WEBVTT 00:00:00.004 --> 00:00:01.924 If we want to design AI in a way 00:00:01.924 --> 00:00:03.804 that's positive for society, 00:00:03.804 --> 00:00:07.044 how do we think beyond just the user level of design, 00:00:07.044 --> 00:00:09.364 but to think of the broader community that it impacts 00:00:09.364 --> 00:00:10.844 and even society. 00:00:10.844 --> 00:00:12.524 James, thank you so much for being here. 00:00:12.524 --> 00:00:14.164 Thank you for having me. 00:00:14.164 --> 00:00:18.924 So the Stanford Institute for Human-Centered AI. 00:00:18.924 --> 00:00:21.124 Explain to me what the mandate is. 00:00:21.124 --> 00:00:23.324 What are you looking to do at SHAI? 00:00:23.324 --> 00:00:27.404 At HAI, what we're really interested in is making sure 00:00:27.404 --> 00:00:32.324 to guide and shape AI for the best outcomes for humanity. 00:00:32.324 --> 00:00:35.724 And so it really comes down to understanding 00:00:35.724 --> 00:00:37.684 how is this gonna impact us? 00:00:37.684 --> 00:00:40.004 So think about guiding it in terms of ethics 00:00:40.004 --> 00:00:42.684 and other social impact, how to design it, 00:00:42.684 --> 00:00:46.004 and how to actually do the basic underlying science 00:00:46.004 --> 00:00:48.764 of the new technology as it's developed. 00:00:48.764 --> 00:00:50.124 What are some of the most interesting things 00:00:50.124 --> 00:00:50.964 that you're seeing? 00:00:50.964 --> 00:00:54.524 Well, I mean, obviously the number of interesting 00:00:54.524 --> 00:00:58.124 research innovations in different labs are incredible. 00:00:58.124 --> 00:01:01.244 Whether it's thinking about how to do these large models 00:01:01.244 --> 00:01:03.804 at a smaller scale that could run, you know, on your phone 00:01:03.804 --> 00:01:06.884 and not leak all your private information 00:01:06.884 --> 00:01:09.684 into the infrastructure, you know, things like that, 00:01:09.684 --> 00:01:12.644 but all the way to applications of how we use AI 00:01:12.644 --> 00:01:13.764 to improve health care, 00:01:13.764 --> 00:01:16.484 improve mental health, everything happens. 00:01:16.484 --> 00:01:20.364 And so we founded the institute with the recognition- 00:01:20.364 --> 00:01:23.524 before this journey of AI even happened- 00:01:23.524 --> 00:01:25.524 that AI was now at the level 00:01:25.524 --> 00:01:27.524 that it was going to change all of society, 00:01:27.524 --> 00:01:29.804 and therefore we needed all of the fields 00:01:29.804 --> 00:01:30.964 that you could only find 00:01:30.964 --> 00:01:33.364 in a great research university like Stanford 00:01:33.364 --> 00:01:36.244 working on this together. So we're very interdisciplinary. 00:01:36.244 --> 00:01:39.044 We'll have artists working with computer scientists, 00:01:39.044 --> 00:01:41.885 health care folks working with lawyers 00:01:41.885 --> 00:01:43.725 and economists and technologists. 00:01:43.725 --> 00:01:45.565 Like every possible combination 00:01:45.565 --> 00:01:47.645 of these different fields working together 00:01:47.645 --> 00:01:50.245 to define how this future could be. 00:01:50.245 --> 00:01:52.685 I'm curious, what are some of the things, 00:01:52.685 --> 00:01:53.605 what are some of the topics, 00:01:53.605 --> 00:01:56.005 that you are personally grappling with right now? 00:01:56.005 --> 00:01:58.885 So, you know, I've had a background 00:01:58.885 --> 00:02:01.125 in human computer interaction and design. 00:02:01.125 --> 00:02:03.285 So how do we design the future technology 00:02:03.285 --> 00:02:04.165 that people would use? 00:02:04.165 --> 00:02:07.845 So I've actually been for 30 years working with AI 00:02:07.845 --> 00:02:10.805 and how it would happen in our future tools. 00:02:10.805 --> 00:02:12.925 Today, that really happens in two ways. 00:02:12.925 --> 00:02:15.285 One, I work on applications of the future. 00:02:15.285 --> 00:02:18.165 For example, can I create a great health coach, 00:02:18.165 --> 00:02:20.165 a fitness coach that's personalized to you 00:02:20.165 --> 00:02:22.525 and understands your goals, your barriers, 00:02:22.525 --> 00:02:24.565 your long-term desires 00:02:24.565 --> 00:02:26.205 and then helps you get to those goals. 00:02:26.205 --> 00:02:28.205 And we have a paper that's coming out 00:02:28.205 --> 00:02:30.245 that showed how well this can work. 00:02:30.245 --> 00:02:31.565 All the way to, 00:02:31.565 --> 00:02:35.405 if we want to design AI in a way that's positive for society, 00:02:35.405 --> 00:02:38.605 how do we think beyond just the user level of design, 00:02:38.605 --> 00:02:40.925 but to think of the broader community that it impacts 00:02:40.925 --> 00:02:44.165 and even society. And what's a new design process 00:02:44.165 --> 00:02:47.405 that we can develop so that our partners, companies, 00:02:47.405 --> 00:02:49.445 and employees could design in that way 00:02:49.445 --> 00:02:53.485 and have AI that has a more positive impact on societies. 00:02:53.485 --> 00:02:57.005 And you talked a little bit about companies and partners. 00:02:57.005 --> 00:03:00.725 SHAI has always had this real sort of ecosystem approach 00:03:00.725 --> 00:03:03.005 to how you do things. 00:03:03.005 --> 00:03:03.925 But talk a little bit more 00:03:03.925 --> 00:03:05.325 about how you work with companies. 00:03:05.325 --> 00:03:10.325 Yeah, so at HAI, if we wanna really change the world 00:03:10.485 --> 00:03:13.125 we can't just change it through writing research papers 00:03:13.125 --> 00:03:15.485 that only academics are gonna read. 00:03:15.485 --> 00:03:16.765 The real change happens 00:03:16.765 --> 00:03:19.445 through people producing real products. 00:03:19.445 --> 00:03:22.766 So we have a very important industrial affiliates program 00:03:22.766 --> 00:03:27.286 where companies join with us and they help sponsor research, 00:03:27.286 --> 00:03:30.086 so it helps us in terms of funding of the research, 00:03:30.086 --> 00:03:33.286 but it also helps us be grounded in what the real problems 00:03:33.286 --> 00:03:34.966 that companies are facing 00:03:34.966 --> 00:03:37.006 as they try to deploy AI in the real world. 00:03:37.006 --> 00:03:38.526 So it's really a two-way, 00:03:38.526 --> 00:03:41.126 we do our research that helps inform them 00:03:41.126 --> 00:03:43.726 and helps them learn how to do responsible 00:03:43.726 --> 00:03:45.486 and human-centered AI, but at the same time, 00:03:45.486 --> 00:03:48.086 it's helping our research be better by being grounded 00:03:48.086 --> 00:03:50.606 in real problems that companies are facing. 00:03:50.606 --> 00:03:53.326 James, Dr. Landay, thank you so much for your time here. 00:03:53.326 --> 00:03:55.366 Wonderful to learn about HAI 00:03:55.366 --> 00:03:56.726 and all of the things that you're doing. 00:03:56.726 --> 00:03:58.006 Thank you for having me.