFF/speaker_6" (890799616_6) 00:57:09.129 --> 00:57:09.629 I don't see.
530 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:57:09.629 --> 00:57:10.629 And it should.
531 "TRN6-29-BANFF/speaker_2" (890799616_2) 00:57:10.629 --> 00:57:11.129 Just.
532 "TRN6-29-BANFF/speaker_5" (890799616_5) 00:57:12.129 --> 00:57:13.629 Because it is learned.
533 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:57:13.629 --> 00:57:24.120 My preferences over time. Do you do that as a query or not? I mean it is over time, I can provide it more information. It has learnt my preferences.
534 "TRN6-29-BANFF/speaker_2" (890799616_2) 00:57:24.120 --> 00:57:41.870 Or a map application that you're walking down the street and it says, by the way, Warren I know you like capecinos. Here's a great capecinos shop. Is that a query? I don't know that that to me that's dynamically created content based upon knowledge of the the entity I'm interacting with. It's not a search in many ways.
535 "TRN6-29-BANFF/speaker_5" (890799616_5) 00:57:41.870 --> 00:57:42.370 Possibly.
536 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:57:42.370 --> 00:57:53.370 Not, but it is something that I would like to continue to use and to get better. And so it's not, it's definitely not leaderhoods, but it is.
537 "TRN6-29-BANFF/speaker_3" (890799616_3) 00:57:53.370 --> 00:57:53.870 Something.
538 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:57:53.870 --> 00:57:55.870 Which I think is getting.
539 "TRN6-29-BANFF/speaker_2" (890799616_2) 00:57:55.870 --> 00:57:56.370 Is.
540 "TRN6-29-BANFF/speaker_7" (890799616_7) 00:57:56.370 --> 00:57:58.370 Of value to the person.
541 "TRN6-29-BANFF/speaker_2" (890799616_2) 00:57:58.370 --> 00:58:12.370 Use it. Let me be clear, I'm not trying to cut it out. What I'm identifying is that I think what we have is a separate case that we should capture as a, as a preference that you can express.
542 "TRN6-29-BANFF/speaker_5" (890799616_5) 00:58:12.370 --> 00:58:15.780 Oh, thank you. Thanks. Tim and Robot.
543 "TRN6-29-BANFF/speaker_1" (890799616_1) 00:58:15.780 --> 00:58:45.060 Tim Edrova, I'm confused about what we're trying to solve for. I thought we were trying to solve ensuring users are brought to the original content versus extracting that content and keeping users away from it. And, that we're trying to avoid misrepresenting content, you know, that a news headline gets changed and the whole thing gets changed due to a poor board choice.
544 "TRN6-29-BANFF/speaker_1" (890799616_1) 00:58:45.060 --> 00:59:00.540 I didn't think we were trying to create like a little early ox bubble for ourself, like to, like, like preserve a a an experience in time.
545 "TRN6-29-BANFF/speaker_1" (890799616_1) 00:59:00.540 --> 00:59:18.570 But to address these issues with with content and assets or whatever we want to call them. I I still fail to see why if the content is being respected in a way that this category requires.
546 "TRN6-29-BANFF/speaker_1" (890799616_1) 00:59:18.570 --> 00:59:37.090 Why there's any issue, kind of what the context of that use is, right? Like if it's bringing users to the original content and it's not misrepresenting that content, like we, if that's been solved, aren't we happy? Yeah, thank you.
547 "TRN6-29-BANFF/speaker_5" (890799616_5) 00:59:37.590 --> 00:59:38.090 Added.
548 "TRN6-29-BANFF/speaker_3" (890799616_3) 00:59:38.090 --> 00:59:46.389 Oh, I think pretty similar to what was saying, but also I think when we.
549 "TRN6-29-BANFF/speaker_5" (890799616_5) 00:59:46.389 --> 00:59:46.889 Thing to use.
550 "TRN6-29-BANFF/speaker_3" (890799616_3) 00:59:46.889 --> 01:00:02.389 Stability of these preferences out in the wild, people who want their content to be findable, the search content that includes the rankings and the, here's other content related to what you might want is in that pocket, and so if we're making choices where we expect.
551 "TRN6-29-BANFF/speaker_3" (890799616_3) 01:00:02.889 --> 01:00:13.979 People to pick findability as a bucket, I think it makes sense to include more than just, you know, 2010 search.
552 "TRN6-29-BANFF/speaker_6" (890799616_6) 01:00:13.979 --> 01:00:18.229 Do do you think that nate's definition fails on that measure?
553 "TRN6-29-BANFF/speaker_3" (890799616_3) 01:00:18.229 --> 01:00:20.729 So I think, I think this is.
554 "TRN6-29-BANFF/speaker_7" (890799616_7) 01:00:20.729 --> 01:00:21.229 It's.
555 "TRN6-29-BANFF/speaker_6" (890799616_6) 01:00:21.229 --> 01:00:23.229 Yeah, ok, good. That's.
556 "TRN6-29-BANFF/speaker_7" (890799616_7) 01:00:23.229 --> 01:00:23.729 This.
557 "TRN6-29-BANFF/speaker_6" (890799616_6) 01:00:23.729 --> 01:00:26.229 A lot of negativity floating around and that's what I'm.
558 "TRN6-29-BANFF/speaker_1" (890799616_1) 01:00:26.229 --> 01:00:27.360 These are the.
559 "TRN6-29-BANFF/speaker_7" (890799616_7) 01:00:27.360 --> 01:00:28.610 Eight.
560 "TRN6-29-BANFF/speaker_6" (890799616_6) 01:00:28.610 --> 01:00:30.610 I think it's good. Yeah, I think it's I think it's.
561 "TRN6-29-BANFF/speaker_7" (890799616_7) 01:00:30.610 --> 01:00:31.610 I'm gonna take up.
562 "TRN6-29-BANFF/speaker_3" (890799616_3) 01:00:31.610 --> 01:00:32.110 Primary but.
563 "TRN6-29-BANFF/speaker_7" (890799616_7) 01:00:32.110 --> 01:00:33.110 Do you think it's good?
564 "TRN6-29-BANFF/speaker_5" (890799616_5) 01:00:33.110 --> 01:00:35.814 Thanks, Brad?
565 "Brad Silver" (352249088) 01:00:35.814 --> 01:01:10.019 Yeah, so I I I get the concern and maybe some of the confusion around why publishers might not necessarily want an expansive definition. I think there's a couple of things. The 1st thing is that if we're thinking about this in the sort of classic search context where there is something sort of like search plus, you know, that you're, there's a search engine and then the search engine also recommends, content for you in various ways. So there's the sort of pool and then there's the.
566 "TRN6-29-BANFF/speaker_6" (890799616_6) 01:01:10.019 --> 01:01:10.769 Push.
567 "TRN6-29-BANFF/speaker_2" (890799616_2) 01:01:11.269 --> 01:01:11.769 Yeah.
568 "TRN6-29-BANFF/speaker_4" (890799616_4) 01:01:12.269 --> 01:01:13.461 It it it really is.
569 "Brad Silver" (352249088) 01:01:13.461 --> 01:01:32.060 It's gonna depend on the party doing that, whether or not there is actual benefit for the publisher. It depends how it's presented. There are some news aggregation services that may fall into this category that are not particularly helpful for publishers. There is the question of whether.
570 "Brad Silver" (352249088) 01:01:32.060 --> 01:01:53.270 The publishers. Ultimately it's in their interest to be pushing more towards building direct relationships with end users as opposed to kind of being constantly sucked into aggregation or intermediary type relationships. And I think that whilst some might see that as a forthooting inclination, I think it also is kind of a longer term self preservation inclination.
571 "Brad Silver" (352249088) 01:01:53.270 --> 01:02:13.270 In a world where, you know, kind of being reliant on intermediaries for, for, for your survival hasn't necessarily served the news industry that, that well. So it's a bit more complicated than just saying, well, why wouldn't publishers just want, you know, their stuff to appear in every sort of way?
572 "Brad Silver" (352249088) 01:02:13.270 --> 01:02:45.710 Which people could then find it's find find the path back to them. I think yes generally, but there are going to be a lot of instances where they don't. So I don't know if that's I'm not saying that I don't, that I disagree with Nate's suggestion, I think it was it's responsive to to to the question that's raised, but I do think it needs more thought as to how expensive we want this category to actually be to solve for what we wanted to solve for.
573 "TRN6-29-BANFF/speaker_5" (890799616_5) 01:02:47.520 --> 01:02:50.770 Thanks. Glenn, did you already say your part?
574 "TRN6-29-BANFF/speaker_2" (890799616_2) 01:02:50.770 --> 01:02:54.820 I, I I I did what I took my hand down just a second ago.
575 "TRN6-29-BANFF/speaker_6" (890799616_6) 01:02:54.820 --> 01:02:57.070 Yeah, can we talk about this primary purpose?
576 "TRN6-29-BANFF/speaker_7" (890799616_7) 01:02:57.070 --> 01:03:24.120 Yeah I that's what I'd like to focus on is, is it's I I I'm hearing a fair amount of approach support or at least willingness to consider something like what Nate suggested. I put up because I I heard a little pushback on primary purpose, so I did another version in the chat which is without the primary purpose qualifier. Can we talk through primary purpose for a few minutes and see if we need that or not?
577 "TRN6-29-BANFF/speaker_5" (890799616_5) 01:03:24.120 --> 01:03:25.370 Nick is it related?
578 "TRN6-29-BANFF/speaker_2" (890799616_2) 01:03:25.370 --> 01:03:26.870 Yeah, it's related.
579 "TRN6-29-BANFF/speaker_1" (890799616_1) 01:03:26.870 --> 01:04:00.870 So I I'm happy with starting with what Nate is presenting. I I think the risk here and where we're risking overlapping is, is, that classic search with queries is kind of expanding into this chat window world and we, we're risking the rag world where you're starting with the chat interface that can act like search, and there's there's potentially an overlap there, and that's the boundary I think we need to draw up with this text is from search that.
580 "TRN6-29-BANFF/speaker_2" (890799616_2) 01:04:00.870 --> 01:04:01.870 Also have.
581 "TRN6-29-BANFF/speaker_1" (890799616_1) 01:04:01.870 --> 01:04:02.370 As capability.
582 "TRN6-29-BANFF/speaker_2" (890799616_2) 01:04:02.370 --> 01:04:03.370 Abilities of short.
583 "TRN6-29-BANFF/speaker_1" (890799616_1) 01:04:03.370 --> 01:04:20.699 Summarization and whatnot to full kind of chatbot views that can do search like functionalities and so I think people will have different preferences for these two and that's where I think we should be careful drawing the line. I don't unfortunately have text.
584 "TRN6-29-BANFF/speaker_5" (890799616_5) 01:04:20.699 --> 01:04:21.949 Nate?
585 "TRN6-29-BANFF/speaker_4" (890799616_4) 01:04:21.949 --> 01:04:37.230 So 1st of all I'll say I'm sensitive to all the concerns Brad raised and and I do want to think about those more. I put this forward as a good faith effort to try to just find, find the path forward. I do think primary purpose is important here though, because absent if we don't have primary purpose.
586 "TRN6-29-BANFF/speaker_4" (890799616_4) 01:04:37.230 --> 01:04:57.230 Then all it says is an application that's the use of an asset in an application and then that's qualified by saying that selects and directs users to the location of those assets. All it means is that somewhere in that application, there has to be a function of that application that performs this sort of search function, but that could be like.
587 "TRN6-29-BANFF/speaker_4" (890799616_4) 01:04:57.230 --> 01:05:23.929 1 % of the application and then 99 % of the other, and then we're using that asset in that because the applies to the application, not to the act of selecting assets and directing users to location of those assets, if that makes sense, it effectively gets at this concern, you're saying, so you need that primary purpose to be like that's really what this application is about, not just some random feature that was tacked on in order to qualify for this, this category.
588 "TRN6-29-BANFF/speaker_6" (890799616_6) 01:05:23.929 --> 01:05:24.929 Yeah.
589 "TRN6-29-BANFF/speaker_5" (890799616_5) 01:05:24.929 --> 01:05:27.179 And the middlebot.
590 "TRN6-29-BANFF/speaker_1" (890799616_1) 01:05:27.179 --> 01:05:49.169 Name is Timo and Robot. I don't think we can identify a commercial application whose primary purpose is search. I think the primary purpose is, is to add impressions. The primary purpose is surveillance appalism.
591 "TRN6-29-BANFF/speaker_7" (890799616_7) 01:05:49.169 --> 01:05:49.669 No, but.
592 "TRN6-29-BANFF/speaker_1" (890799616_1) 01:05:49.669 --> 01:06:13.380 Seriously, right? Like, I think this idea of like locking it into like I I think we're, I worry that we are not grounded in in in reality here and and again, I think that when it comes to things like.
593 "TRN6-29-BANFF/speaker_1" (890799616_1) 01:06:13.380 --> 01:06:30.060 You know, does this quo, you know, does a chat program qualify for this category? I think that's a really important question, not because we need to discern between different applications, but we need to protect.
594 "TRN6-29-BANFF/speaker_1" (890799616_1) 01:06:30.060 --> 01:06:52.989 The use of the content and that that, that is the goal is protecting kind of the, the, the, the agency of the content holder and how that content is, is used both these to be visits and, and, and representation of that content. Yeah. Thank you.
595 "TRN6-29-BANFF/speaker_2" (890799616_2) 01:06:52.989 --> 01:06:54.760 Aaron?
596 "TRN6-29-BANFF/speaker_3" (890799616_3) 01:06:54.760 --> 01:07:11.100 Yeah, just to respond to the concern Nate raised, I I think the like the rest of the definition of the category is to address that issue. So like the introductory sentence can be broader.
597 "TRN6-29-BANFF/speaker_3" (890799616_3) 01:07:11.100 --> 01:07:29.540 Which then says, you know, the use of the asset is conditioned on the following things. And so that's why I feel like we don't need to be very restrictive about the nature of the application because the nature of the use is more specified.
598 "TRN6-29-BANFF/speaker_2" (890799616_2) 01:07:30.290 --> 01:07:31.290 Chris?
599 "TRN6-29-BANFF/speaker_2" (890799616_2) 01:07:32.290 --> 01:07:46.909 I was just gonna say that I understand where he's trying to go with the primary purpose, but I think given what Tim and Robot said and which I agree with this, there's probably another way to say that or a better way to say because the primary purpose could be, you know.
600 "TRN6-29-BANFF/speaker_6" (890799616_6) 01:07:46.909 --> 01:07:47.409 The number I mean.
601 "TRN6-29-BANFF/speaker_2" (890799616_2) 01:07:47.409 --> 01:07:49.909 It's supposed to be primary, but it's there's any number.
602 "TRN6-29-BANFF/speaker_6" (890799616_6) 01:07:49.909 --> 01:07:50.409 World.
603 "TRN6-29-BANFF/speaker_2" (890799616_2) 01:07:50.409 --> 01:07:52.540 Things that, you know.
604 "TRN6-29-BANFF/speaker_1" (890799616_1) 01:07:52.540 --> 01:07:54.290 This could be used for, so.
605 "TRN6-29-BANFF/speaker_2" (890799616_2) 01:07:54.290 --> 01:08:00.790 Yeah, if there's I don't know if we're moving it as an answer, but maybe phrasing it another way or finding what the real.
606 "TRN6-29-BANFF/speaker_7" (890799616_7) 01:08:00.790 --> 01:08:06.879 Or or who the primary purpose is for. Yeah. Yeah.
607 "TRN6-29-BANFF/speaker_5" (890799616_5) 01:08:06.879 --> 01:08:34.129 Yeah, I got in line to say, I think about your thing, your point is like extremely cynical in a way, so my the way I thought about it is that somebody who consumes his preference is probably gonna be self classifying themselves as the primary purpose being searched, right? Like it's at least how I thought about it, which is like the other angle into this. So assuming that's done we are going into, I like the formulation with primary purpose in it, right? Like, but.
608 "TRN6-29-BANFF/speaker_7" (890799616_7) 01:08:34.129 --> 01:08:34.629 I.
609 "TRN6-29-BANFF/speaker_5" (890799616_5) 01:08:34.629 --> 01:08:42.640 I hear you, but I think it's probably outside our scope to kind of talk about it that way. Annik.
610 "TRN6-29-BANFF/speaker_7" (890799616_7) 01:08:42.640 --> 01:08:45.259 Yeah, I.
611 "TRN6-29-BANFF/speaker_4" (890799616_4) 01:08:45.259 --> 01:08:59.569 I am like I'm sensitive both Aaron and Tim and Robot's Point I think that maybe primary, maybe it's better to apply the primary purpose to be used because that's the thing that we care about and not to the application.
612 "TRN6-29-BANFF/speaker_4" (890799616_4) 01:09:00.069 --> 01:09:06.569 To address, maybe that would address Timmy's point now, as to whether or not the later on.
613 "TRN6-29-BANFF/speaker_1" (890799616_1) 01:09:06.569 --> 01:09:07.569 Yeah.
614 "TRN6-29-BANFF/speaker_6" (890799616_6) 01:09:07.569 --> 01:09:08.069 The.
615 "TRN6-29-BANFF/speaker_4" (890799616_4) 01:09:08.069 --> 01:09:17.069 Concern is that the 1st, if, if the 1st I I think that in any sort of when you're trying to retroactively interpret any sort of legislation or anything like the 1st.
616 "TRN6-29-BANFF/speaker_4" (890799616_4) 01:09:17.569 --> 01:09:31.569 Line of text is important, does a lot of work, and if the 1st line is everything, and then we're gonna narrow everything, I I think that's different than start like I just think it should be a little bit more defined than that and I.
617 "TRN6-29-BANFF/speaker_1" (890799616_1) 01:09:31.569 --> 01:09:32.569 Is that.
618 "TRN6-29-BANFF/speaker_4" (890799616_4) 01:09:32.569 --> 01:09:44.020 Maybe that's not, you know, maybe it could just, I don't know, I will think about that, but I I I I think that if it's you can't just start off by saying literally everything and then let's narrow that down.
619 "TRN6-29-BANFF/speaker_5" (890799616_5) 01:09:44.020 --> 01:09:46.859 Yeah, sure.
620 "TRN6-29-BANFF/speaker_3" (890799616_3) 01:09:46.859 --> 01:09:57.609 I, I know that there was a reference to search application before and is leveraging I know that there's and Suresh you raised this very issue that.
621 "TRN6-29-BANFF/speaker_7" (890799616_7) 01:09:57.609 --> 01:09:59.109 Search application in this.
622 "TRN6-29-BANFF/speaker_3" (890799616_3) 01:09:59.109 --> 01:10:26.330 December draft is already defined, and I know that this group can also tweak that definition, but keeping the search application here, would that like help scope this better in and if contoring and and bringing more clarity, I actually really appreciated Nathan's points as well. So I wonder if that would like help bring a bit more clarity to the definition in a way that doesn't confuse.
623 "TRN6-29-BANFF/speaker_3" (890799616_3) 01:10:26.830 --> 01:10:33.160 The industry.
624 "TRN6-29-BANFF/speaker_7" (890799616_7) 01:10:33.160 --> 01:10:35.979 So.
625 "TRN6-29-BANFF/speaker_5" (890799616_5) 01:10:35.979 --> 01:10:36.729 Chris.
626 "TRN6-29-BANFF/speaker_2" (890799616_2) 01:10:37.729 --> 01:10:46.580 I mean it also should just a loophole, right? So if you say we're primary purposes one thing is really something. That's fine.
627 "TRN6-29-BANFF/speaker_2" (890799616_2) 01:10:47.330 --> 01:10:47.830 If there's.
628 "TRN6-29-BANFF/speaker_1" (890799616_1) 01:10:47.830 --> 01:10:48.330 Another ways.
629 "TRN6-29-BANFF/speaker_2" (890799616_2) 01:10:51.720 --> 01:10:52.970 Why.
630 "TRN6-29-BANFF/speaker_7" (890799616_7) 01:10:52.970 --> 01:10:56.319 Sorry.
631 "TRN6-29-BANFF/speaker_6" (890799616_6) 01:10:56.319 --> 01:10:57.569 Where's it not second primary pur.
632 "TRN6-29-BANFF/speaker_7" (890799616_7) 01:10:57.569 --> 01:11:02.140 I don't know, you're the editor.
633 "TRN6-29-BANFF/speaker_6" (890799616_6) 01:11:02.140 --> 01:11:05.080 Because that's not what I have in my headed buffer.
634 "TRN6-29-BANFF/speaker_7" (890799616_7) 01:11:05.080 --> 01:11:07.479 I, I've.
635 "TRN6-29-BANFF/speaker_5" (890799616_5) 01:11:07.479 --> 01:11:12.490 I I think that was suggestion, right? So it's probably too new between.
636 "TRN6-29-BANFF/speaker_2" (890799616_2) 01:11:12.490 --> 01:11:14.490 Is this usually when we take a break?
637 "TRN6-29-BANFF/speaker_7" (890799616_7) 01:11:14.490 --> 01:11:21.490 Yeah, we're about to take a break. I I think we are getting to diminishing returns. I, I do think we need to get some techs out there to let people think about it.
638 "TRN6-29-BANFF/speaker_3" (890799616_3) 01:11:21.490 --> 01:11:22.990 Should put a suggestion in.
639 "TRN6-29-BANFF/speaker_7" (890799616_7) 01:11:22.990 --> 01:11:23.990 Okay.
640 "TRN6-29-BANFF/speaker_6" (890799616_6) 01:11:23.990 --> 01:11:26.990 If that may have got lost.
641 "TRN6-29-BANFF/speaker_7" (890799616_7) 01:11:26.990 --> 01:11:27.990 Is that ok.
642 "TRN6-29-BANFF/speaker_3" (890799616_3) 01:11:27.990 --> 01:11:30.490 Request? Last thing in the chat right now.
643 "TRN6-29-BANFF/speaker_7" (890799616_7) 01:11:30.490 --> 01:11:40.509 Okay let's look at that suggestion and talk about it, and then we're gonna take a break. So talk us through it.
644 "TRN6-29-BANFF/speaker_3" (890799616_3) 01:11:40.509 --> 01:12:10.209 So to address the concern that search application is the whole thing kind of out, independent of the specific like presentation of the assets, instead of talking about a search application, we would just be talking about search results and defining those results as like the links which are directing users to the result and the associated content that like kind of brings in some of this idea that there's other stuff that can be used to give context and help.
645 "TRN6-29-BANFF/speaker_6" (890799616_6) 01:12:10.709 --> 01:12:11.209 Staff.
646 "TRN6-29-BANFF/speaker_3" (890799616_3) 01:12:11.209 --> 01:12:17.939 But that is all still like in service of directing users to the location of the assets.
647 "TRN6-29-BANFF/speaker_7" (890799616_7) 01:12:17.939 --> 01:12:31.100 So, I think it depends a lot on your drafting because if if you just say use of an asset in search results, one of the readings of that is, is that it's only if it shows up, we get back to that problem.
648 "TRN6-29-BANFF/speaker_3" (890799616_3) 01:12:31.100 --> 01:12:35.850 In search results and the processing.
649 "TRN6-29-BANFF/speaker_3" (890799616_3) 01:12:36.850 --> 01:12:38.350 To enables search for.
650 "TRN6-29-BANFF/speaker_2" (890799616_2) 01:12:38.850 --> 01:12:39.850 Right.
651 "TRN6-29-BANFF/speaker_6" (890799616_6) 01:12:39.850 --> 01:12:42.850 I think we need more than the idea here. We need the.
652 "TRN6-29-BANFF/speaker_1" (890799616_1) 01:12:42.850 --> 01:12:44.350 Text yep so.
653 "TRN6-29-BANFF/speaker_6" (890799616_6) 01:12:44.350 --> 01:12:45.350 But it.
654 "TRN6-29-BANFF/speaker_7" (890799616_7) 01:12:45.350 --> 01:12:48.350 Potentially promising maybe I'll be during the break work on something.
655 "TRN6-29-BANFF/speaker_1" (890799616_1) 01:12:48.350 --> 01:12:48.850 Okay.
656 "TRN6-29-BANFF/speaker_7" (890799616_7) 01:12:48.850 --> 01:13:01.629 Alright, let's go ahead and break for now. Let's come back in 15 min at eleven and we'll we'll try and if if if we do have a proposal there, we'll talk about it. Otherwise, we'll move on to the training term.
657 "TRN6-29-BANFF/speaker_6" (890799616_6) 01:13:01.629 --> 01:13:08.629 I've also proposed a change to the final paragraph that I think is responsive to some of the discussions that were had here.
658 "TRN6-29-BANFF/speaker_7" (890799616_7) 01:13:08.629 --> 01:13:10.629 So folks, take a look at that.
659 "TRN6-29-BANFF/speaker_6" (890799616_6) 01:13:10.629 --> 01:13:12.629 Well yep ok.
660 "TRN6-29-BANFF/speaker_7" (890799616_7) 01:13:12.629 --> 01:13:16.629 And, and Martin, is this what you intend that that you're current reading?
661 "TRN6-29-BANFF/speaker_1" (890799616_1) 01:13:16.629 --> 01:13:18.129 The 1st 72.
662 "TRN6-29-BANFF/speaker_2" (890799616_2) 01:13:18.129 --> 01:13:19.629 The other.
663 "TRN6-29-BANFF/speaker_7" (890799616_7) 01:13:20.129 --> 01:13:21.129 Great.
664 "TRN6-29-BANFF/speaker_6" (890799616_6) 01:13:21.129 --> 01:13:23.629 I still have primary purposes.
665 "TRN6-29-BANFF/speaker_7" (890799616_7) 01:13:23.629 --> 01:13:24.629 Something's not right. Oh.
666 "TRN6-29-BANFF/speaker_6" (890799616_6) 01:13:24.629 --> 01:13:25.629 Oh, NO, I don't. Yeah.
667 "TRN6-29-BANFF/speaker_7" (890799616_7) 01:13:25.629 --> 01:13:27.629 Oh ok.
668 "TRN6-29-BANFF/speaker_5" (890799616_5) 01:13:27.629 --> 01:13:30.629 We'll be back in 16 min at 11aM.
669 "TRN6-29-BANFF/speaker_2" (890799616_2) 01:13:30.629 --> 01:13:31.129 Plan.
670 "TRN6-29-BANFF/speaker_6" (890799616_6) 01:13:31.129 --> 01:13:33.629 I think something happened with.
671 "TRN6-29-BANFF/speaker_1" (890799616_1) 01:33:06.580 --> 01:33:26.630 Right? Yeah, yeah, yeah, I wasn't sure whether I think Christopher, that's why we have.
672 "TRN6-29-BANFF/speaker_1" (890799616_1) 01:33:26.630 --> 01:33:52.280 Okay, everybody, let's get started. Right? That's very Dance for us as pennates. The dancing will continue until consensus. I do believe The we have a special exception.
673 "TRN6-29-BANFF/speaker_1" (890799616_1) 01:33:52.280 --> 01:34:12.060 Alright, come on, let's go. So do we have any further immediate substantive thoughts or concrete proposals about the search text or can we take it off to the list to allow people to gentlemen to simmer?
674 "TRN6-29-BANFF/speaker_1" (890799616_1) 01:34:12.060 --> 01:34:28.590 Can you. You're out of order. What these whole courtrooms out of order? It's that time of the meeting where everybody gets a little bit silly.
675 "TRN6-29-BANFF/speaker_1" (890799616_1) 01:34:28.590 --> 01:34:44.730 Sometimes that time is on the 1st day, but today is our time. Okay, let's put a pin in this for now then. Let it simmer.
676 "TRN6-29-BANFF/speaker_1" (890799616_1) 01:34:44.730 --> 01:35:00.029 Next up, we were discussing the towering Majesty that is the foundation bottle term, formerly known as the foundation bottle term. And I believe.
677 "TRN6-29-BANFF/speaker_1" (890799616_1) 01:35:00.029 --> 01:35:20.579 Into a place, where despite this comment, we selected this text B There is a pull request for this in a few minutes. What's the PR?
678 "TRN6-29-BANFF/speaker_1" (890799616_1) 01:35:20.579 --> 01:35:40.579 Give me a minute, two SF? Yeah, now there's more to this pull request unfortunately than just the definition because I had to do some housekeeping. But but this is the the the meet up.
679 "TRN6-29-BANFF/speaker_1" (890799616_1) 01:35:40.579 --> 01:35:59.489 Correct? That is the meant of it. So get rid of all that fine text and we have this very simple, the act of using an asset in the production or requirement of the weights of an AI model that can generate content in one or more modalities, text, image, audio et cetera. And we are renaming the category to AI training.
680 "TRN6-29-BANFF/speaker_1" (890799616_1) 01:35:59.489 --> 01:36:14.699 Sorry, AI model trends, thank you. Now, there was discussion about removing the phrase.
681 "TRN6-29-BANFF/speaker_1" (890799616_1) 01:36:14.699 --> 01:36:34.699 Weight solve, the weights off, which I think is a implication that would not necessarily harm the definition if there's NO objections to that but I wanted to test the waters? Yeah, there was a counterproposal to use parameters, right? I don't know who made that.
682 "TRN6-29-BANFF/speaker_1" (890799616_1) 01:36:34.699 --> 01:36:45.149 I did, but I I think the general sense was that just removing it. This this level of specificity doesn't add anything to the definition. Is that a controversial statement?
683 "TRN6-29-BANFF/speaker_1" (890799616_1) 01:36:45.149 --> 01:37:05.149 Do we have any hands? Yeah, bomb. Is the idea that this would Speaker freak up is the idea that this would include fine tuning that that category is implicit within this. Refinement. That's what refinement captures. Sure, ok, so then maybe it's still helpful to have that.
684 "TRN6-29-BANFF/speaker_1" (890799616_1) 01:37:05.149 --> 01:37:25.219 Sentence, that short sentence seeing what fine tuning is as part of this. I don't know if that's helpful. Would it also help if it was production and again we're getting to wordsmithing here. Production or subsequent refinement? Yeah, yeah. Just to to to emphasize that again I'm with you on the I don't I don't really wanna get in.
685 "TRN6-29-BANFF/speaker_1" (890799616_1) 01:37:25.219 --> 01:37:47.279 The weeds on that particular thing, I think that that would be a helpful addition to setting like a sentence just to keep in there but I think I think if you use the word to submit it implies that someone who takes a model from someone else is not covered by that. It's still subsequent. Yeah, I know, but.
686 "TRN6-29-BANFF/speaker_1" (890799616_1) 01:37:47.279 --> 01:38:07.279 Doesn't say by the same party. Yeah. And or subsequent. I I think I like it as is simpler the better. Sure. That is the simple form. Yeah, we often get in trouble when we add extra words. I think what we're trying to avoid is, being.
687 "TRN6-29-BANFF/speaker_1" (890799616_1) 01:38:07.279 --> 01:38:33.839 Avoid for veganness, which I don't believe we offer specific, then you miss any future things we didn't think of. I think the 1st line of this is beautiful. Train or fine tune whatever that fine tuning is, could be weight, could be, but that's fine tuning, so I like it the way. Good Kevin? Like specifically we've got a deleted line just to be careful.
688 "TRN6-29-BANFF/speaker_1" (890799616_1) 01:38:33.839 --> 01:38:50.879 You were reading the 1st deleted line. The act of using or train or fine tune was this the current text is the act of using an acid in the production. I'm also that ok good then sorry Kevin didn't catch me, thank you.
689 "TRN6-29-BANFF/speaker_1" (890799616_1) 01:38:50.879 --> 01:39:07.349 I don't think this is a substantive comment necessarily, and that I think we're all trying to talk about the same thing. I don't think they're trying to draw a distinction between pre training and post training or fine tuning or whatever you want to call it is very useful.
690 "TRN6-29-BANFF/speaker_1" (890799616_1) 01:39:07.349 --> 01:39:27.179 I think a definition where we say training is anything that changes the parameters of a model, catches everything and doesn't try to draw technical distinctions where there really aren't any. I mean even the refinement of weights, right? Right. You train a model you're you're initializing the weights somehow.
691 "TRN6-29-BANFF/speaker_1" (890799616_1) 01:39:27.179 --> 01:39:42.749 Let's say, and then you are refining them and at some point you decide you're done, but there's NO, that's all the refinement of weights, right? I'm wondering I'm sorry.
692 "TRN6-29-BANFF/speaker_1" (890799616_1) 01:39:42.749 --> 01:40:02.489 Any disagreement with that? I'm wondering if, if it would be helpful to add kind of a non normative explanatory note along those lines, something to the effect of in practice any use of the asset that has the effect of changing the model is.
693 "TRN6-29-BANFF/speaker_1" (890799616_1) 01:40:02.489 --> 01:40:17.699 Encompassed by this category or something like that. Would you like me to propose them? Okay. Nate? I think Kevin covered it. Excellent. Very good? So Martin, you'd get a.
694 "TRN6-29-BANFF/speaker_1" (890799616_1) 01:40:17.699 --> 01:40:34.379 PR or like a yeah where do you want to put that suggestion Kevin? What would be most helpful for you? Where I can find it. Put it in chat and I can put it in the edit thing straight away, that's ok. Yeah. I'll take it somewhere.
695 "TRN6-29-BANFF/speaker_1" (890799616_1) 01:40:34.379 --> 01:40:53.819 Justin, just clarify this will cover both the original producer of a model and a 3rd party that may choose to refine. Yes yes yes. Thank you. Or someone who's employing Laura or any of those other techniques that involve the creation of weights?
696 "TRN6-29-BANFF/speaker_1" (890799616_1) 01:40:53.819 --> 01:41:11.489 I'm good with that, thank you. I think that's good, ok? I would say we have conversions on that. Okay. Should I refresh this more?
697 "TRN6-29-BANFF/speaker_1" (890799616_1) 01:41:11.489 --> 01:41:30.359 I removed the off the weights thing. Okay, that's all. If Kevin can get me some contextualization, then I will add that and I gave you one minor editorial.
698 "TRN6-29-BANFF/speaker_1" (890799616_1) 01:41:30.359 --> 01:41:48.869 The job included at your leisure. So I'm not looking at that oh.
699 "TRN6-29-BANFF/speaker_1" (890799616_1) 01:41:48.869 --> 01:42:05.279 Yeah, I'm not sure that's right. It is right. We have to be cons whatever word we're gonna use should be consistent, but I don't think it needs I I think this is specifically not consistent. Right so.
700 "TRN6-29-BANFF/speaker_1" (890799616_1) 01:42:05.279 --> 01:42:25.279 I flagged the word, so in the definition it says the act of using an asset that can generate content. And my point was that whatever word where you end up using for assets, we generate assets so that we are consistent with in the wording that we can't use asset in one place content.
701 "TRN6-29-BANFF/speaker_1" (890799616_1) 01:42:25.279 --> 01:42:45.389 Another or get, whatever word we bit, it applies in both cases. Well, my my my choi my my reason for choosing the word content in this case was to avoid dragging in the the definition we have of asset or artifact or whatever it is that we have right because.
702 "TRN6-29-BANFF/speaker_1" (890799616_1) 01:42:45.389 --> 01:43:00.899 When we're talking about an asset, it's like the whole unit, right? Yes, exactly. Whereas content in this case is specifically some portion of material, not the whole thing. Not necessarily I mean not necessarily, but.
703 "TRN6-29-BANFF/speaker_1" (890799616_1) 01:43:00.899 --> 01:43:16.409 But definitely possibly, this is this is a narrower, a narrower thing than the, the, the full thing. If you limit this to producing of assets, which is in our definition an entire thing, we, we miss something.
704 "TRN6-29-BANFF/speaker_1" (890799616_1) 01:43:16.409 --> 01:43:35.669 I get that. Then we ought to define content as well in this case. We do. What we do where fixed. Audio, video. Right. We need to do some more explicitly. We can we have to be a little yeah just.
705 "TRN6-29-BANFF/speaker_1" (890799616_1) 01:43:35.669 --> 01:43:56.839 I think at this stage we're using our placeholder, we need to go through the entire to see that we're consistent about these things that out with regards to individual sentences possibly add definitions and so forth and so on. Yeah. I'm happy to use a different word here, but I don't I don't think we're gonna use the same words that we have. Okay.
706 "TRN6-29-BANFF/speaker_1" (890799616_1) 01:43:56.839 --> 01:44:17.329 Yeah, because the intent is to to say you're creating material or something that is not that is not necessarily the same thing. No, that I, I totally I understand what you're trying to accomplish. Let's let's take it offline. Alright, that's done. Yeah, let's let's move on then to.
707 "TRN6-29-BANFF/speaker_1" (890799616_1) 01:44:17.329 --> 01:44:36.599 You are glorious friends section three.two. Where did the issue go? There it is. And yesterday we had, I believe two competing proposals for this as well.
708 "TRN6-29-BANFF/speaker_1" (890799616_1) 01:44:36.599 --> 01:44:55.949 And I don't believe we did select one of those as a way to go forward and I'm wondering where they are hiding.
709 "TRN6-29-BANFF/speaker_1" (890799616_1) 01:44:55.949 --> 01:45:11.219 She was that. So we're talking about one 60 now three.two. Not quite yet. So we had this one, two oh four and two oh six.
710 "TRN6-29-BANFF/speaker_1" (890799616_1) 01:45:11.219 --> 01:45:29.969 Yeah, great. Let me see if I can find the one where. So two oh six is the one that I'm.
711 "TRN6-29-BANFF/speaker_1" (890799616_1) 01:45:29.969 --> 01:45:46.199 Suggesting we we've got. So this is just a transcription of what Brad suggested and we moved past that. Okay. The two oh six takes the present text.
712 "TRN6-29-BANFF/speaker_1" (890799616_1) 01:45:46.199 --> 01:46:04.019 Alyssa's suggestion, Brad's suggestion and tries to take all of the discussion we had yesterday and synthesize that into something that's a little, more responsive to all of the concerns that were raised.
713 "TRN6-29-BANFF/speaker_1" (890799616_1) 01:46:04.019 --> 01:46:22.559 Let's do this. And diff is very difficult to read on. I'm gonna look at the. This is what I want. Make it bigger, the worst thing.
714 "TRN6-29-BANFF/speaker_1" (890799616_1) 01:46:22.559 --> 01:46:39.929 And so, 3.2 yeah I'm looking, three.one isn't affected, ok. So let's let's look through this. Oh boy, that's taking a while, but.
715 "TRN6-29-BANFF/speaker_1" (890799616_1) 01:46:39.929 --> 01:46:59.929 Oh, is it? There we go. So this specification provides a set of definitions for different categories of use plus a system for associating simple preferences to each other. The categories of use in the vocabulary describe concrete observable outcomes that depend on the use of assets. They seek to avoid des.
716 "TRN6-29-BANFF/speaker_1" (890799616_1) 01:46:59.929 --> 01:47:04.019 Grabbing internal details of implementations or their.
717 "TRN6-29-BANFF/speaker_1" (890799616_1) 01:47:04.019 --> 01:47:24.019 This specification only seeks to ensure that preferences can be understood, not provide a means of ensuring that preferences are followed. I feel like we're bearing the lead a bit there, but sure. Yep. I like it. Enforcement is not provided by this specification, Preferences do not themselves create rights obligations or.
718 "TRN6-29-BANFF/speaker_1" (890799616_1) 01:47:24.019 --> 01:47:43.469 Or prohibitions either in the positive or the negative. Yeah. Other mechanisms technical, legal contractual or otherwise? Traction I never mind, might enforce adherence to these preferences and thereby determine the consequences of not following the state of preference.
719 "TRN6-29-BANFF/speaker_1" (890799616_1) 01:47:43.469 --> 01:48:00.929 An entity that receives usage preferences has a choice about whether to respect the preferences it is discovered. It makes this choice according to the understanding of how the asset is used, how that usage corresponds to the usage categories, where preferences have been stated and the applicable legal context.
720 "TRN6-29-BANFF/speaker_1" (890799616_1) 01:48:00.929 --> 01:48:24.559 Historically entities have used assets for a variety of reasons. This has included a wide array of uses in applications including safety, advertising, education, scholarship, research, market intelligence, preservation, surveillance, interoperability, and accessibility. These decisions were based on a subjective analysis of preexisting signals about.
721 "TRN6-29-BANFF/speaker_1" (890799616_1) 01:48:24.559 --> 01:48:45.629 Differences and an understanding of how the asset would be used in that application. The mechanisms in this document might be more direct than previous ways of expressing preferences, but the same fundamental challenges remain when deciding whether or not to use an asset.
722 "TRN6-29-BANFF/speaker_1" (890799616_1) 01:48:45.629 --> 01:49:07.339 Okay go for it. Yeah, just put a suggestion in the chat, but since this is in the 1st paragraph that's visible right now since it's that there are NO rights obligations or prohibitions and trying to be very like balanced about.
723 "TRN6-29-BANFF/speaker_1" (890799616_1) 01:49:07.339 --> 01:49:26.669 No obligations in either direction, then I'd suggest that on a mode. Then I suggest other mechanisms.the.might enforce adherence or non adherence. So, that's.
724 "TRN6-29-BANFF/speaker_1" (890799616_1) 01:49:26.669 --> 01:49:42.929 That help contribution Timmid robot? Timod robot. In the same vein Mark, when you read the last sentence, you added or not, whether or not to use an asset.
725 "TRN6-29-BANFF/speaker_1" (890799616_1) 01:49:42.929 --> 01:50:00.299 My fault, well, but perhaps kind of in a similar vein to what was just said, like adding that would be helpful. Right, keeping it neutral. Martin, take that as a friendly. Yep. Last sentence, where is that weather?
726 "TRN6-29-BANFF/speaker_1" (890799616_1) 01:50:00.299 --> 01:50:15.629 Thank you. Leonard? Yeah, so I just submitted a couple of changes, but one of them I wanted to talk through, which is, the.
727 "TRN6-29-BANFF/speaker_1" (890799616_1) 01:50:15.629 --> 01:50:33.899 Some of the wording in here doesn't make it clear to whom or to what this text is referring. So, e.g., let's just look at that last slide. The same fundamental challenges remain when deciding whether to use an asset, when.
728 "TRN6-29-BANFF/speaker_1" (890799616_1) 01:50:33.899 --> 01:50:53.899 What is deciding when who is deciding, to use it? Is it the end users? I mean obviously we know we wrote this, but there's NO it there's it's missing a, you know, subject, if you will. And there's another place that I also flagged where the same thing is true, where we don't give direction as to who this is.
729 "TRN6-29-BANFF/speaker_1" (890799616_1) 01:50:53.899 --> 01:51:18.019 Is being directed at. So, I think we need some, some subjects there. Do you have a suggestion of how to identify the entity that you're addressing? In my propose in my, in the comments I just sent you, I use the word system because it seems like the most agnostic, but I am absolutely up for other words. I think we can take it for granted that these are gonna.
730 "TRN6-29-BANFF/speaker_1" (890799616_1) 01:51:18.019 --> 01:51:33.258 I need a couple of passes editorially for for English grammar, you know, the clarity's organization Brad?
731 "Brad Silver" (352249088) 01:51:33.258 --> 01:51:57.119 The last paragraph isn't really accurate to say, and I think this is the implication is that there's, there's historical context that is relevant here, and I wonder if that's correct because while there has been use of assets historically, we're creating something new. And so our.
732 "Brad Silver" (352249088) 01:51:57.119 --> 01:52:17.076 Are we saying, you know, that these sorts of uses sort of should continue that there I mean I it's I'm not, I'm not exactly sure what guidance this actually gives.
733 "TRN6-29-BANFF/speaker_1" (890799616_1) 01:52:17.076 --> 01:52:32.516 So I I think that the purpose here is to relate it to the context in used including the historical context, not not to guide in a sense, but more to caution overreading what what's, what this framework is providing.
734 "Brad Silver" (352249088) 01:52:32.516 --> 01:52:49.769 Yeah, it sort of reads a little bit like that, but if if you're saying don't overread this, then it sort of doesn't where when where does it leave the reader? And I don't know if we Then we're sort of saying.
735 "Brad Silver" (352249088) 01:52:49.769 --> 01:53:08.900 You know, I think we've said so much about this being uninforceable, and now we're sort of saying and don't over overread even what, what is, what is uninforceable. I mean, how much further than being uninforceable can you go?
736 "TRN6-29-BANFF/speaker_1" (890799616_1) 01:53:08.900 --> 01:53:14.001 So I think this is kind of the key paragraph.
737 "Brad Silver" (352249088) 01:53:14.001 --> 01:53:20.875 No, I think it's actually the last I'm I'm talking about the last, last paragraph, but.
738 "TRN6-29-BANFF/speaker_1" (890799616_1) 01:53:20.875 --> 01:53:25.359 Yeah, I mean. Yes Yep.
739 "Brad Silver" (352249088) 01:53:25.359 --> 01:53:47.900 Yeah, but that's sort of the meats of it. I think it almost feels like you try and over explain something and in overexplaining it, you make it a little bit more fuzzy and I I, I think that the last sentence in an attempt to be clear actually makes things more fuzzy.
740 "TRN6-29-BANFF/speaker_1" (890799616_1) 01:53:47.900 --> 01:54:00.616 So I, I I hear you Brad. I'm I also hear that a lot a number of folks in the room feel that it's important to have text like this. Okay. Chris.
741 "Brad Silver" (352249088) 01:54:00.616 --> 01:54:07.109 I and I I I hear that, that, I hear the importance of that. I, I and I I'm not.
742 "Brad Silver" (352249088) 01:54:07.109 --> 01:54:22.259 I'm not pushing back on, on, on, on the need for it, but I I I wanted to be, to be clear, I mean because historically entities have used assets for a lot a variety of reasons. They've also used those assets in the absence of preferences.
743 "Brad Silver" (352249088) 01:54:22.259 --> 01:54:43.822 So, you know, I don't want the implication to be like these preferences really shouldn't mean anything to those that have used these assets in the past for a historical context that is now evolved. So I that that that that's all I don't, I don't want to oversell what we're doing here. I also don't wanna like undersell it.
744 "TRN6-29-BANFF/speaker_1" (890799616_1) 01:54:43.822 --> 01:55:02.939 I, I could see that that that last paragraph could be re rewritten in a way that doesn't lose, you know, the nuance and the examples, but it's perhaps a bit crisper. I think part of the, what I was attempting to do here was, was ensure that we.
745 "TRN6-29-BANFF/speaker_1" (890799616_1) 01:55:02.939 --> 01:55:19.199 We didn't we weren't too specific about those historical things. I think there have been preferences about how things are are used in the past. It's just we've had NO good ways of expressing those preferences.
746 "TRN6-29-BANFF/speaker_1" (890799616_1) 01:55:19.199 --> 01:55:39.199 And so that's really what it's what it's saying and and acknowledging that the, the, the challenges that existed previously in terms of understanding what preferences were persist even when the, there are means of expressing preferences more clearly.
747 "TRN6-29-BANFF/speaker_1" (890799616_1) 01:55:42.269 --> 01:56:02.269 Brad like doesn't the last sentence kind of go a little bit into what Martin just said, like to make it like differentiation from what was there before, right? Like, you know, that we are being more direct and this is this is a piece that I took from Alyssa's proposed text and tried to work it in. So if I did so imperfectly, that's, that's on.
748 "TRN6-29-BANFF/speaker_1" (890799616_1) 01:56:02.269 --> 01:56:10.546 My team worked on it a bit but yeah thanks Brad Chris.
749 "Chris Needham" (1284876288) 01:56:10.546 --> 01:56:39.539 So see, yeah, I'm I I also struggle with this paragraph. I I mean I sort of said yesterday about about that. I I think the rest of this section reads pretty well to me and I think that conveys the point about, you know, the, the, the, the sort of the the surrounding context for this. I think if we wanna talk about history, then we need to be.
750 "Chris Needham" (1284876288) 01:56:39.539 --> 01:57:01.945 A bit more honest about what we're saying because, you know, the recent history suggests that assets have been used like with NO regard whatsoever for you know site rep, you know, site preferences. And so I think this sort of presents a version of history which is incomplete and.
751 "Chris Needham" (1284876288) 01:57:01.945 --> 01:57:34.180 I think I think is pointing towards the sort of surround, the surrounding sort of regulatory context, and I and and I still maintain that that's not, really what we should be attempting to do in this document. That's, you know, the what the kinds of uses for education and and research and so on. You know th these are things that that have sort of existing provisions that, that we don't need to, to.
752 "Chris Needham" (1284876288) 01:57:34.180 --> 01:57:54.259 I think this paragraph really, as Brad says it it just it just adds I think confusion to the, to, to, to what we're saying and and I think we're better off leaving it out of the document and because the rest of what we say in this section actually.
753 "Chris Needham" (1284876288) 01:57:54.259 --> 01:57:56.500 Conveys what we do mean.
754 "TRN6-29-BANFF/speaker_1" (890799616_1) 01:57:56.500 --> 01:58:13.846 So Chris, can, can I ask, would it would it give you more comfort if, if we weren't couching this in historical terms, but we we said something like, you know, in practice entities of use assets for a variety of reasons and so forth and so on. So try not to relate it to historical, but just to common practice.
755 "Chris Needham" (1284876288) 01:58:13.846 --> 01:58:25.339 No, because my ladder point rep applies, which is it's it's the surrounding context in which we're working with that defines those things. It's not for us to define those things.
756 "TRN6-29-BANFF/speaker_1" (890799616_1) 01:58:25.339 --> 01:58:43.665 Well I I just know, you know, we've already had a significant amount of discussion around compromise here and I think that there's there's been a lot of compromise on this text, so I'd ask you to think carefully about whether, you know, this is something that you just think is not terribly helpful versus you're actively against.
757 "Chris Needham" (1284876288) 01:58:43.665 --> 01:58:48.161 Understood. Yeah, thank you.
758 "TRN6-29-BANFF/speaker_1" (890799616_1) 01:58:48.161 --> 01:59:03.449 So I think two things. One is I don't know that it's always super helpful to talk about these as not enforceable because at some level, it's not that they are unenforceable, it's that they aren't self enforcing, right? Like these aren't.
759 "TRN6-29-BANFF/speaker_1" (890799616_1) 01:59:03.449 --> 01:59:22.709 There are, there are going to be ways in which these are in different contexts used in an enforcement way. And so saying that it doesn't matter that we don't talk about these exceptions or these times you might not follow them because they are forceable, we've decided that at some level people care about these because they think they will be in some way.
760 "TRN6-29-BANFF/speaker_1" (890799616_1) 01:59:22.709 --> 01:59:42.029 Enforced. So I think there's that. 2nd part is, we could say in here sort of explicitly, if these are used in a enforcement context, if they're used in a way that is binding, they need to take into account.
761 "TRN6-29-BANFF/speaker_1" (890799616_1) 01:59:42.029 --> 01:59:57.989 Law or like other areas in which you would ultimately legal obligation not to find, not to follow them because that is the piece that I think is as missing here, right? That if these become enforced or enforceable, they're going to conflict with other.
762 "TRN6-29-BANFF/speaker_1" (890799616_1) 01:59:57.989 --> 02:00:13.319 Legal and in some situations I think ethical obligations. I think the current text says like we don't provide a mechanism to enforce, I don't think he says unenforceable, he's in here. No, but in the discussion I think people said well like it doesn't matter if there's exceptions cause.
763 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:00:13.319 --> 02:00:28.679 They're not enforced, but they think will be. Thank you. And then, so I wanna say my comments are partially about the overall times, but largely about what we talked about.
764 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:00:28.679 --> 02:00:45.839 The the the the highlighted facts that they're the the paragraph that's highlighted up there. As I read through this, we are being so balanced on both sides of the thought of are they in, are they out? You have to do it, you do not have to do it.
765 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:00:45.839 --> 02:01:04.229 I worry that the reader is going to get lost in in in the way we've expressed our meaning. And that may likely lead as it has in the past to people both reading, misreading our intent.
766 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:01:04.229 --> 02:01:24.229 That's currently written, but worse, people deliberately misreading the intent to fit their expectations of what they want to achieve, and that's always a troublesome outcome because it means we fail to convey concisely what we're really trying to say in an unambiguous way. And I do wear that we are bouncing around a little bit.
767 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:01:24.229 --> 02:01:45.869 I have a suggestion, that's why I was looking for the link to this actual edited text. I was actually gonna try to drop it into an AI system and actually ask you questions about what is it like given this following text, you know, what am I allowed to do? You know, can I do this? Can I do that? Actually, just to use for the lack of a independent.
768 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:01:45.869 --> 02:02:01.559 1st time viewing this expert in the space, which we don't have in this room, to try to have an AI be that entity because I, I do worry that has written that it could lead to high misinterpretation or adult misinterpretation. So I, I hear concerns.
769 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:02:01.559 --> 02:02:19.649 Also I was trying to drop in the AI so I should provide you the pull request link. Is there an where's the editor's copy of this? It's in GitHub. Okay, i found GitHub but I didn't see it right. I don't know that as a group we should walk down the road of using AI to to.
770 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:02:19.649 --> 02:02:34.979 I'm not suggesting that. I'm not suggesting whatever tools you use, then that's, that's on you. Who's next? Justin?
771 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:02:34.979 --> 02:02:51.689 I battled with it historically part of the way some other people did, but I kind of come to the mind that I think that the speaking about the historical part points to the fact that historically this may have been done differently.
772 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:02:51.689 --> 02:03:11.689 Take notice that the mechanisms now are more direct than they were. So if you thought we were doing this one way before, take note that we're doing it differently now. And I think that's what I believe we're trying to say here, which is good. I proposed a version in chat that simply removes that this has included a bunch of stuff because I don't think we necessar.
773 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:03:11.689 --> 02:03:29.639 You really need to say that and maybe it makes it more clear. You mean the entire sentence? Pretty much that's what I, Yeah, like I just, NO, and the next sentence actually. So the two sentences, so it's just historically we did this for a bunch of reasons.
774 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:03:29.639 --> 02:03:46.169 This is different now. That's it. Alright, it's NO terrible.
775 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:03:46.169 --> 02:04:02.279 We cannot hear you if you're speaking. But I just want to push back very hard on that. I just very strongly disagree that.
776 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:04:02.279 --> 02:04:22.279 If what we are saying here is we are now enunciating preferences in a different way, do not acknowledge directly that those may still be limited by valid accessibility research for whatever reasons. I think this will be read to over, at least by many people to clawback those.
777 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:04:22.279 --> 02:04:51.697 It's existing uses. And so that's why I think it's important to say here that you may look at these preferences and still determine that your uses historically for things like accessibility are preserved. I respond. Very quickly. I think some of that was, I think was addressed yesterday in some of the other parts of the document, but but I definitely for search though, but not sure. Thank you. Victoria, you wanna try again?
778 "Tori Noble" (1142273792) 02:04:51.697 --> 02:05:16.876 Yes, sorry I didn't hear my name the 1st time. I struggly agree that for the same reasons we both said just said and also that we discussed yesterday the list of these exceptions really does need to be here in both sections, and I think that seemed to be the consensus of the group yesterday, thinks that there are new concerns about that.
779 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:05:16.876 --> 02:05:20.159 Please don't try to characterize consensus the chair.
780 "Tori Noble" (1142273792) 02:05:20.159 --> 02:05:39.059 We'll do that. Sorry, sorry I don't mean consensus, but I mean, to the extent that there are concerns we didn't discuss yesterday, I mean there those might be worth considering, but other than that it seems like there was support, NO, not trying to say consensus, but.
781 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:05:39.059 --> 02:05:42.386 Recognizing that these are good uses.
782 "Tori Noble" (1142273792) 02:05:42.386 --> 02:06:00.502 The technology that we don't want to imply should be clawed back now or that we're changing the system in a way like what we're doing now that's different is a reason to NO longer recognize these uses.
783 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:06:00.502 --> 02:06:21.059 Yeah, I think one point Peter, like that, I think Kevin made yesterday was like kind of combine good, I I don't think this characterizes this good uses, it's just historical users, right? It's kind of a combination of good are like not good stuff. So I just wanted to make sure that that's noted somewhere. So I, I think where we're at here is, is we're hearing from some folks.
784 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:06:21.059 --> 02:06:41.059 That this text doesn't add something or it may cause confusion. It's not well stated. We're hearing from other folks that this is very important to them and that that they need to see something like this in the draft. So I think the paths forward are either to try and clarify.
785 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:06:41.059 --> 02:07:02.129 By this text and address the 1st concerns or to convince the the other folks that this text isn't necessary or create a strong argument that the text isn't necessary that we've used to declare that in the rough. You know, that we've been going back and forth on this for, for a long time, so.
786 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:07:02.129 --> 02:07:21.599 Please consider carefully, you know, how do we make progress here? Paul, go 1st. I also want to reiterate the point that I made yesterday that I think that historically, the historically in here is very strange language wise, it implies that there is a shared history and that I think.
787 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:07:21.599 --> 02:07:41.599 A lot of concerns that are sort of embedded in this thing are very particular US legal system concerns that are in here. Historically, other legal systems have dealt with this being very explicit about what's allowed and a sort of like a sort of unquality.
788 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:07:41.599 --> 02:07:59.069 The fight historically in here sounds, I think to a lot of readers outside of this room, this community will be very strange. Thank you. My.
789 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:07:59.069 --> 02:08:16.439 My concern is again within the the realm of enforceability, I understand that we're not defining forceability, but I don't think you can ignore it. And I.
790 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:08:16.439 --> 02:08:32.069 I think that, we haven't I I I don't currently think we have a, a vocabulary in which good actors can.
791 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:08:32.069 --> 02:08:48.480 Comply because there there's not enough carve outs, like, I think that this.
792 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:08:48.480 --> 02:09:03.810 Is I think that this paragraph is inadequate to meet that need, but I'm reluctant to remove it at like this stage of the document myself.
793 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:09:03.810 --> 02:09:22.830 That I think that it's, maybe it's a thorn, but it's like a phone I don't want us to remove until kind of later in the process. Maybe it should live somewhere else. Yeah, there was discussion when we were.
794 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:09:22.830 --> 02:09:39.450 Working this out in the search category of maybe refactoring the mention of accessibility and kind of other considerations from that category up into the vocabulary overall, and I wonder if that might be a way kind of out of this.
795 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:09:39.450 --> 02:09:54.450 Because I think that language was a bit more acceptable to folks, but also I agree I think those are really important considerations to reflect.
796 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:09:54.450 --> 02:10:10.110 Somewhere in the vocabulary and I I don't know that like a, a reference to historical practice is the best way to do it rather to say that, you know, there there maybe valid motivations for.
797 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:10:10.110 --> 02:10:30.110 Not following a preference, including but not limited to accessibility or preservation or, you know, so this is a little weird text and I I think that's part of the tension that we're getting here is that.
798 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:10:30.110 --> 02:10:45.870 We're trying to walk this line between saying that there are things that people will use to justify basically ignoring a preference and not coming out and trying to enumerate.
799 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:10:45.870 --> 02:11:05.870 Those things. I think Aaron's point here about find the specific carve out about accessibility doesn't really address the core problem that we have here. It would help if we sort of raified this idea that accessibility was was the thing, but then that privilege is accessibility under some.
800 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:11:05.870 --> 02:11:27.800 Are the other things that people care about. And I think that doesn't really get to the core of the reasons that we have this sort of text in the document. I think with Alyssa's proposal here, and this is largely based on on that, the idea is to try to walk that line by saying that there are a bunch of reasons that people have chosen to effectively ignore people's.
801 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:11:27.800 --> 02:11:47.160 References of the stuff is used without explicitly endorsing those practices. And part of I think Kevin's suggestion here was helpful in doing was identifying things that some people might object to, and so I included the advertising and surveillance.
802 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:11:47.160 --> 02:12:03.390 Items in that list for specific that specifically that reason, because then it gives us that sort of, here are some of the things that people have in the past, in some jurisdictions chosen to use this justification for ignoring preferences.
803 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:12:03.390 --> 02:12:21.270 But we're not doing anything to endorse particular practices one way or or another. And that was a deliberate choice here following in the sort of suggestion that Alyssa had. But I I recognize that this doesn't go far enough.
804 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:12:21.270 --> 02:12:39.810 For some people and it doesn't address the sorts of concerns that Meredith and and Tim and Robot have have raised here. And I don't know how much protection we need to add in, in the document for these practices. I know a lot of people are looking for this document to sort of.
805 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:12:39.810 --> 02:13:10.420 Endorse some set of practices that they consider important to them, but fundamentally I don't know that we can do that in in this document. And so by saying historically and recognizing that, that, that these exceptions exist without endorsing any of them, hopefully walks that line, but I'm not, I I recognize this is weird text as a result. This.
806 "Chris Needham" (1284876288) 02:13:10.420 --> 02:13:35.790 Yeah I thought I wanted to respond to something that a previous speaker. Was it Victoria said, about the work here sort of not making it explicit that what we're not trying to do is to influence that kind of surrounding regulatory context for our work. Now what we, you know, I think that would be a mistake, like if we end up with something here which ends up.
807 "Chris Needham" (1284876288) 02:13:35.790 --> 02:13:51.090 You know, being more restrictive seems like a bad outcome, so to saying something to the effect that, you know, we're, that we're not trying to influence that sort of one way or the other.
808 "Chris Needham" (1284876288) 02:13:51.090 --> 02:14:06.090 Perhaps maybe a helpful, helpful way to, to, to frame it. Because if that's if because if that's what the underlying concern is, then then you know let's you know then then I can think we can sort of speak directly to that.
809 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:14:17.000 --> 02:14:39.360 Thank you. So I think one of the things we're grappling with here with is like because we deal with these, the nature of this process is that we deal with very little chunks of the text at a time. And so there's been some discussion of like where in the whole structure of the document, these concerns go.
810 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:14:39.360 --> 02:14:56.310 And so one of the things I would say is maybe, you know, it's 1145, maybe we can take some time between now and our next online meeting to look and think about where these go in the whole structure of the document. There's also that section about potential impacts.
811 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:14:56.310 --> 02:15:14.160 Which I don't totally understand how that reads into the spect all, I would say. But I can look at this drafting. This is weird bad, and I think they're just like a bunch of things matched together, but I know there's a few of us to middle about me, Lila, who could work on this and try to have something.
812 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:15:14.160 --> 02:15:30.510 That isn't too ambiguous but just that does communicate some concrete sentences. Thank you. And Yeah, I.
813 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:15:30.510 --> 02:15:49.140 I think the effort to be vague by referring to history is vague in the wrong way that I'm and I am understanding the discussion here.
814 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:15:49.140 --> 02:16:05.100 As us saying this room is trying to communicate to future users of the vocabulary and future possible enforcers of the vocabulary, that we understand that the vocabulary may not like.
815 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:16:05.100 --> 02:16:20.700 Apply in all circumstances or that there, there maybe considerations or contexts in which it is valid not to follow the preferences. And can we agree on and just say that?
816 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:16:20.700 --> 02:16:36.060 I think I think that that is what the section does. I don't. Okay, so the paragraph above. Yeah.
817 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:16:36.060 --> 02:16:51.960 Again, this is descriptive of what might happen in future and not a statement about how to interpret the preferences. It's that's a good point, right?
818 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:16:51.960 --> 02:17:11.960 Can I request that you attempt to just give us a very direct. Right. I think I I'm from a FinTech background and FinTech surprisingly has a lot of models that they train being.
819 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:17:11.960 --> 02:17:35.030 Based off of in in the ecosystem, mouse movements when I purchase my banana republic, to ensure that I'm not a fraudster and and my I'm not typing things too fast and whatnot. And I don't think that, and the reason from a data privacy standpoint, we, there's not controls to allow fraudsters to opt out and if the legal.
820 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:17:35.030 --> 02:17:56.940 Basis for processing is not consent and rather it's legitimate interest is because you don't want to let the fractures opt out. None of us, we all want to continue to participate in the ecosystem without getting defrauded, right? And so I think that sentiment applies for, you know, some of the discussion we previously previously had regarding CSAM.
821 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:17:56.940 --> 02:18:16.940 And such. So I wonder if the solution here is to with respect to the training definition cause I feel like this language is kind of more like pre picture language that folks might not pay attention to, but if we can add clarity to the main document around training control cause I I suspect.
822 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:18:16.940 --> 02:18:38.700 That the group is predominantly focused on adding an opt out associated with Gen AI models that deal with the output of, you know, images and texts and all of those things. I don't think this group is trying to like provide an opt out associated and a preference opt out associated with training of.
823 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:18:38.700 --> 02:18:57.450 You know, the models that frankly in our day to day, we don't even realize is happening in the background that protect us from fraud as, you know, as the fin tech industry, you know, relies so heavily on. So I wonder if it's about going back to the training definition and narrowing that to.
824 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:18:57.450 --> 02:19:16.920 Foundational models that produce the type of content that are is focused on Gen A gen AI around text and image and video and, audio. Bulk? Yeah, I think it's broader, like this is essentially about.
825 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:19:16.920 --> 02:19:36.920 A public interest acknowledged-exception and maybe exception is too strong this word. And I think this document is going to be incomplete when we are not having something in there. Like this is a constituency that is not very well represented in these discussions that's.
826 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:19:36.920 --> 02:19:55.350 Sort of like, likely to fall by the wayside. I think pointed out that these are generally very, very, very risk averse organizations who want to do the good thing and we need to give, we need to speak to them.
827 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:19:55.350 --> 02:20:12.570 In order for them to understand their place in this. And that needs to be by I I think honestly like the enumerative list that wasn't complete is more or less the only way you can do that by clearly pointing out what this is supposed.
828 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:20:12.570 --> 02:20:30.630 Not to restrict, and I think we need to be mindful of the concerns that were expressed yesterday that this can't veer into the territory where using the preferences equals granting a licensed or or a positive right to them, but like you need to.
829 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:20:30.630 --> 02:20:47.610 Speak to them. Otherwise, this will create collateral damage and language like historically, like they will read this as like they are so risk averse, they will read this as in the past this was ok and now it's not ok anymore. And this this just doesn't.
830 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:20:47.610 --> 02:21:05.280 This just doesn't help. Like I think we in the end, when I think Meredith suggested that we really need a piece of public interest language in here and I very gladly see something coming from you that tries to address that, that doesn't go as far as I'll be sort of like.
831 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:21:05.280 --> 02:21:22.410 Yeah, well, yeah. So we'll cut the queue at Glenn. Yeah, thank you. I might have go ahead and say whatever you wanted to say to You know I was just promising Paul it'd be a very careful small drafter. Thank you. Thanks.
832 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:21:22.410 --> 02:21:38.310 Max, yeah, I wanted to just respond a little bit to what I think Aaron and were saying, two steps, so I don't, I don't really agree that we should be given.
833 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:21:38.310 --> 02:21:54.510 Giving guidance to enforcement, like to enforcers on like when exceptions exist or like dictate why they like I understand some of the instics here but I I don't think that's what we should be doing. And I think it's firmly.
834 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:21:54.510 --> 02:22:14.510 On individual publishers to make those decisions on what they do want and do not want to accept in terms of exceptions. And I wanna like especially address like Kathy's point around like FinTech models. I don't think that's a very good metaphor in this case. I don't think.
835 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:22:14.510 --> 02:22:33.060 This is like you're, you're sort of expressing like user consent in in processing models and I don't think that model really applies to like the site owner model creator or relationship. I think it's different. Thank you. Did I drop off?
836 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:22:33.060 --> 02:22:51.960 Tim and robot, I just wanted to respond that it was my understanding that we, we can't restrict training down to foundational models based on the conversation yesterday because like or just generative models because.
837 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:22:51.960 --> 02:23:09.300 The models that you wouldn't call generative still have the capacity to be generated. And so while it that basically that category exists in our minds, but once we try to nail it down, it.
838 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:23:09.300 --> 02:23:28.080 It's nearly impossible. Thank you. Wow, pressure. I'm not against coming up with some kind of language here. I will say that as a industry that has suffered greatly at people try to use these little.
839 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:23:28.080 --> 02:23:46.350 Public interest exceptions to say movies should be free. Content should be free, right? I I am worried that the the selection of the language here while I appreciate it opens up support for the public interests could also be misused for support for the folks.
840 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:23:46.350 --> 02:24:06.350 Under I'm a researcher or I'm a public just cause want to make other things available. And so I believe it can be done, but if you want to take a shot out, I won't say that. Okay. So it's clear that that we need to continue discussing this. It's clear we need.
841 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:24:06.350 --> 02:24:26.580 Proposals. I I hope everyone now has a more of a sense of the challenges we face in this area. I think that there's a very delicate balance to be had. The question in my mind right now is do we think that the the current text and the specification is the best place for us to launch that?
842 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:24:26.580 --> 02:24:44.790 Development from or is this text and improvement over that? Sorry, I have three I have three suggestions. Sure because Aaron made a suggestion in the chat that I think is, is reasonable. Okay. 1st is the the text in the current draft.
843 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:24:44.790 --> 02:25:04.790 Second is this, 3rd is this with the final paragraph removed and a sentence added, and there's a different chat, but that the the sentence that Aaron suggested was there maybe considerations that take precedence over any stated preference.
844 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:25:04.790 --> 02:25:30.480 Especially when taking the public public interest into account and that follows on from the, the, the sentence at the very top. Let me see if I can find this. So I see a longish diff ish looking thing from you, is that it? That's it. And I'm looking for that in the top with the like it's the 3rd line from the top. Okay, hold on.
845 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:25:30.480 --> 02:25:46.620 I'm dealing with the excellence. Your face here. What's the best way to do this? Just for clarity, Aaron's proposal is to make this a like two sentence paragraph.
846 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:25:46.620 --> 02:26:01.950 The entire section. So just would minus out. I know you suggested that, but I'm just simply taking the sentence and trying to preserve the rest of it because I think that there is stuff that we've added to the rest of this that still carries value in, in that context.
847 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:26:01.950 --> 02:26:18.210 That would be the 4th option here, I guess, which is to remove remove the entire section and replace it with a very short statement like you suggest. At least we could talk about that. Yeah. So your, your formulation would be.
848 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:26:18.210 --> 02:26:34.530 Current formulation would be just that one sentence at the top of that section that the yes, that one plus plus that. And I'm suggesting that there's a version of this that takes this, yeah.
849 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:26:34.530 --> 02:26:55.850 Yuck, that takes those two things and simply puts them in place. So remove remove the historically paragraph and make this modification. So that's that's one option. What Aaron is suggesting is that this is the entirety of that section that replace.
850 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:26:55.850 --> 02:27:00.840 Is all the other paragraph like remove all the other paragraphs aside from this.
851 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:27:00.840 --> 02:27:20.840 Which I don't think is going to work because we've had a lot of questions about what it means to follow right and respect preferences. I think the rest of it has that reflects a fair amount of of discussion that we've had. Right. I think at least the next paragraph is, I think essentially, like the other mechanisms.
852 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:27:20.840 --> 02:27:36.240 Yeah. I I do have questions about I think this is a little bit elaborate, but that's an editorial comment, not a substantive one. We could probably trim those introductory progress, but basically what you're saying is those things I think are.
853 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:27:36.240 --> 02:27:55.560 Yeah, more, some stand up, yeah, yeah. Oh paragraph markers. Hey, who style sheet did that? They're not supposed to be copiable. I think you're using the wrong browser. Or maybe I'm using the Oh, NO, it's the GitHub thing.
854 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:27:55.560 --> 02:28:10.620 So that's, that's that proposal, more or less.
855 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:28:10.620 --> 02:28:28.170 I I'm conscious of the time, but if, if folks especially the folks who were very keen to have the the list in the specification, any initial reactions?
856 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:28:28.170 --> 02:28:44.550 I think I do, I do think that the, the, it seems like the historically paragraph kind of cuts perhaps all the wrong ways and I'm.
857 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:28:44.550 --> 02:29:00.510 Happy. I I like the addition I'm happy to remove it with the under with the assumption that we'll be working on bringing something else. Of course, you know, none of this is set in stone. We're just looking for a way to move forward and, and improve the specification over what we have now.
858 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:29:00.510 --> 02:29:16.680 If you can come up with text that we can get comfort around, you know, we're not gonna say NO, we can't consider that, so I don't know. So I think, having some, like if this is the.
859 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:29:16.680 --> 02:29:35.850 That are out here. I think am I missing the like any it's like a. I would put like one clause after that sentence especially when taking the public into your account. I think one of the things you can say here to make the drafting more clear is that.
860 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:29:35.850 --> 02:29:51.630 Individual actors who do this and they have to make their own determination. Like this is I think the piece that is missing, right? Is that like if you are the person doing this work for accessibility, for research, whatever, you have to make your determination so oh that the last paragraph. Not the, not the 1st sentence?
861 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:29:51.630 --> 02:30:11.630 And do not think it does that work. Really? Has a choice? Yes. Can you say more? Make a choice about whether to respect preferences and make that it makes a choice according to their understanding of.
862 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:30:11.630 --> 02:30:28.110 Yes, what they're gonna do and I will, I can, I don't think it does the work. I can try to write something. I just can't quote the words right now like it sounds like it's a matter of redrafting that last paragraph to make it a little crisper, not a fundamental change.
863 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:30:28.110 --> 02:30:44.460 I think it includes some specific examples. So like to the extent that the dis disagreement is out there, take out all specific examples like accessibility and research or not, I think you need to have those, not to say that they are always permitted, but to say those are the types of things you are talking about here and not.
864 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:30:44.460 --> 02:30:59.880 Something else. Actually, I don't think this is talking about anything specific at all. Yeah, it's it's simply saying that if you have information that the person expressing the preference doesn't have.
865 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:30:59.880 --> 02:31:19.880 You use that information in making a decision about whether to respect that preference. I just disagree, I'm sorry can you make a proposal? I know it's not now, but like I think we can discuss it then because I think this text moves us forward. I'm sure.
866 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:31:19.880 --> 02:31:29.040 Sure like we can do like more passes at it and and discuss them as they go because I think it's a strict improvement on what we have today.
867 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:31:29.040 --> 02:31:48.810 Megan I think I'm next. I I think it's a little bit totalogical to say that the like, you know, whether a preference applies to a use depends on how the asset is used and how that usage corresponds to the usage categories and then saying.
868 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:31:48.810 --> 02:32:03.930 But it also depends on the applicable legal context is really intention with the statement just above that these preferences do not create rights obligations or prohibitions. So I I feel like they do not themselves. I don't.
869 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:32:03.930 --> 02:32:20.700 I, I I can see the 1st sentence adding useful clarity, but I think that the others and I and I think that Meredith's concerns about.
870 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:32:20.700 --> 02:32:37.860 Giving a little bit more comfort to users who want like users of the vocabulary, not end users but to services that want to do accessibility or some of those other potentially challengeable activities like could that could be added to the 1st sentence.
871 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:32:37.860 --> 02:32:55.800 We're following from it and we don't need the, the next one. Sorry, I'm not sure I understood the point. Would would you mind just like repeating just I I'm not sure I understood what you meant. Sure, so, saying that an entity will decide.
872 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:32:55.800 --> 02:33:11.100 Like whether a preference applies based on how the asset is used and how the usage corresponds to the usage categories, that's just saying, you know, all of the things that the usage categories have already set, like.
873 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:33:11.100 --> 02:33:28.140 These are the uses in the category. And so I don't think that that is useful edition. Then saying that the entity makes that determination based on the legal context is introducing a notion of legal obligation that we have just disclaimed in the proceeding paragraph.
874 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:33:28.140 --> 02:33:48.140 And so that creates ambiguity and also I think doesn't help. Yeah, do you want to chime in on that? Yeah I don't think we're disclaiming that in the 1st sentence, it's the on its own it does. Like if both in the illegal content.
875 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:33:48.140 --> 02:34:10.280 It's like there's both the situation that AI training doesn't restrict AI training for research purposes if you meet the conditions of being a research organization, you can just ignore that based on that thing. I think that is describing on its own it doesn't say something you can if you assess your legal background.
876 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:34:10.280 --> 02:34:36.480 You understand your legal position in the system, you can ignore it. The other way around if you contact probably some of these qualify as an opt out for the four TM thing and it creates sort of legal liabilities if you don't follow that thing. And I think that is what we're trying to say on itself, they don't. Legal contacts into account.
877 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:34:36.480 --> 02:34:54.840 They do we have to say that? Because the legal context already said that. If we are gonna mention what context is appropriate to take into account, then why is it only the legal context? If Aaron, we need to say this because we have been asked this question again and again and again, and if we fail to address it.
878 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:34:54.840 --> 02:35:13.320 Having a positive right under EU law to engage in a research use of AI is one of the considerations that may take precedence over any state of preference. Right, it's too, too vague. We've had too many questions about this to, to not at least.
879 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:35:13.320 --> 02:35:28.625 Provide some indication of beyond the very abstract thing that you suggested. Thank you. Victoria.
880 "Tori Noble" (1142273792) 02:35:28.625 --> 02:35:36.390 Yeah, I just sort of wanted to echo with the concerns that I'm not sure this is moving in a better direction and I'm also a little.
881 "Tori Noble" (1142273792) 02:35:36.390 --> 02:35:52.740 I'm Concerned about again process is not my role, so I understand that maybe I just don't know what process is, but it seems like it might be more productive to have you know whoever is interested in coming up with an alternative for us to consider.
882 "Tori Noble" (1142273792) 02:35:52.740 --> 02:36:08.176 You know, take time to go do that rather than trying to draft sort of by committee because I think that we're maybe rejecting each other's suggestions preemptively before we see the whole thing. I just think it's hard to do or to write this way.
883 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:36:08.176 --> 02:36:29.510 I think. So I think if you can kind of talk to them and come up with a proposal, I think that would be the best way forward. Yeah, we're definitely not gonna come up with text here. I think it's more we're trying to poke around what the sensitivities are, what the considerations that people have are so that people who do that drafting have a better chance of coming.
884 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:36:29.510 --> 02:36:31.501 If it's something that might succeed.
885 "Tori Noble" (1142273792) 02:36:31.501 --> 02:36:36.958 Perfect. Thank you. I think.
886 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:36:36.958 --> 02:36:53.280 I think that's we could address both sides by just adding, including how the usage corresponds to blah blah blah, the applicable context etc. Because that makes it clear that there are others. I still think they're dropping it, but I understand that's not.
887 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:36:53.280 --> 02:37:13.280 Necessary, but that makes it clear that these aren't the only considerations that there might be other ones. I think the word otherwise allows it one right like at least let's, let's call it there. I.
888 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:37:13.280 --> 02:37:37.760 Clearly we're gonna need more discussion around this area. My only remaining question is, you know, the the current text in the, in the document doesn't seem to be helping. We know that this text isn't there either, but if if we should should we just leave it as is for now? I think we should and but but but this is a distinct improvement of what we.
889 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:37:37.760 --> 02:37:58.860 Previously for sure like yeah. I I'd I'd I'd be supportive of going ahead and merging this as long as there's a very clear indicator that that text is still not done and that we're gonna have issues against it. So I think we have, promises from Meredith and Timod Robot to try to address.
890 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:37:58.860 --> 02:38:17.280 That aspect of it. I think Aaron is gonna give us a an attempt at trying to narrow it, yeah. Clarify it in some way. I don't know what that was, but I did in the chat. I'm asking that you do something more concrete than that and give us.
891 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:38:17.280 --> 02:38:33.450 Our concrete change, same for these guys. Right. What Do you want to merge something to, I think we should merge what has been proposed here with the understanding that there are outstanding changes suggested. Yes on the screen?
892 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:38:33.450 --> 02:38:51.300 Effectively this on the screen. I'll I'll send you a URL. No, I have that already transcribed. Okay, cool. Alright, let's put a pin in that then, and if folks could work on those proposals and bring them to the list.
893 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:38:51.300 --> 02:39:08.220 We've got 20 min left. No, sorry. We have 50 min left. We have, we have 15 min left. If people are willing. Let's move on to the use discussion. We had a long discussion yesterday.
894 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:39:08.220 --> 02:39:25.530 We hopefully have people who are gonna go off and have chats and come up with proposals. Is there anything else we can usefully use this time for to guide that? Especially in light of developments we've had around the training term and around the search term?
895 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:39:25.530 --> 02:39:40.920 And also a discussion of purpose based and so forth and so on. I have on the relevant issue which I under, which was.
896 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:39:40.920 --> 02:40:00.090 Which was it? Category for rack and grounding 172I have made a suggestion for a new category if you scroll all the way to the bottom, 1702?
897 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:40:00.090 --> 02:40:18.120 1702 Yes. Come on.
898 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:40:18.120 --> 02:40:35.610 This is very much intended as a starting point for discussion. It takes the label AI input and the 1st sentence.
899 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:40:35.610 --> 02:40:55.290 A more or less from what's in the RSL Specification uses for the content signals, and then it adds a exclusion that tries to draw a line between assets that are being retrieved by systems.
900 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:40:55.290 --> 02:41:10.530 Versus assets that are being directly provided as input by end users, either as assets or as specific pointers to locations where so it tries to.
901 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:41:10.530 --> 02:41:28.440 Draw this line between trivial augmented generation and user like direct user inputs in order to keep those out of the scope of the control.
902 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:41:28.440 --> 02:41:49.640 Okay. Any initial reactions? What's the name? No, sir. So I think that I do think that it would be helpful for me as, you know, I do, it's when I was thinking of proposing something to understand this, talk a little bit about.
903 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:41:49.640 --> 02:42:11.510 It's end user thing. I understand where it's coming from. The way of many applications are being developed though it becomes another one of these like line blur blurred lines sort of thing because, you know, like, is it user providing it if the application asks, you want me to go out and scrape these 10000 websites?
904 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:42:11.510 --> 02:42:39.800 And, you know, and do this and then you just say yes, and I mean they don't maybe not 10000, but they very frequently are applications that will, you know, suggest, you know, using something. And so it's like, where does the because all actions by any sort of system is directed by human at some level, at the highest level of it. And so we have to draw the line somewhere around there or else it's just literally everything is directed by.
905 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:42:39.800 --> 02:42:57.210 Because because there aren't automated systems that operate in the absence of some sort of human setting them into motion. If I may, in talking to Paul about this one, I think the idea was that, where, where assets are directly provided.
906 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:42:57.210 --> 02:43:12.390 Either, provided or or or referenced, it would be singular assets. So directly identified or directly provided assets as opposed to, yeah.
907 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:43:12.390 --> 02:43:27.510 Like singular things rather than, you know, please go to this website and, and do these things or look at these, the following collection of things. Those would be included in the retrieval side of this rather than directly provided.
908 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:43:27.510 --> 02:43:47.510 Sorry. Does that, does that help? I don't think so because what what is that asset? If that asset is a URL, then it includes everything at that site and just just the things of that URL in that case. So if it's a sing, if it's a single URL, that is a single resource that you're.
909 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:43:47.510 --> 02:44:05.760 Right, you're you're asking the model going and retrieve. Rather than going off and following links and and what have what have you. And we would have to work on I guess Yeah I guess my question is more about what like these systems, what like the the model itself may suggest assets.
910 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:44:05.760 --> 02:44:24.300 And then the user makes sort of confirm that like that this happens a lot with these models where they're asking for a sort of confirmation of do you want me to do this? I mean the intent is to keep that outside, right? Like to say, NO, to keep that inside the AI input category. So like if the system like this, this is a very.
911 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:44:24.300 --> 02:44:42.090 Quick this morning on the way here drawn up thing, we need to draw this line somewhere and the intent is sort of like user directly inputs whether whether system starts selecting something, so that I think that is clearly on the side of what should be.
912 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:44:42.090 --> 02:44:57.570 Within the control of this. So, so, so the scenario that you give where the system stuck like like the the language isn't there yet, but like the intent here is where the system starts suggesting things to you, it's not direct user input anymore.
913 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:44:57.570 --> 02:45:17.570 Yeah, but if but if I say yes, and it responds to a suggestion by the system. Yeah, but it's not directly you say yes to something that's I, yeah, I mean I I think that's it that this is I think we recognize the problem in that I I would also be interested in in prep.
914 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:45:17.570 --> 02:45:40.740 Why it is that the fact that user, a user provides something means we don't want to allow the asset owner to express a preference because e.g., you know, user could provide, could take a photo that a user has posted on their personal website and ask an application to do all sorts of things to that photo and, and.
915 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:45:40.740 --> 02:46:00.740 There might be a reason why the owner of that asset who published it would want to have a preference attached to that asset, that applications which are being used by, by the user could say, NO NO NO, I'm sorry, I can't follow that request. And in fact, that would note that several prominent applications currently do this if you try to upload.
916 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:46:00.740 --> 02:46:30.920 Like a photo of say a famous mouse or something like that. They will say, you know, I can't, I can't modify that because there there is some sort of a sort of, you know, thing in there and so I I would think that we could offer a preference that would extend that same sort of respect or ability to express a preference to everyone even in the user provided context. So who's, so can I just respond to that? Yeah. I think the the you're asking for the intent because.
917 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:46:30.920 --> 02:46:51.360 I I I am very much concerned that, that anything that is going to enable overreach in with regards to that, that will prevent me to do things that I'm legally morally, that's nobody else's business to prevent me from doing to things that I haven't.
918 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:46:51.360 --> 02:47:07.530 Legal access to, and I think we wanna, we wanna find a balance there and we want to introduce safeguards there that sort of, in the absence of such a category, the systems you describe already do that.
919 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:47:07.530 --> 02:47:18.696 I, I feel like I think you made your point. I think we'll go along when we come back to it, right? But we just need to hear other people. So Brad?
920 "Brad Silver" (352249088) 02:47:18.696 --> 02:47:45.050 Thanks. Yeah, I mean this is a, this is a difficult one. I mean Nate hits on a bunch of the issues. I mean, so, you know, I, I I I I think we're sort of closest to addressing the concern where the asset is provided directly by an end user. I think that where it starts to be an asset, in a remote location or one that is facilitated by sort of a scraping activity by.
921 "Brad Silver" (352249088) 02:47:45.050 --> 02:48:05.050 That model, that is that's where things start to get a bit messy for a couple of reasons is that that's exactly how some users are bypassing you know password controls or capture mechanisms that specifically are designed to kind of.
922 "Brad Silver" (352249088) 02:48:05.050 --> 02:48:25.050 Ensure that it's really users that are visiting websites as opposed to, as opposed to bots and crawlers. That's number one. So you may disagree with, with, with that practice, but I just want to put it out there because that's, that's kind of what, what, what, what, what happens and there is an impact on publishers.
923 "Brad Silver" (352249088) 02:48:25.050 --> 02:48:52.800 On that front. The other, other element which we haven't gotten into a lot of or any discussion about is that obviously we're going to be talking in the next phase about attachment mechanisms and robots.TXT and there's you know a stated policy out there from a number of companies that they're fetching.
924 "Brad Silver" (352249088) 02:48:52.800 --> 02:49:12.800 Is not going to recognize robots.txt, as opposed to regular crawling. So I wanna kind of just for shadow that if you think about like because now we're starting we we don't have to have that discussion now obviously that'll be an attachment mechanism discussion, but there's this bifurcation of thinking.
925 "Brad Silver" (352249088) 02:49:12.800 --> 02:49:27.499 Thinking about how content acquisition is happening and what controls that are put around that should, should be respected or not that plays into how we think about a used preference. Thank you.
926 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:49:27.499 --> 02:50:01.299 Thank you, thanks Brad. Brad, I think one of the points of this is like we don't have anything in this regard, right? So like, you know, just Paul is throwing up like a straw man proposal to kind of take a look at. So it's, it's totally fine if like you and Nate I think you promised yesterday to go look at something to make, make a proposal, make that proposal. I, I think like if you cannot get consensus, this doesn't get in, right? Whatever that is. So like, you know, if you want an alternate proposal, please make it. And, and I want to remind folks that, that our our scope here is to allow people to express their preferences, not necessarily to enforce them, yeah.
927 "Brad Silver" (352249088) 02:50:01.299 --> 02:50:11.257 Of course, yeah, I just want to sort of be sort of opening this up so let's let's put all the the different issues on the table for us to, to pick at.
928 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:50:11.257 --> 02:50:30.710 Thank you, Leonard. Yeah, Paul is you probably may guess I have significant objections to 2nd sentence up there. That's something that as you know, from day one of working with you before this committee and on this committee that I have expressed a huge desire.
929 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:50:30.710 --> 02:50:49.470 Based from our customers to do which is that absolutely assets provided as input by end users must have their, I shouldn't say must, let me phrase that, should have their preferences respected when the application in question believes it is appropriate to do so.
930 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:50:49.470 --> 02:51:04.710 So, and this was noted, I forget by whom I apologize in chat that that's NO different than respecting or not respecting those preferences in any other scenario that we are covering, so I would.
931 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:51:04.710 --> 02:51:20.100 Requests very strongly that you remove that second sentence. Thank you Renut. I think we can get back to that so Timoth Robot? Hello, Tim atrobot. I.
932 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:51:20.100 --> 02:51:36.960 Again I think this points to the need for some like top level carve outs so that the, the, the preference can or the category can, can stand on its own.
933 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:51:36.960 --> 02:51:56.960 There is a use case for Greg where it could be used as part of a lack of better term word training algorithm, back propagation, retraining, whatever, that I question if this does.
934 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:51:56.960 --> 02:51:57.600 Doesn't.
935 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:51:57.600 --> 02:52:17.600 Prohibited, I might want to allow rag but not allow train. We've talked a lot about training. I could be wrong, but in retrieving assets into a system that generates an output, that output could be some form of training. And it's not specifically for bidding here. Is it forbidden somewhere else am I missing it or not?
936 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:52:17.600 --> 02:52:19.950 No, that's that's a good point. So.
937 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:52:19.950 --> 02:52:38.670 Justin's point here is that if the, if this particular use involves the refinement of the model, which is often the case in things like chat GPT, then maybe, maybe we need to be careful to exclude the training in in explicit. That's a that's a good piece of input. Can, can I.
938 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:52:38.670 --> 02:52:56.400 If you could, we could say that, if, if the output if the if the AI based system is generating outputs for the purposes of training, please see the training parts. Could you have to follow the training? Could you capture that on the issue? Just as a following comment?
939 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:52:56.400 --> 02:53:15.630 If you can answer something. I always make an attempt. Thank you. I will make an attempt. Go ahead. There, there was discussion earlier in the training definition about guys, Previously we had weights as sort of a stop point where where training ended and I.
940 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:53:15.630 --> 02:53:35.630 I can't remember if we ended up deciding that parameters was gonna be sort of included in there instead, but I I think that if we did something like that, then the logical way to approach this, understanding we have to address sort of the, the, the concerns about overreach, but just from a pure like trying to actually define what we're doing.
941 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:53:35.630 --> 02:53:39.750 Would be to continue that forward and say that everything.
942 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:53:39.750 --> 02:53:59.750 After the weights is this category is inference or whatever this is. And then we can have, and you could do that with one sentence by just basically modifying the the training definition that we have and then just applying it to everything kinda after that. And then we could have the debates about like, ok, what, where, where should this apply to.
943 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:53:59.750 --> 02:54:06.330 User initiated things and if it shouldn't, maybe that can be a two categories, one could be for that. We could talk about that.
944 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:54:06.330 --> 02:54:22.380 I mean I realized that's opening up a can of worms, but you could have a way to express preference for those user things and maybe some applications would be more likely to to ignore that preference than to ignore a preference big, you know, than to choose to ignore a preference that's that applies.
945 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:54:22.380 --> 02:54:37.410 To non user, you know, to system or whatever else you wanna call it. I, I don't know if that approach might be productive, but it's certainly from the draft that we have trying to have consistency of approach, I think would be the simplest approach.
946 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:54:37.410 --> 02:54:57.410 I think where we ended up was that we were gonna remove weights or parameters and just talk about use of the data to that that you know has an effect on the model, but to change the model either in its production or subsequently so use of the data that does not change. Exactly, that that would be the mirror. Yeah. And so it just where wherever that line is at the end of training, we just can't.
947 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:54:57.410 --> 02:55:17.410 Forward and then we have a debate about what's overreached, and I tend to agree with the like the approach of I think these things are best resolved in some sort of like the discussion we just had about sort of global exceptions, like if you don't want to use the preference because of whatever or if you want, like, there maybe things that motivate you to use or ignore a preference and.
948 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:55:17.410 --> 02:55:28.830 And I I think that we keep having that same discussion with every single vocabulary issue I think it's better done at the at the level up in the hierarchy. I think the three or two is outside, right? Like all the specific.
949 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:55:28.830 --> 02:55:43.890 Bugs pretty much so could could you as well try and capture that on the issue? I think that's a that's a great insight as well. We're on yep 1702. Okay.
950 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:55:43.890 --> 02:56:03.660 That looks like we've exhausted the queue, but I don't think we have any other proposals on the table. No, I obviously work needs to happen. I I'm inclined to perhaps go ahead and break. If people want to use time to have lunch, that's great if people want to use time to work on proposals or talk to people, that's great.
951 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:56:03.660 --> 02:56:23.660 As we said yesterday, we're planning to have an online only interim in roughly the June timeframe, then we'll meet in Vienna. Some, some folks pointed out that the US Labor Day is unlike the rest of the planet in September so we will we'll try.
952 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:56:23.660 --> 02:56:33.210 Try to be kind to those although I can't tell you how many australian holidays I've traveled on for standards work, but NO, I'm not better. We'll we'll we'll try the last week of August.
953 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:56:33.210 --> 02:56:48.420 If that causes folks concern, please get that to me and Suresh as quickly as possible and we'll start to nail those dates. This is gonna be a hybrid. Which days do you mean when you say the last week? Last week?
954 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:56:48.420 --> 02:57:08.420 That would lead us to if we follow our normal pattern, I think we're looking at the week of 2520 06:27 or perhaps 2620 07:28. August. Yes.
955 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:57:08.420 --> 02:57:14.250 The week of August 24, right? Let's, let's not pick out exact days right now, but the week of 24 August. You're thinking here.
956 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:57:14.250 --> 02:57:32.700 We're thinking Europe yes. Sorry Europe or the United Kingdom. That's their fault. You are you looking for a host? We're looking for, if people have, and again, you know, we have fairly constrained requirements. If people have facilities that they think are appropriate in terms of.
957 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:57:32.700 --> 02:57:48.540 Size, nearness to hotels and restaurants, nearness to international airports. I think that's especially important because people do come a long way, lack of an NDA, good networking, so forth and so on. Please come to Suresh and I We do have.
958 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:57:48.540 --> 02:58:05.880 Yeah, that wasn't a dig at Cisco, sorry. We, we do have possible hosts in Europe/UK already, so it's not urgent, but if people want to put their hands up, please come talk to us and we'll we'll try and find the best place of balancing is not critical.
959 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:58:05.880 --> 02:58:25.880 But it maybe urgent because you want to you wanna close the deal relatively soon. Yeah, sorry. It is not critical that we find someone, but if we if if you wanna put your hand up, do so quickly because we would rather get this settled soon rather than wait for the last minute that has not been good in the past. So in the next couple of weeks, if you could the most we'd want to hear about that.
960 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:58:25.880 --> 02:58:36.060 A number of people last night expressed reservations about the cloud fair option that we met in the past in terms of things like the acoustics in the room or what have you.
961 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:58:36.060 --> 02:58:56.060 Right. So there are always these sorts of things with the I find this room to be too large, but it's pretty good otherwise. It is incredibly rare to find the perfect room, and actually the best rooms I've been in tend to be in universities that have a specific kind of lecture hall. They they're different.
962 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:58:56.060 --> 02:59:04.440 Names for it, but it's basically a U shaped tiered set up, it's fantastic. They're very hard to find. Although in August maybe we'll find a university. I don't know. Maybe.
963 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:59:04.440 --> 02:59:24.440 In rear Harris, what's that exact setup? Well, off you go. Talk to some folks. For that matter, you know, well, NO. Great greatly appreciated. It would. Yeah. See what you can do. Scare that out. But please get to us in the next couple of weeks. Otherwise, we have a couple of default.
964 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:59:24.440 --> 02:59:28.950 Applications that we know we can use. So we'll we'll we'll fall back to one of those.
965 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:59:28.950 --> 02:59:48.950 Bump, the balance would be nice too. But yeah, acoustics is often a very good point. We've been look I remember Missilla Paris, that was a great location with horrible acoustics, so, and that was a smaller group. Exactly. All right, so thank you all for your patience and your your listening and.
966 "TRN6-29-BANFF/speaker_1" (890799616_1) 02:59:48.950 --> 03:00:00.630 And willingness to consider other viewpoints. I think it's been a pretty productive meeting. Let's continue this on the list. We need to make progress outside of these meetings if we want to finish while there are still.
967 "TRN6-29-BANFF/speaker_1" (890799616_1) 03:00:00.630 --> 03:00:20.630 Two twos in the in the year. Thank you to Cisco for hosting us. This has been fantastic. I think everyone's had a good time and we'll see each other online and hopefully in either Vienna or somewhere else in europe/UK. Right.
968 "TRN6-29-BANFF/speaker_1" (890799616_1) 03:00:20.630 --> 03:00:40.630 And thanks to everyone for being like super well behaved and like like being very like empathetic of like other people's concern. I think that's kind of like we kind of crossed the line there so like people can see other people's point of view. I think like it took a lot of time to get built up to that so and being constructive I think a lot of constructive discussion now we're still.
969 "TRN6-29-BANFF/speaker_1" (890799616_1) 03:00:40.630 --> 03:00:43.260 Progress.
970 "TRN6-29-BANFF/speaker_1" (890799616_1) 03:00:43.260 --> 03:01:01.020 Thanks Alan, thanks for everybody online, whatever your time zone is, like so thanks for being very flexible on this. Thanks. All right. Miss you Chris, and hope to see you soon.
971 "TRN6-29-BANFF/speaker_1" (890799616_1) 03:01:01.020 --> 03:01:06.000 I put that in the.