WEBVTT Kind: captions Language: en 00:00:00.000 --> 00:00:02.960 Welcome you all to the second panel 00:00:04.680 --> 00:00:06.720 of the first day and 00:00:09.280 --> 00:00:10.480 this panel 00:00:11.000 --> 00:00:14.400 will aim to discuss 00:00:14.960 --> 00:00:19.440 the current status of natural language processing techniques. 00:00:20.000 --> 00:00:25.200 And during the context of the the web, we can think 00:00:25.480 --> 00:00:27.680 or we know that they can be used 00:00:28.200 --> 00:00:31.440 to generate textual descriptions for images, 00:00:31.440 --> 00:00:36.440 but also for other visual media that's presented on web pages 00:00:37.160 --> 00:00:40.880 and we will focus today our discussion on 00:00:41.840 --> 00:00:45.480 or start to consider aspects such as text adaptation 00:00:45.480 --> 00:00:51.480 and how to provide understandable text to better meet the web user needs. 00:00:51.480 --> 00:00:55.880 And they're in different contexts of of use and also what 00:00:56.160 --> 00:01:00.000 our future perspective for natural language 00:01:00.000 --> 00:01:03.240 processing on web accessibility 00:01:03.240 --> 00:01:05.400 are to support web accessibility. 00:01:06.720 --> 00:01:09.600 So I'm glad to welcome back 00:01:10.720 --> 00:01:13.000 Michael, Shivam and Amy. 00:01:13.200 --> 00:01:16.320 And when 00:01:16.320 --> 00:01:18.840 you are Amy and also 00:01:19.560 --> 00:01:23.600 to welcome Shaomei Wu from AImpower 00:01:26.160 --> 00:01:27.480 who agreed to join 00:01:27.480 --> 00:01:30.920 us on the second panel of the day. 00:01:31.360 --> 00:01:33.840 Welcome you all back and welcome Shaomei. 00:01:34.760 --> 00:01:36.960 For your first intervention 00:01:36.960 --> 00:01:41.400 I ask you just to briefly introduce yourself 00:01:42.000 --> 00:01:44.240 and your three other co panelists 00:01:45.440 --> 00:01:47.600 already done that on the previous panel. 00:01:47.600 --> 00:01:51.400 So no need to to reintroduce yourselves. 00:01:52.800 --> 00:01:57.520 But I will start by 00:01:59.840 --> 00:02:02.760 thinking about once again the quality. 00:02:02.760 --> 00:02:05.560 So we go back to the to the quality topic, 00:02:06.240 --> 00:02:09.040 but now the quality of machine generated descriptions 00:02:11.040 --> 00:02:13.920 now no longer from the perspective of image processing, 00:02:14.200 --> 00:02:18.240 but from the perspective of the natural language generation, right? 00:02:18.240 --> 00:02:21.480 So how can we improve the quality 00:02:21.480 --> 00:02:26.760 of this machine generated descriptions, especially taking into account 00:02:26.880 --> 00:02:30.320 the personalized preferences from users? 00:02:30.760 --> 00:02:33.360 And I will start with you Shaomei 00:02:36.720 --> 00:02:39.360 Hello. So 00:02:41.880 --> 00:02:44.160 thank you all for 00:02:45.160 --> 00:02:47.640 having me here 00:02:47.640 --> 00:02:48.200 today. 00:02:48.200 --> 00:02:53.160 And my name is Shaomei Wu and right now 00:02:53.160 --> 00:02:57.480 I'm the founder and CEO of AImpower.org 00:02:58.080 --> 00:03:02.160 Tech non profit that researches 00:03:02.160 --> 00:03:05.320 and co-create 00:03:06.080 --> 00:03:09.800 in-powering technology for 00:03:09.840 --> 00:03:13.640 other ways marginalized users. 00:03:13.640 --> 00:03:15.800 And first of all, I want to 00:03:16.960 --> 00:03:20.400 and also share, 00:03:20.440 --> 00:03:26.200 that I do have a stutter so that you may hear 00:03:26.440 --> 00:03:30.720 there'll be more pauses and ... when I talk 00:03:32.080 --> 00:03:33.920 and before AImpower. 00:03:33.920 --> 00:03:37.840 I was a research scientist at Facebook 00:03:38.920 --> 00:03:42.080 leading a lot of research and 00:03:43.680 --> 00:03:46.120 product work on 00:03:46.760 --> 00:03:50.480 accessibility, inclusion and equity. 00:03:51.160 --> 00:03:56.400 So one of those product that I shipped is 00:03:56.400 --> 00:03:59.240 automatic alt texts 00:04:02.840 --> 00:04:07.520 and these you're allowed to we will provide 00:04:08.080 --> 00:04:10.320 short and 00:04:11.440 --> 00:04:13.560 machine generated 00:04:13.560 --> 00:04:16.440 description of 00:04:17.960 --> 00:04:21.800 images on Facebook and Instagram 00:04:23.160 --> 00:04:28.880 to screen reader users in real time. 00:04:29.880 --> 00:04:31.880 So when it comes to 00:04:31.880 --> 00:04:35.280 quality of automated alt text 00:04:35.840 --> 00:04:39.280 or other similar 00:04:40.480 --> 00:04:41.880 systems, 00:04:42.240 --> 00:04:44.400 we saw two kind of biggest 00:04:45.480 --> 00:04:50.280 area of development that we wanted to do. 00:04:50.760 --> 00:04:54.240 And the first one is accuracy, 00:04:54.320 --> 00:04:57.480 which I think we talked a lot about 00:04:57.480 --> 00:05:00.960 in the last panel as well. 00:05:02.520 --> 00:05:05.120 But I want to talk 00:05:05.120 --> 00:05:09.840 it'll be more about the second one, which is the 00:05:10.880 --> 00:05:12.960 richness of those 00:05:14.000 --> 00:05:16.000 descriptions. 00:05:16.760 --> 00:05:20.520 So to be honest, 00:05:20.760 --> 00:05:26.480 like the alt texts that we generated 00:05:26.520 --> 00:05:29.560 was quite 00:05:29.880 --> 00:05:32.320 limited, you know, a lot of 00:05:33.440 --> 00:05:37.880 users they will say that it's more like 00:05:38.560 --> 00:05:40.680 teasers, kind of, you know, 00:05:41.960 --> 00:05:44.880 and telling you like, oh yeah, five 00:05:45.720 --> 00:05:49.200 people smiling, pizza, indoor. 00:05:49.960 --> 00:05:51.600 But no more than that. 00:05:51.600 --> 00:05:59.880 No more than what kind of pizza and what kind of like 00:06:00.520 --> 00:06:03.000 indoor environment. Is it at home? 00:06:03.000 --> 00:06:06.200 Is it like a restaurant? 00:06:06.200 --> 00:06:12.720 So I think our users really, when they kind of gathered all of the 00:06:13.000 --> 00:06:17.520 richness of what someone who has eyesight can see, 00:06:17.880 --> 00:06:20.320 you know, can handle access. 00:06:21.120 --> 00:06:24.320 So one particular kind 00:06:26.240 --> 00:06:31.200 an area that users do really want to know more 00:06:31.360 --> 00:06:37.000 is it's about people who they are, how do they 00:06:38.480 --> 00:06:39.120 look? 00:06:39.920 --> 00:06:44.280 Look like... race, gender and 00:06:45.320 --> 00:06:47.640 even how 00:06:48.320 --> 00:06:49.840 attractive they are 00:06:49.840 --> 00:06:53.120 because that is something really kind of socially salient. 00:06:53.880 --> 00:06:58.960 So that was kind of big challenge for us. 00:06:59.040 --> 00:07:01.920 One way was designing our 00:07:03.560 --> 00:07:07.040 our system 00:07:08.280 --> 00:07:11.520 because, you know, like, how can we share 00:07:11.520 --> 00:07:14.560 those kind of attribute in a, like, 00:07:15.520 --> 00:07:19.720 you know, both accurate and kind of socially 00:07:20.880 --> 00:07:22.400 conscious way. 00:07:22.400 --> 00:07:26.840 So we actually opt to not 00:07:28.080 --> 00:07:29.120 kind of 00:07:30.120 --> 00:07:34.200 showing like the race and gender of the, 00:07:35.160 --> 00:07:39.720 the people being photographed, 00:07:40.640 --> 00:07:43.640 but which I you know, 00:07:43.920 --> 00:07:46.600 we got actually a lot of complaint on 00:07:47.160 --> 00:07:52.200 but how to kind of yeah convey those in a kind of socially 00:07:53.400 --> 00:07:57.040 respectful way I think it's something we 00:07:57.840 --> 00:08:00.520 we should really work like work on 00:08:00.960 --> 00:08:07.840 and now I can see kind of like a few ways that we can make that better. 00:08:07.880 --> 00:08:12.080 For example, like, considering the like... 00:08:16.800 --> 00:08:18.480 relationship 00:08:18.480 --> 00:08:21.760 between kind of people in the photo and 00:08:23.040 --> 00:08:27.120 yours, for example, like if they are friends 00:08:27.120 --> 00:08:32.600 and then we can just tell them the name, and, you know, other things about those people 00:08:32.920 --> 00:08:36.320 and then another thing is to kind of, give people 00:08:38.560 --> 00:08:41.840 progress details. 00:08:42.040 --> 00:08:48.200 So then you know, I have some kind of like option to kind of allow the consumer 00:08:48.200 --> 00:08:52.200 of those alt texts to kind of request more 00:08:54.840 --> 00:08:57.960 details that we just cannot provide 00:08:57.960 --> 00:09:02.840 by our systems. 00:09:03.680 --> 00:09:07.800 So I will be done here and let 00:09:09.280 --> 00:09:12.560 others talk. 00:09:12.560 --> 00:09:12.960 Thank you Shaomei. 00:09:12.960 --> 00:09:16.680 Shivam, your thoughts on 00:09:17.080 --> 00:09:21.680 how can we improve the quality of machine generated descriptions? 00:09:23.080 --> 00:09:26.280 Okay, so this is a two part thing. 00:09:26.280 --> 00:09:29.080 So when you come to technically implementing more, say 00:09:30.000 --> 00:09:32.400 how you have design your model, how you train them, 00:09:32.720 --> 00:09:35.800 and what the whoever the stakeholders of designing a particular model 00:09:35.800 --> 00:09:39.760 is very much necessary in how and when to get the quality machine 00:09:39.760 --> 00:09:41.200 generated description. 00:09:41.200 --> 00:09:44.400 Now when you when we take into account users personalized 00:09:44.400 --> 00:09:47.520 preferences, this is a two part... 00:09:47.560 --> 00:09:50.320 So first, let's take an example. 00:09:50.320 --> 00:09:54.360 I am a person who knows Spanish and my model, 00:09:54.800 --> 00:09:58.360 a very famous model gives descriptions in English right, now 00:09:58.720 --> 00:10:01.560 that model or whatever the consumption of that model 00:10:01.680 --> 00:10:04.600 is like let's say you're using an API to consume the model. 00:10:05.040 --> 00:10:08.680 So that should take into account the user's personalized preferences 00:10:08.680 --> 00:10:12.400 of his language and write the output based on that as well. 00:10:12.760 --> 00:10:17.080 So this diversity of a model to prepare output in multiple 00:10:17.080 --> 00:10:21.280 formats, multiple languages is something that can be looked into. 00:10:21.280 --> 00:10:25.160 This is how the quality of the machine generated description increases. 00:10:25.440 --> 00:10:28.600 Now, you did not train the complete model separately. 00:10:28.600 --> 00:10:29.640 What you can do 00:10:29.640 --> 00:10:33.440 is just a post-processing scripts for your models and that can help. 00:10:33.440 --> 00:10:38.680 And you it's not much of an effort when we say as a model training input, 00:10:38.680 --> 00:10:42.600 but it's very simple solution to what can be a best solution. 00:10:42.920 --> 00:10:45.520 The other thing is how you prepare your quality data. 00:10:45.520 --> 00:10:50.760 Now you should fully carefully categorize it is strictly if needed. 00:10:51.080 --> 00:10:55.640 And let's say you have input data that are blurred images and all sorts of thing. 00:10:55.920 --> 00:10:59.560 So you need to carefully prepare your model, creating data and 00:10:59.800 --> 00:11:02.360 based on that data, your description would be 00:11:02.640 --> 00:11:07.400 a bit more clearer and the population will also be factored in. 00:11:07.400 --> 00:11:08.280 When you 00:11:08.400 --> 00:11:13.000 look into how you can both process your data for a certain group of people. 00:11:13.760 --> 00:11:17.160 So that's how I see it. 00:11:17.160 --> 00:11:17.760 Thank you. 00:11:17.760 --> 00:11:21.880 And Amy, want to share your experiences? 00:11:22.920 --> 00:11:23.400 Sure. 00:11:23.400 --> 00:11:25.920 So a couple of ways that I've seen 00:11:26.720 --> 00:11:31.440 that I think are sort of promising maybe to use NLP to improve quality. 00:11:31.800 --> 00:11:34.600 One thing I started seeing recently is people, 00:11:35.200 --> 00:11:39.080 you know, starting to consider context around the 00:11:39.480 --> 00:11:42.280 the image that's going to be described, to maybe 00:11:42.280 --> 00:11:45.880 create a description that's more that's more helpful. 00:11:45.880 --> 00:11:47.760 So imagine, you know, 00:11:47.760 --> 00:11:51.720 someone writes a post on Twitter and they couple that post with an image. 00:11:51.880 --> 00:11:57.560 So considering the the post and the image together, maybe might inform 00:11:57.720 --> 00:12:01.080 models on how to create something that's more informative. 00:12:01.080 --> 00:12:05.600 So for instance, if I posted a picture of myself 00:12:06.440 --> 00:12:08.600 snowboarding and I said I learned a new trick, 00:12:08.600 --> 00:12:11.760 then it might be important to tell me what trick you learned. 00:12:12.320 --> 00:12:15.560 Whereas on the other hand, I said, I just went on vacation. 00:12:15.560 --> 00:12:17.800 You know, the exact trick might not matter as much. 00:12:18.360 --> 00:12:22.560 And so I think that the idea of like using language understanding 00:12:22.560 --> 00:12:25.920 to get more information about the context before making a prediction is promising 00:12:26.560 --> 00:12:27.480 in another way. 00:12:27.480 --> 00:12:30.600 I've sort of seen it used to maybe improve the quality kind of 00:12:30.640 --> 00:12:31.680 goes back to the other 00:12:32.840 --> 00:12:36.080 the other answers that were given so maybe you can use question 00:12:36.080 --> 00:12:39.080 answering about the image to gain more information when you need it. 00:12:40.440 --> 00:12:42.480 Oh, one thing 00:12:42.480 --> 00:12:48.040 I've also thought about is seeing if maybe users could give examples 00:12:48.040 --> 00:12:52.640 or their preferences about descriptions in in natural language. 00:12:52.640 --> 00:12:55.800 So this is an example of a description maybe we can copy the style 00:12:55.800 --> 00:12:59.160 of this description when we're applying it to other descriptions. 00:12:59.400 --> 00:13:03.960 So maybe I like to hear about the costumes someone wears in a 00:13:04.920 --> 00:13:08.320 in in a video, and I wish that future descriptions 00:13:08.800 --> 00:13:12.560 might include more information about that rather than summarizing them. 00:13:12.880 --> 00:13:16.400 And then finally, one other way I've seen, 00:13:17.240 --> 00:13:20.120 I've used NLP to improve 00:13:20.120 --> 00:13:23.000 quality, is also based on summarization. 00:13:23.640 --> 00:13:27.520 So there can be times when there's more to describe than time 00:13:27.520 --> 00:13:28.680 you have to describe it. 00:13:28.680 --> 00:13:32.720 So especially in videos, there's often a really small amount 00:13:32.720 --> 00:13:36.120 of time to describe without overlapping the other audio. 00:13:36.640 --> 00:13:42.080 So one way you can use use NLP to improve the quality 00:13:42.360 --> 00:13:45.320 is by trying to summarize those descriptions so they fit 00:13:45.320 --> 00:13:49.640 within the time you have and they don't decrease the experience of people 00:13:50.560 --> 00:13:53.080 trying to watch the video and hear the audio at the same time. 00:13:53.080 --> 00:13:55.560 Yeah, yeah. 00:13:55.560 --> 00:14:00.600 That's that's definitely a good use for NLP. 00:14:01.440 --> 00:14:03.800 Michael, still in this topic 00:14:03.920 --> 00:14:08.560 and I would like to have your perspective 00:14:08.560 --> 00:14:12.560 on initiatives from WAI that 00:14:14.200 --> 00:14:17.000 might assist users in 00:14:17.400 --> 00:14:19.920 providing their preferences 00:14:20.760 --> 00:14:23.560 so that eventually 00:14:23.560 --> 00:14:26.760 models can use those 00:14:27.600 --> 00:14:33.440 or anything that might be ongoing in that regard. 00:14:33.440 --> 00:14:34.200 Okay. 00:14:34.560 --> 00:14:39.600 So first of all, just give the disclaimer for anybody new to this session 00:14:39.600 --> 00:14:42.320 that I'm not a machine learning professional, I'm speaking from the 00:14:43.080 --> 00:14:46.080 of my work and the Web accessibility initiative. 00:14:46.440 --> 00:14:47.640 I do want to talk briefly. 00:14:47.640 --> 00:14:52.280 I think the other panelists covered almost anything 00:14:52.280 --> 00:14:54.760 I would have said 00:14:54.920 --> 00:15:00.720 one thing that based on my knowledge of how machine learning works generally 00:15:00.720 --> 00:15:04.080 today, our models tend to be focused 00:15:04.080 --> 00:15:06.480 on, you know, of our particular ability, 00:15:07.680 --> 00:15:09.080 and it's not universal. 00:15:09.080 --> 00:15:13.640 And the future models will have more abilities combined. 00:15:13.640 --> 00:15:18.440 But, you know, so there may be one model that can recognize this is a human. 00:15:18.440 --> 00:15:22.440 And here are those attributes, another one that you can say this is this human, 00:15:23.280 --> 00:15:26.800 and yet another one that can say this human plus that human 00:15:26.800 --> 00:15:28.320 equals this relationship. 00:15:29.320 --> 00:15:29.920 So all of 00:15:29.920 --> 00:15:32.360 that information, I believe is separate right now. 00:15:33.000 --> 00:15:36.800 So the ability for models to share contexts, 00:15:37.080 --> 00:15:40.960 I think is going to be a part of the solution that we need. 00:15:40.960 --> 00:15:47.760 So what I can speak up for in the Web accessibility initiative. 00:15:47.760 --> 00:15:52.680 So we are only beginning to explore 00:15:52.960 --> 00:15:57.120 what AI and accessibility means. 00:15:57.120 --> 00:16:00.200 And so this symposium is a part of that process. 00:16:01.160 --> 00:16:04.400 We have a practice of doing 00:16:04.920 --> 00:16:07.520 research papers, sort of literature reviews, 00:16:08.040 --> 00:16:11.000 and then proposing accessibility user requirements. 00:16:11.000 --> 00:16:16.120 So that would be something that, you know, we could be working on to 00:16:17.160 --> 00:16:18.680 start gathering this information. 00:16:18.680 --> 00:16:22.640 And from there we decide what to do, whether the content goes into guidelines 00:16:22.640 --> 00:16:26.000 or into new technologies or whatever. 00:16:26.000 --> 00:16:30.840 But I think most of the resources around AI are 00:16:31.760 --> 00:16:36.200 would fit into new resources for those categories. 00:16:36.200 --> 00:16:38.520 Okay, great. Thanks. And 00:16:39.480 --> 00:16:42.920 I would like now to move on to addressing something 00:16:42.920 --> 00:16:47.400 that was basically the core of Jutta’s keynote. 00:16:47.400 --> 00:16:48.720 So it's 00:16:49.080 --> 00:16:52.800 discrimination bias or any other type of bias. 00:16:53.080 --> 00:16:55.280 And here 00:16:55.720 --> 00:16:58.760 also looking at something that Antonio Gambabari 00:16:59.160 --> 00:17:02.640 has entered in the Q&A for the previous panel. 00:17:02.640 --> 00:17:04.680 But I think it's also a 00:17:05.760 --> 00:17:07.840 very well fit. 00:17:07.840 --> 00:17:12.280 It fits very well into this topic and it brought out 00:17:13.760 --> 00:17:15.320 the use 00:17:15.960 --> 00:17:19.280 large language models, LLMs 00:17:19.280 --> 00:17:21.560 and which are currently 00:17:22.160 --> 00:17:25.240 getting a lot of traction and a lot of spotlight. 00:17:25.760 --> 00:17:29.120 And and do you think these LLMs 00:17:30.240 --> 00:17:33.560 can open up 00:17:33.600 --> 00:17:39.000 new avenues, as Antonio Gambabari was mentioning, for reducing 00:17:39.320 --> 00:17:42.480 the different type of bias that we see 00:17:43.160 --> 00:17:48.480 as a result of the use of AI trained models? 00:17:49.240 --> 00:17:53.280 And Shivam you want to go first, this time? 00:17:53.280 --> 00:17:54.680 Yeah, sure, sure. 00:17:54.680 --> 00:17:57.880 So this is quite a question which is 00:17:58.160 --> 00:18:01.760 and has been close to my heart as well, how can we address social 00:18:01.760 --> 00:18:05.600 bias, in largely any model. As part of industry 00:18:05.600 --> 00:18:08.960 have seen a lot of ML models trainings how the output comes. 00:18:09.360 --> 00:18:11.960 So social ML model results of data that they have 00:18:12.440 --> 00:18:15.920 and how the social attitudes are represented within that model. 00:18:16.520 --> 00:18:20.520 And most of that available data is used between models publicly sold, 00:18:21.280 --> 00:18:23.840 which continuously degree of bias that you can see, 00:18:23.840 --> 00:18:27.480 because most of the data that are generated on Internet is basically 00:18:28.480 --> 00:18:30.280 those people who can consume it, right? 00:18:30.280 --> 00:18:35.160 It it's not that everybody once would who doesn't even know what it is. 00:18:35.400 --> 00:18:36.840 They cannot create data over there. 00:18:36.840 --> 00:18:42.040 So most of the data that is available to train the model, it's built out of that. 00:18:42.040 --> 00:18:45.920 So that's how you see a bias in one way. The other 00:18:45.920 --> 00:18:49.720 instance I can give an example is you will see a lot of street violence, 00:18:49.720 --> 00:18:53.760 homelessness, drug overdose and, all those things overrepresented 00:18:54.080 --> 00:18:58.280 in the text that discuss mental illness, although these both are not similar, 00:18:58.960 --> 00:19:01.360 but you will find this kind of representation 00:19:01.600 --> 00:19:03.720 in ML outputs like 00:19:05.040 --> 00:19:06.080 how can we address this? 00:19:06.080 --> 00:19:10.040 Now there there is a novel way of human in the loop. 00:19:10.040 --> 00:19:14.480 Our human feedback loop on an existing models where you can provide some feedback 00:19:14.480 --> 00:19:18.840 to the already existing model that this is the sort of output is not correct. 00:19:18.840 --> 00:19:22.600 This can be a correct version and this can be a good version some some human 00:19:23.600 --> 00:19:24.480 interface 00:19:24.480 --> 00:19:24.920 is needed or 00:19:24.920 --> 00:19:29.280 what that now the other aspect of it is the representational 00:19:29.280 --> 00:19:30.320 training of ML models. 00:19:30.320 --> 00:19:32.320 Now, the underlying data is models. 00:19:32.560 --> 00:19:34.400 It's the main source of the issue here. 00:19:34.400 --> 00:19:37.960 So you need to correctly source your data and at least up to date, 00:19:38.800 --> 00:19:41.320 you're not all representing one section of data. 00:19:41.600 --> 00:19:44.000 For example, let's say you have a bigger society. 00:19:44.560 --> 00:19:48.920 This society can be underprivileged, overprivileged and maybe some rigid persons. 00:19:48.920 --> 00:19:50.960 They just account society. 00:19:50.960 --> 00:19:53.800 Now, you cannot just take the data from one section of society and train 00:19:53.840 --> 00:19:57.680 the model and say that I can give you a complete picture of the area. 00:19:58.160 --> 00:20:00.920 There's much a separate section which are underrepresented. That's 00:20:00.920 --> 00:20:04.600 what is happening with all the models right from the start of ML. 00:20:05.040 --> 00:20:10.720 You can see. Now what we can also do to mitigate is you can create inclusive buckets 00:20:10.720 --> 00:20:13.960 where the developer of ML models or designer of ML models 00:20:13.960 --> 00:20:17.080 you can give the inclusive bucket training to them. 00:20:17.080 --> 00:20:17.440 You can 00:20:18.720 --> 00:20:19.160 get them 00:20:19.160 --> 00:20:22.800 aware that what is happening and what can how we can mitigate this. 00:20:23.000 --> 00:20:26.680 So all the person who are included in ML generation 00:20:26.720 --> 00:20:31.120 or there are a lot of you still going on, I mean a lot of data extraction goes on. 00:20:31.320 --> 00:20:33.640 So all those people can be trained for inclusiveness. 00:20:34.520 --> 00:20:36.920 There are multiple tools that help us do that. 00:20:37.600 --> 00:20:39.760 Like if you are creating a model, you can test in 00:20:39.960 --> 00:20:42.800 Google helps us in 00:20:43.080 --> 00:20:43.680 analyzing the 00:20:43.680 --> 00:20:46.320 models like Google has a lot of tools, AI fairness, 00:20:46.440 --> 00:20:49.360 So how your models are performing when talk about 00:20:49.360 --> 00:20:52.560 a lot of including inclusive outputs of your data. 00:20:53.440 --> 00:20:55.960 Also you need to do a thorough testing of your models 00:20:55.960 --> 00:21:02.040 whenever you go ahead to include that all the outputs are properly 00:21:02.840 --> 00:21:06.280 aligned, properly represented, all the sections of your model 00:21:06.560 --> 00:21:10.240 which it is intended to be used should be represented well. 00:21:10.480 --> 00:21:14.520 Your testing should be that model in case of any models you're creating, 00:21:15.000 --> 00:21:16.880 because now we're not at that stage that 00:21:17.840 --> 00:21:19.080 AI and ML is 00:21:19.080 --> 00:21:21.840 in the starting off state, it's quite mature 00:21:21.840 --> 00:21:23.080 right now. We are seeing 00:21:23.080 --> 00:21:26.280 a lot of breakthrough technology so we can do this going forward. 00:21:26.280 --> 00:21:28.440 I guess this can be a solution. 00:21:30.000 --> 00:21:31.800 Okay, thank you. 00:21:31.800 --> 00:21:36.400 Shivam. Shaomei, can we have your input on how 00:21:36.480 --> 00:21:41.200 can we address that social bias or other types of bias? 00:21:41.200 --> 00:21:44.160 Yeah. So 00:21:44.520 --> 00:21:44.880 yeah. 00:21:44.880 --> 00:21:48.120 So then on these, I want to kind of go back to 00:21:48.160 --> 00:21:52.400 I just kind of talk about before in 00:21:53.520 --> 00:21:55.640 particular on 00:21:56.400 --> 00:21:58.560 sensitive social 00:21:58.880 --> 00:22:03.400 identities, you know, about people 00:22:03.400 --> 00:22:07.080 on the photos. 00:22:07.080 --> 00:22:10.200 I, I don't see I kind of the 00:22:10.200 --> 00:22:12.480 way for the 00:22:13.880 --> 00:22:16.920 a current 00:22:16.920 --> 00:22:19.080 and machine learning 00:22:21.000 --> 00:22:23.240 system to kind of accurately 00:22:24.320 --> 00:22:26.880 come out with those labels. 00:22:26.880 --> 00:22:31.800 I think the key kind of issue here is a lot of those 00:22:33.760 --> 00:22:34.720 systems will 00:22:34.720 --> 00:22:37.560 kind of really assume these like fixed 00:22:37.720 --> 00:22:41.200 and definite need 00:22:43.800 --> 00:22:47.560 these kind of social categorizations 00:22:48.840 --> 00:22:53.800 such as race and gender. 00:22:53.800 --> 00:22:56.840 So I think maybe we should think 00:22:56.840 --> 00:22:59.240 be an kind of 00:23:00.320 --> 00:23:03.720 a machine learning systems 00:23:03.720 --> 00:23:05.760 and kind of find a way to 00:23:07.320 --> 00:23:10.000 to kind of attribute people 00:23:10.880 --> 00:23:13.800 respective race back fully 00:23:15.520 --> 00:23:17.920 through by doing this kind of like 00:23:17.920 --> 00:23:20.600 having to like agencies 00:23:21.000 --> 00:23:23.840 of those being kind of 00:23:25.880 --> 00:23:29.280 photographed and being 00:23:29.520 --> 00:23:31.040 described. 00:23:31.040 --> 00:23:35.480 For example, I think now a lot of people has been kind of 00:23:36.560 --> 00:23:38.960 specifying their 00:23:39.760 --> 00:23:42.720 pronouns, for example in their 00:23:42.720 --> 00:23:46.360 social media bios 00:23:46.800 --> 00:23:49.960 and I think those kind of 00:23:51.000 --> 00:23:52.680 information should be 00:23:52.680 --> 00:23:55.720 made use of or could be kind of 00:23:56.040 --> 00:23:59.000 made use of one way 00:23:59.000 --> 00:24:01.800 of assigning on, you know, 00:24:01.800 --> 00:24:04.760 one way to describing 00:24:04.760 --> 00:24:07.200 the gender of somebody in the photo. 00:24:07.920 --> 00:24:11.040 And also another kind of 00:24:13.200 --> 00:24:17.080 interactions that we have been kind of exploring 00:24:17.080 --> 00:24:20.120 is to just kind of describing 00:24:20.120 --> 00:24:23.120 the appearances 00:24:23.320 --> 00:24:25.440 instead of identities. 00:24:26.200 --> 00:24:28.480 For example, what kind of 00:24:28.480 --> 00:24:31.200 describe 00:24:31.200 --> 00:24:35.200 skin tones or hair style 00:24:35.640 --> 00:24:38.560 and outfit 00:24:39.600 --> 00:24:41.320 instead of 00:24:41.320 --> 00:24:45.480 assigning a kind of race or gender label of somebody. 00:24:46.320 --> 00:24:51.240 But I don't think any of those solutions can really address 00:24:51.240 --> 00:24:56.280 the kind of the real cause of the problem. 00:24:56.360 --> 00:25:01.200 So I don't really have a very good answer on this 00:25:01.800 --> 00:25:04.960 I think maybe we should, 00:25:04.960 --> 00:25:08.520 you know, like maybe the alternative is to kind of think of the way 00:25:08.840 --> 00:25:12.880 to come away and kind of share who we are. 00:25:13.480 --> 00:25:16.880 We saw so much relying on the kind of 00:25:21.640 --> 00:25:25.120 images like we are today. 00:25:25.240 --> 00:25:28.360 So, you know, like, how can we convey 00:25:28.880 --> 00:25:30.680 the kind of 00:25:31.800 --> 00:25:34.640 information that we want to share online, 00:25:35.080 --> 00:25:37.840 not so visual centric way. 00:25:38.400 --> 00:25:39.440 I think that's a kind of 00:25:40.520 --> 00:25:41.640 bigger 00:25:44.520 --> 00:25:47.320 question, the way I saw it, too. 00:25:47.320 --> 00:25:48.400 Thank you. 00:25:49.560 --> 00:25:50.640 Thank you, Shaomei. 00:25:50.640 --> 00:25:53.720 And Amy next to you. 00:25:54.640 --> 00:25:58.600 I think the prior the prior answer is mostly covered. 00:25:58.600 --> 00:26:00.480 The things I was going to mention I loved. 00:26:00.480 --> 00:26:03.040 Shaomei’s answer about, 00:26:03.040 --> 00:26:06.240 you know, describing ourselves in ways that are like figuring out ways 00:26:06.240 --> 00:26:10.320 that don't rely on the visual information and giving agency to people 00:26:10.320 --> 00:26:14.840 to just to add their own identities that they want to be shared. 00:26:15.000 --> 00:26:18.520 I will say that I think that that depends in different contexts. 00:26:18.520 --> 00:26:19.680 You might want to share 00:26:19.680 --> 00:26:23.800 different parts of your identity if it's important to you and you might. 00:26:24.240 --> 00:26:29.040 And so I think that even things that give like end users agency 00:26:29.040 --> 00:26:33.720 might have a lot of subtlety and how they would be applied in different cases. 00:26:34.640 --> 00:26:37.800 And I like the idea, you know, of describing, 00:26:37.800 --> 00:26:39.560 you know, aspects of appearance. 00:26:39.560 --> 00:26:42.520 I think you're missing one one challenge with that is 00:26:42.520 --> 00:26:43.920 you might be sort of trading off 00:26:43.920 --> 00:26:47.280 between these like aspects of appearance that you're describing and 00:26:48.080 --> 00:26:51.520 and the efficiency with which someone can like 00:26:51.720 --> 00:26:52.760 maybe they're not going to get 00:26:52.760 --> 00:26:56.280 the information as quickly as a sighted person would perceiving that person. 00:26:56.760 --> 00:27:00.040 And just because, you know, audio occurs over time. 00:27:00.040 --> 00:27:04.200 So so I think there's it's an extremely difficult, difficult challenge. 00:27:05.000 --> 00:27:07.280 And and in some cases it can matter. 00:27:07.280 --> 00:27:10.920 Like I can imagine, you know, seeing a photograph of the leadership 00:27:10.920 --> 00:27:14.080 of a company, you might want to know some some quick 00:27:15.360 --> 00:27:18.120 details about about the demographics of who's who's leading it. 00:27:18.120 --> 00:27:22.680 For instance, one one thing that I've noticed that 00:27:23.240 --> 00:27:26.160 is is sort of related to this is that, you know, 00:27:26.280 --> 00:27:29.640 when I'm when I am asking. 00:27:29.640 --> 00:27:31.200 So I sometimes, 00:27:31.200 --> 00:27:35.280 you know, have people describe videos and there can be a lot of differences 00:27:35.280 --> 00:27:38.360 in which aspects, even if they're going to describe the aspects of someone's appearance 00:27:39.080 --> 00:27:42.360 the way they describe those based on who is in front of them 00:27:42.720 --> 00:27:45.880 can also differ based on biases that people have. 00:27:45.880 --> 00:27:49.680 So if people see a woman, they might describe her differently 00:27:49.680 --> 00:27:50.760 than they would describe a man. 00:27:50.760 --> 00:27:54.360 They might focus on different aspects of of appearance. 00:27:54.360 --> 00:27:58.320 And so I think even things that go towards describing aspects of appearance 00:27:58.320 --> 00:28:02.280 will have to be, you know, very carefully, very carefully designed. 00:28:02.280 --> 00:28:05.680 And it really feels like a challenging a challenging problem. 00:28:05.720 --> 00:28:08.040 Yeah. 00:28:08.040 --> 00:28:11.080 Thank you so much, Amy. 00:28:11.080 --> 00:28:14.200 Michael, any thoughts on this? 00:28:14.200 --> 00:28:15.680 And I would 00:28:16.680 --> 00:28:18.200 add something 00:28:18.200 --> 00:28:21.920 here, especially for you, ... 00:28:22.160 --> 00:28:26.600 do you see any future role in accessibility guidelines 00:28:27.600 --> 00:28:29.760 in contributing to 00:28:31.560 --> 00:28:34.320 preventing bias in machine 00:28:34.320 --> 00:28:37.880 learning, generated descriptions or whatever 00:28:38.400 --> 00:28:43.840 that results from these models? 00:28:43.840 --> 00:28:46.040 I know my answer to that question. 00:28:46.040 --> 00:28:49.640 It could be longer than my prepared answers. 00:28:49.640 --> 00:28:53.040 So let's see where we go. 00:28:53.040 --> 00:28:56.760 I just want to add a couple of thoughts to what the others have been saying. 00:28:57.400 --> 00:28:59.880 I want to first to categorize bias 00:29:00.920 --> 00:29:04.240 as we're talking so far mainly about bias and recognition. 00:29:04.240 --> 00:29:08.360 You know, this... are there biases of how machine learning recognizes 00:29:08.360 --> 00:29:10.680 objects, people, etc., contexts 00:29:12.240 --> 00:29:17.160 in that, one thing that magnifies the challenge and accessibility context 00:29:17.160 --> 00:29:20.600 is that the sample size of people with disabilities 00:29:20.600 --> 00:29:23.160 can be smaller in various training sets. 00:29:24.120 --> 00:29:25.440 And so there is a risk 00:29:26.600 --> 00:29:29.040 that images of people with 00:29:29.040 --> 00:29:34.440 disabilities on a training set or contexts that are important for them, like wheelchair ramps 00:29:34.440 --> 00:29:38.400 or something will be excluded as outliers 00:29:38.400 --> 00:29:43.440 or will be less well recognizable by the AI 00:29:43.440 --> 00:29:46.040 than, you know, images of other people are. 00:29:46.760 --> 00:29:49.600 So, you know, that's just another 00:29:50.920 --> 00:29:55.360 another dimension to the aspects of bias that we need to look at. 00:29:55.720 --> 00:30:00.200 But then we also need to look at the own bias in the application of this. 00:30:00.920 --> 00:30:03.400 You know, we've talked a few times during the session 00:30:03.400 --> 00:30:07.520 about the risk of relying on machine generated 00:30:08.640 --> 00:30:11.320 descriptions, captions as being good enough, 00:30:12.520 --> 00:30:15.520 whereas content that has more of a mainstream 00:30:15.520 --> 00:30:18.440 audience might also have captions as descriptions. 00:30:18.440 --> 00:30:22.680 But get more curated, you know, quality assurance. 00:30:23.280 --> 00:30:27.760 So, you know, that kind of bias could creep in and 00:30:28.920 --> 00:30:29.440 that can 00:30:29.440 --> 00:30:32.400 magnify the impact on disability bias, 00:30:33.360 --> 00:30:37.640 you know, because it can cause people to be excluded from the fora 00:30:38.040 --> 00:30:42.400 that, often, which people are recruited to be part of training sets, etc.. 00:30:42.720 --> 00:30:45.960 So, you know, again, the ethical principles from where machine learning 00:30:46.240 --> 00:30:47.040 speaks to that. 00:30:47.040 --> 00:30:50.480 And I think that we may by identifying some content 00:30:50.480 --> 00:30:52.680 that we need to add to that. 00:30:53.960 --> 00:30:58.560 So moving on to what we can do about that, you know, 00:30:58.560 --> 00:31:02.280 I do believe that it's within the scope of the Web accessibility initiative 00:31:02.560 --> 00:31:05.640 or the W3C to provide guidance 00:31:05.640 --> 00:31:10.680 in some form about how AI and accessibility 00:31:10.680 --> 00:31:13.320 should work together, addressing many of these things. 00:31:14.640 --> 00:31:16.600 You know, typically this sort of thing 00:31:16.600 --> 00:31:19.800 would be a working group node, which means that it's a 00:31:21.000 --> 00:31:23.000 it is a a formal document 00:31:23.520 --> 00:31:27.040 published by the W3C that's had a certain level of review. 00:31:27.600 --> 00:31:30.120 There's even opportunities for versions 00:31:30.120 --> 00:31:32.880 that have had more review and signoff. 00:31:33.480 --> 00:31:35.680 So I think that's one thing we might like to do. 00:31:36.320 --> 00:31:39.000 I'll also talk briefly about the work that we're doing 00:31:39.000 --> 00:31:42.000 on the Web content accessibility guidelines 00:31:42.000 --> 00:31:45.200 3.0 sorry, the W3C accessibility 00:31:45.200 --> 00:31:48.520 guidelines 3 00:31:48.520 --> 00:31:50.600 or WCAG 3. 00:31:50.600 --> 00:31:54.360 We it's a it's a substantial re-envisioning 00:31:54.360 --> 00:31:59.000 and it's been a core requirement 00:31:59.000 --> 00:32:01.680 from the beginning that we wanted to address, you know, 00:32:03.000 --> 00:32:06.120 addressed equity in the approach, in the guidelines. 00:32:06.120 --> 00:32:09.280 How are we going to make sure that they're equitable to people with disabilities? 00:32:09.520 --> 00:32:13.120 We've been exploring that in specific ways and within the working group, 00:32:13.960 --> 00:32:16.480 really unpacking that to understand, 00:32:16.640 --> 00:32:19.560 you know, the relationship of equity and accessibility and bias. 00:32:19.560 --> 00:32:20.800 And in the other dimension. 00:32:20.800 --> 00:32:25.360 So that's turning, you know, we're connecting that with other work 00:32:25.360 --> 00:32:31.480 W3C has been doing to make itself a more equitable organization. 00:32:31.480 --> 00:32:36.120 And so, you know, this is to say that I believe WCAG 3, 00:32:36.560 --> 00:32:40.120 will also have some structure built in and support resources, 00:32:40.120 --> 00:32:43.920 addressing issues of bias specifically. 00:32:45.040 --> 00:32:47.320 Now, these are 00:32:47.480 --> 00:32:50.880 hopes, not promises, but you know, 00:32:51.480 --> 00:32:54.240 that's the direction from activities like this. 00:32:56.280 --> 00:32:57.120 Thank you so much. 00:32:57.120 --> 00:33:01.200 And yes, those are really some exciting avenues that we 00:33:01.440 --> 00:33:05.400 we hope that can come to fruition in the near future. 00:33:06.360 --> 00:33:08.640 So I guess final question 00:33:08.640 --> 00:33:12.800 for everyone, and it is 00:33:13.640 --> 00:33:18.080 I would like to know a bit about your future perspectives on the use 00:33:18.080 --> 00:33:23.800 of natural language processing for the field or in the field of accessibility. 00:33:24.360 --> 00:33:27.200 And I'll start with you this time, Amy. 00:33:30.920 --> 00:33:33.680 Yeah, So I think this is a really exciting area. 00:33:33.680 --> 00:33:39.200 And one thing, one one shift I've found recently among people in NLP 00:33:39.240 --> 00:33:42.960 who I talked to is that, you know, as the models are getting better 00:33:42.960 --> 00:33:47.880 at just creating like fluent, fluent text that looks reasonable, 00:33:48.240 --> 00:33:49.120 that lot of people 00:33:49.120 --> 00:33:52.920 are becoming more interested in what are the actual applications of this 00:33:52.920 --> 00:33:56.560 and how can we build tools that actually support those applications 00:33:56.720 --> 00:33:59.440 rather than relying on, you know, automated metrics for, 00:34:00.360 --> 00:34:03.520 but that might not, you know, capture people's experiences. 00:34:03.520 --> 00:34:08.520 So I wanted to to note that that's a direction I found exciting. 00:34:08.520 --> 00:34:12.800 So I guess a couple a couple of things I think could be promising are 00:34:13.200 --> 00:34:17.600 and I've kind of mentioned them before in my other in my other responses, But, 00:34:17.960 --> 00:34:23.040 you know, as gain the ability to describe more and more about the image, I, 00:34:23.240 --> 00:34:28.680 I think that NLP can provide a really good opportunity to personalize 00:34:29.040 --> 00:34:33.480 those descriptions based on the person and what they want as well as the context 00:34:33.680 --> 00:34:34.080 there is. 00:34:34.080 --> 00:34:35.760 You know, if you think about walking in a room, 00:34:35.760 --> 00:34:38.720 there's like so much you could possibly describe. 00:34:38.720 --> 00:34:41.720 If we can make it easier for people to get the information that they're 00:34:41.720 --> 00:34:46.360 looking for quickly from their media, that that would be a 00:34:47.560 --> 00:34:48.840 a great improvement. 00:34:48.840 --> 00:34:52.320 You know, combining computer vision to recognize things 00:34:52.320 --> 00:34:56.640 in the underlying image and using something like NLP to 00:34:57.520 --> 00:35:01.680 to summarize that description I think is is promising and exciting. 00:35:02.120 --> 00:35:04.120 And one other way I think I'm excited about 00:35:04.120 --> 00:35:09.080 it is in its opportunities to maybe help people with their own description tasks. 00:35:09.080 --> 00:35:12.600 So when we have humans working on descriptions, it's really hard. 00:35:13.400 --> 00:35:16.360 So, you know, novices sometimes have a hard time remembering 00:35:16.360 --> 00:35:18.560 and applying the guidelines that exist. 00:35:18.560 --> 00:35:23.280 You know, maybe we could rewrite people's descriptions of videos to be more in line 00:35:23.280 --> 00:35:24.760 with how an expert would write them 00:35:24.760 --> 00:35:28.560 by making them more concise or changing the grammar a bit 00:35:28.560 --> 00:35:32.640 so that it fits what people are expecting from their guidelines. 00:35:32.640 --> 00:35:36.480 Or we might alert people to aspects of their own descriptions 00:35:36.480 --> 00:35:39.920 that that might need to that could be changed a little bit 00:35:39.920 --> 00:35:44.480 to perhaps reduce something like bias that they have in the description. 00:35:44.480 --> 00:35:47.280 So I think there's there's really lots of exciting opportunities 00:35:47.280 --> 00:35:50.680 in terms of authoring descriptions as well as making those end descriptions 00:35:50.680 --> 00:35:51.640 a little bit better. Yeah. 00:35:53.280 --> 00:35:53.760 Great. 00:35:53.760 --> 00:35:54.120 Yeah. 00:35:54.120 --> 00:35:58.120 Thanks a lot. Shivam? 00:35:58.120 --> 00:35:59.800 Yeah, so 00:36:00.080 --> 00:36:04.200 I see it a bit of more it now rather than earlier, 00:36:04.600 --> 00:36:07.040 because now the models, the engines are quite advanced 00:36:08.040 --> 00:36:10.040 so I see a good context 00:36:10.040 --> 00:36:13.960 aware solution that gives you faster processing some efficient data. 00:36:13.960 --> 00:36:14.440 Right. 00:36:14.440 --> 00:36:17.840 And that works on text, video and as well as audio. 00:36:17.840 --> 00:36:20.400 So I see this happening to be a reality. 00:36:21.360 --> 00:36:23.400 A good use case I would have been following up 00:36:23.400 --> 00:36:27.280 also is how we can make the academic textbooks. 00:36:27.280 --> 00:36:29.920 And we have academic assignments, right? 00:36:29.920 --> 00:36:33.600 There are multiple charts, bar chart graphs, all associated data. 00:36:33.920 --> 00:36:38.360 If some of these models or technologies can create better 00:36:39.120 --> 00:36:42.120 understanding of those things, it would help a lot of 00:36:43.160 --> 00:36:46.960 people in understanding that we have difficulty just by reading it. 00:36:46.960 --> 00:36:50.400 Or maybe in absence of good quality 00:36:50.680 --> 00:36:53.320 descriptions of these charts, bars and all those things. 00:36:53.600 --> 00:36:55.840 I see this happening in the next few years. 00:36:56.320 --> 00:36:58.720 A better description of the generation. 00:36:59.080 --> 00:37:02.400 And as a closing comment, I would say there are different types of consumers 00:37:02.760 --> 00:37:05.880 of media that some can easily read but not comprehend. 00:37:05.880 --> 00:37:09.440 Some comprehend easily, but have difficulty consuming, 00:37:09.720 --> 00:37:11.200 consuming it visually. 00:37:11.200 --> 00:37:16.080 Now in that sense, the coming NLP technologies would help designers 00:37:16.080 --> 00:37:19.920 have contextual description of outputs and that I would see in simple terms. 00:37:20.280 --> 00:37:25.000 If you give me a simple, efficient faster output of a media and it's correct, 00:37:25.000 --> 00:37:29.080 then it will be the pinnacle of what I see as the NLP. 00:37:29.360 --> 00:37:32.440 And these are for natural language processing understanding 00:37:32.560 --> 00:37:37.200 as well as generation for all key technologies. 00:37:37.200 --> 00:37:38.200 Thank you so much. 00:37:38.200 --> 00:37:40.560 It's exciting times ahead. Definitely. 00:37:41.400 --> 00:37:43.680 Michael, you want to share your vision. 00:37:46.120 --> 00:37:49.920 So based on my knowledge of 00:37:50.560 --> 00:37:52.800 how machine learning in the present day works, 00:37:54.360 --> 00:37:56.720 you know, the tools tend to be more 00:37:56.720 --> 00:38:00.120 focused on specific abilities, 00:38:00.120 --> 00:38:02.760 which means that the context is a bit isolated. 00:38:03.680 --> 00:38:09.840 So I think I'm speaking as a as a person working the field, 00:38:09.840 --> 00:38:13.600 identifying a need rather than something that may necessarily be a technological 00:38:15.360 --> 00:38:16.640 potential. 00:38:16.640 --> 00:38:20.400 But the Internet of Things used as APIs 00:38:20.400 --> 00:38:23.520 to exchange data between different types of devices. 00:38:24.160 --> 00:38:27.640 And if we could model some structure like that so that these tools 00:38:27.640 --> 00:38:32.280 could share contexts with each other and negotiate a better group description, 00:38:32.520 --> 00:38:35.800 I think that that might be an opportunity for an early 00:38:37.160 --> 00:38:39.960 evolution of this field. 00:38:39.960 --> 00:38:41.560 You know, the long term, of course, 00:38:41.560 --> 00:38:45.920 I think tools will emerge with greater sense of context built in, 00:38:46.960 --> 00:38:49.400 but that'll probably be, you know, another 00:38:49.400 --> 00:38:52.400 tier slash singularity or whatever. 00:38:52.960 --> 00:38:56.520 So yeah, that's my view on that near term future based on my knowledge. 00:38:57.800 --> 00:38:58.520 Yeah. 00:38:58.920 --> 00:39:01.400 Good, good suggestions too to look at 00:39:01.400 --> 00:39:04.760 also. And Shaomei? 00:39:04.760 --> 00:39:05.600 Yeah. 00:39:05.600 --> 00:39:08.120 So yeah. So 00:39:08.720 --> 00:39:10.680 looking into the future 00:39:10.680 --> 00:39:13.760 I can see kind of two 00:39:16.680 --> 00:39:22.080 areas 00:39:22.080 --> 00:39:26.720 that I think will have a lot of potentials. 00:39:26.760 --> 00:39:28.960 And the first one it's from the 00:39:30.840 --> 00:39:32.480 technology 00:39:33.240 --> 00:39:39.400 perspective which I agree with Michael that I can see 00:39:39.760 --> 00:39:42.480 a lot of gain in kind of 00:39:44.520 --> 00:39:47.640 incorporating the context 00:39:47.640 --> 00:39:50.040 surrounding photos 00:39:50.040 --> 00:39:53.720 and by you know like taking advantage 00:39:53.720 --> 00:39:56.480 of the reason and, 00:39:57.600 --> 00:39:59.800 and progressing 00:40:00.800 --> 00:40:05.680 and deep learning models that kind of have 00:40:05.680 --> 00:40:08.160 what kind of math models are 00:40:10.080 --> 00:40:14.720 representations space So you know like we can embed both 00:40:14.800 --> 00:40:18.680 the kind of image as well as the kind of tags 00:40:19.680 --> 00:40:21.840 surrounding it and then 00:40:21.840 --> 00:40:24.040 and then add a kind of 00:40:24.320 --> 00:40:27.760 metadata such as the author or the time 00:40:27.760 --> 00:40:29.440 when, you know, 00:40:29.880 --> 00:40:35.520 when the photo was taken, all kind of posted. 00:40:35.520 --> 00:40:39.360 So, you know, a lot of those can be kind of drawing in a kind of big 00:40:40.040 --> 00:40:43.920 represent patient space that kind of that 00:40:44.960 --> 00:40:47.640 provides a lot more than just kind of 00:40:49.040 --> 00:40:50.400 visual 00:40:51.520 --> 00:40:53.800 information alone. 00:40:53.800 --> 00:40:57.760 So I think that's a kind of big 00:40:57.840 --> 00:40:59.840 technology break 00:40:59.840 --> 00:41:03.520 through that we can see in the kind of near-term future. 00:41:03.960 --> 00:41:08.960 But the kind of second thing I think and more important to me 00:41:08.960 --> 00:41:12.360 is the kind of use case 00:41:12.920 --> 00:41:15.000 perspectives. 00:41:15.000 --> 00:41:18.440 I think right now when we think about all kind of talk 00:41:18.440 --> 00:41:19.320 about 00:41:22.160 --> 00:41:24.360 the media 00:41:25.480 --> 00:41:27.520 accessibility 00:41:27.520 --> 00:41:30.920 we are mostly kind of think about our consumption case, 00:41:31.200 --> 00:41:34.720 like how do I help some people who can not see to kind of 00:41:35.640 --> 00:41:40.560 to kind of consume photos that posted by 00:41:42.000 --> 00:41:45.720 others and mostly by kind of sighted folks. 00:41:45.720 --> 00:41:49.920 But I think it's equally important but largely kind of 00:41:51.720 --> 00:41:54.000 overlook is, 00:41:54.000 --> 00:41:56.520 is these kind of media 00:41:57.600 --> 00:42:02.760 creation 00:42:03.680 --> 00:42:06.440 use cases, you know, like how can we support 00:42:07.680 --> 00:42:10.000 people with visual 00:42:10.320 --> 00:42:12.960 impairment to kind of 00:42:13.600 --> 00:42:19.080 create and kind of share photos and videos 00:42:20.440 --> 00:42:22.800 in my own work into, you know, 00:42:22.960 --> 00:42:26.000 these use cases, which is why you know 00:42:27.240 --> 00:42:30.000 there's like such a gap in what the kind of 00:42:31.800 --> 00:42:34.120 current technology can do. 00:42:34.120 --> 00:42:37.360 For example one like all a kind of modern 00:42:38.520 --> 00:42:41.720 AI models really failed 00:42:41.720 --> 00:42:45.120 when it came to processing photos 00:42:46.280 --> 00:42:52.200 taken by people with visual impairments 00:42:53.320 --> 00:42:56.840 because they are just not the same kind of photo that are used to 00:42:56.880 --> 00:43:00.080 train those 00:43:00.080 --> 00:43:01.360 models. 00:43:01.360 --> 00:43:06.000 So, you know, there's a huge gap in what kind of current like the kind of 00:43:07.240 --> 00:43:09.080 fundamentals of 00:43:09.080 --> 00:43:11.760 those models and then what they can do. 00:43:12.000 --> 00:43:15.840 And then second is there is a lot need for more 00:43:17.640 --> 00:43:20.560 personalized and 00:43:20.920 --> 00:43:24.600 and aesthetic needs. 00:43:24.920 --> 00:43:27.920 Right after I take ten 00:43:28.840 --> 00:43:31.960 selfies, I wanna find out why 00:43:32.200 --> 00:43:35.800 I wanna post that kind of share who I am 00:43:36.440 --> 00:43:40.320 and that it's or it's like we cannot do at all. 00:43:40.320 --> 00:43:46.080 We can, you know we can kind of tell you're like, okay, you'll have ten 00:43:47.120 --> 00:43:50.680 photos and are kind of containing your face, 00:43:51.600 --> 00:43:53.760 but you know but, but, 00:43:54.400 --> 00:43:59.880 but then how like, how can we change kind of the models 00:43:59.880 --> 00:44:04.400 that can really represent somebody's space and then, you know, 00:44:04.440 --> 00:44:09.240 somebody’s kind of aesthetics and I think that's another interesting 00:44:10.560 --> 00:44:11.840 future 00:44:12.280 --> 00:44:14.640 development that I want to see. 00:44:15.680 --> 00:44:17.840 So that's all. 00:44:17.840 --> 00:44:19.400 Thank you so much, Shaomei. 00:44:19.400 --> 00:44:23.960 And I think we we only have 4 minutes more. 00:44:23.960 --> 00:44:26.800 So I won’t risk another question 00:44:27.520 --> 00:44:31.160 because we need to to to end at the top of the hour. 00:44:31.560 --> 00:44:36.240 And so I will take the opportunity to once again, thanks 00:44:36.680 --> 00:44:42.360 thank our our panelists and I hope everyone enjoyed it as much as I did. 00:44:42.360 --> 00:44:43.080 And it was 00:44:44.280 --> 00:44:45.480 really interesting 00:44:45.480 --> 00:44:51.880 and very, very optimistic perspectives so that 00:44:53.240 --> 00:44:56.600 we can see that's not just the, uh, the, 00:44:56.960 --> 00:45:00.200 um, risky, uh, 00:45:00.360 --> 00:45:05.400 or risk enabling, uh, outputs that A.I. 00:45:05.400 --> 00:45:06.040 can have. 00:45:06.040 --> 00:45:09.160 So it's nice to, to have these perspectives. 00:45:09.160 --> 00:45:10.600 So thank you once again. 00:45:10.600 --> 00:45:15.720 So Shaomei, Shivam, Amy and Michael, it was brilliant to have you here, 00:45:17.160 --> 00:45:19.880 and thanks 00:45:19.880 --> 00:45:21.200 who attended.