WEBVTT 00:00:00.308 --> 00:00:01.664 [TEASER]  00:00:01.664 --> 00:00:08.550 RAFAH HOSN: What has changed is that in the old days, we had the luxury of creating something, 00:00:08.550 --> 00:00:14.179 going and piloting for three months until we know whether it works or not, and then 00:00:14.179 --> 00:00:16.219 taking one year to productize! 00:00:16.219 --> 00:00:18.930 That … that, that doesn’t work anymore! 00:00:18.930 --> 00:00:19.960 Because guess what? 00:00:19.960 --> 00:00:25.890 In three months, this innovation is, like, topped by four other innovations, be it at 00:00:25.890 --> 00:00:28.300 Microsoft or elsewhere. 00:00:28.300 --> 00:00:33.740 So that speed is really shifting the mindset and the spirit of people. 00:00:33.740 --> 00:00:34.740 [TEASER ENDS] 00:00:34.740 --> 00:00:39.460 GRETCHEN HUIZINGA: You’re listening to Ideas, a Microsoft Research Podcast that dives deep 00:00:39.460 --> 00:00:45.540 into the world of technology research and the profound questions behind the code. 00:00:45.540 --> 00:00:53.180 I’m Dr. Gretchen Huizinga. 00:00:53.180 --> 00:00:58.859 In this series, we’ll explore the technologies that are shaping our future and the big ideas 00:00:58.859 --> 00:01:00.190 that propel them forward. 00:01:00.190 --> 00:01:01.390 [MUSIC FADES] 00:01:01.390 --> 00:01:03.260 My guest today is Rafah Hosn. 00:01:03.260 --> 00:01:08.409 She’s a partner, group product manager for AI Frontiers at Microsoft Research. 00:01:08.409 --> 00:01:13.990 I’d call Rafah a sort of organizational conductor, working both with leaders to drive 00:01:13.990 --> 00:01:19.100 clarity around the mission as well as program managers to make sure they have solid operational 00:01:19.100 --> 00:01:21.400 strategies to execute on it. 00:01:21.400 --> 00:01:26.229 Rafah has mad skills in bringing research ideas from lab to life, and I’m thrilled 00:01:26.229 --> 00:01:27.909 to talk to her today. 00:01:27.909 --> 00:01:29.470 Rafah Hosn, welcome to Ideas! 00:01:29.470 --> 00:01:30.530 RAFAH HOSN: Thank you, Gretchen. 00:01:30.530 --> 00:01:33.810 Oh, my goodness, I have to live up to this introduction now! 00:01:33.810 --> 00:01:34.810 [LAUGHTER] 00:01:34.810 --> 00:01:40.479 HUIZINGA: Well, before we talk about research ideas, let’s talk about you and your own 00:01:40.479 --> 00:01:44.020 sort of “reason for being” in the research world. 00:01:44.020 --> 00:01:47.829 How would you describe your motivation for working in research and—assuming there was 00:01:47.829 --> 00:01:51.290 one—what was the “big idea” or animating “what if?” 00:01:51.290 --> 00:01:52.730 behind what you’re doing today? 00:01:52.730 --> 00:01:55.710 HOSN: Yeah, you know, I don’t know. 00:01:55.710 --> 00:01:59.450 There are so many big ideas, to be honest! 00:01:59.450 --> 00:02:05.560 Every day, I wake up and I often tell my husband how lucky, like so totally lucky and humbled, 00:02:05.560 --> 00:02:11.530 I am to be where I am right now in this moment, like right now when society as we know it 00:02:11.530 --> 00:02:15.560 is being totally disrupted by this huge leap in AI. 00:02:15.560 --> 00:02:16.560 And why research? 00:02:16.560 --> 00:02:18.540 Well, I’ve tried it all, Gretchen! 00:02:18.540 --> 00:02:24.519 I’ve been in research, I went to product, I did engineering, and I did full circle and 00:02:24.519 --> 00:02:25.680 came back to research. 00:02:25.680 --> 00:02:32.770 Because, you know, for me personally, there’s no other environment that I know of, for me, 00:02:32.770 --> 00:02:37.890 that has this amount of creativity and just infinite curiosity and intellect. 00:02:37.890 --> 00:02:44.930 So working with people that are asking “what next?” and trying to imagine the next world 00:02:44.930 --> 00:02:49.670 beyond where AI is today is just … this is the big idea. 00:02:49.670 --> 00:02:51.060 This is why I’m here. 00:02:51.060 --> 00:02:54.019 This is why I’m excited to come to work every day. 00:02:54.019 --> 00:02:55.019 HUIZINGA: Yeah. 00:02:55.019 --> 00:02:58.840 Well … and I want to drill in a little bit just, sort of, personally because sometimes 00:02:58.840 --> 00:03:07.470 there’s a story, an origin story, if you will, of some pivotal aha moment that you 00:03:07.470 --> 00:03:12.480 say, oh, that’s fascinating, that’s cool, that’s what I want to do. 00:03:12.480 --> 00:03:16.830 Anything that piqued your interest way back when you were a kid or, sort of, a pivotal 00:03:16.830 --> 00:03:19.290 moment in your educational years? 00:03:19.290 --> 00:03:27.299 HOSN: Yeah, you know, so many different things that inspire you along the journey, right. 00:03:27.299 --> 00:03:30.549 It’s not just one thing, Gretchen. 00:03:30.549 --> 00:03:31.630 My dad was a doctor. 00:03:31.630 --> 00:03:35.500 He was my biggest inspiration growing up. 00:03:35.500 --> 00:03:40.629 And the reason is because he had a lot of depth of knowledge in his domain. 00:03:40.629 --> 00:03:41.629 And I wanted that. 00:03:41.629 --> 00:03:45.549 I wanted to have depth of knowledge in a domain. 00:03:45.549 --> 00:03:48.159 So I went engineering against his advice. 00:03:48.159 --> 00:03:50.620 He really wanted me to be a doctor. 00:03:50.620 --> 00:03:53.450 [LAUGHTER] So he was not too happy. 00:03:53.450 --> 00:04:00.409 But, you know, throughout my education, you know, I was there when smartphones came about, 00:04:00.409 --> 00:04:04.479 when the internet was a thing. 00:04:04.479 --> 00:04:10.209 And now, like with generative AI, I feel like I’ve lived through so many disruptions, 00:04:10.209 --> 00:04:13.890 and every one of those was, “Oh my gosh! 00:04:13.890 --> 00:04:17.180 Like, I am exactly where I want to be!” 00:04:17.180 --> 00:04:23.710 So multiple inspirations, and every day, I wake up and there’s new news and I’m saying 00:04:23.710 --> 00:04:25.820 to myself, “OK, that’s great.” 00:04:25.820 --> 00:04:26.889 I love it! 00:04:26.889 --> 00:04:28.510 HUIZINGA: What a time to be alive! 00:04:28.510 --> 00:04:30.240 HOSN: It is amazing! 00:04:30.240 --> 00:04:31.960 HUIZINGA: Yeah. 00:04:31.960 --> 00:04:38.020 Well, you recently took on this new role in AI Frontiers at Microsoft Research. 00:04:38.020 --> 00:04:42.699 And that very word “frontiers” evokes images of unexplored, uncharted territories 00:04:42.699 --> 00:04:47.259 like the Wild West or for Trekkies, maybe “space: the final frontier.” 00:04:47.259 --> 00:04:52.410 So what does it mean to you to be working at the frontier of artificial intelligence, 00:04:52.410 --> 00:04:56.020 and what’s the big idea behind AI Frontiers? 00:04:56.020 --> 00:05:01.780 HOSN: You know, it’s my biggest and most exciting adventure so far! 00:05:01.780 --> 00:05:10.720 Working under Ece Kamar’s leadership in this AI Frontiers is really trying to push 00:05:10.720 --> 00:05:16.840 ourselves to think, what’s beyond what there is right now in artificial intelligence? 00:05:16.840 --> 00:05:21.090 Where can we push more, from a scientific perspective? 00:05:21.090 --> 00:05:26.990 How do we translate these scientific discoveries into capabilities that people can actually 00:05:26.990 --> 00:05:29.220 use and derive value from? 00:05:29.220 --> 00:05:34.770 It’s a big responsibility, as well, because we just don’t want to push the boundaries 00:05:34.770 --> 00:05:37.100 of AI for the sake of pushing. 00:05:37.100 --> 00:05:40.450 We want to push it in a safe and responsible way. 00:05:40.450 --> 00:05:42.780 So it is a big responsibility. 00:05:42.780 --> 00:05:43.800 HUIZINGA: Yeah … 00:05:43.800 --> 00:05:51.790 HOSN: And fundamentally, you know, the unifying big idea in this team is to explore, you know, 00:05:51.790 --> 00:05:58.870 how far can we push intelligence further into models and encapsulations of those models 00:05:58.870 --> 00:06:07.340 so that we can, you know, have not just sort of an assistant but really a personal assistant, 00:06:07.340 --> 00:06:15.600 an agent that can, kind of, do tasks for us, with us, seamlessly across multiple domains? 00:06:15.600 --> 00:06:18.040 So this is what we’re trying to push for. 00:06:18.040 --> 00:06:19.040 HUIZINGA: Mmm. 00:06:19.040 --> 00:06:22.740 Rafah, do you feel like you’re at the frontier of artificial intelligence? 00:06:22.740 --> 00:06:29.380 I mean, what are the emotions that crop up when you are dealing with these things—that 00:06:29.380 --> 00:06:34.580 you and your teams basically know about but the rest of us don’t? 00:06:34.580 --> 00:06:38.210 HOSN: For most days, it’s excitement. 00:06:38.210 --> 00:06:41.810 Sometimes it’s [LAUGHTER] … it ranges, to be honest. 00:06:41.810 --> 00:06:46.180 I would say there’s a spectrum of emotions. 00:06:46.180 --> 00:06:51.400 The dominating one is really just excitement. 00:06:51.400 --> 00:06:58.699 There’s so much that has happened with GenAI, but I feel like it has opened up so many different 00:06:58.699 --> 00:07:04.569 paths, as well, for us to explore, and that’s the excitement. 00:07:04.569 --> 00:07:09.800 And then every time the world accomplishes something, you’re like in astonishment. 00:07:09.800 --> 00:07:11.569 You’re like, wow, wow. 00:07:11.569 --> 00:07:12.569 HUIZINGA: Yeah … 00:07:12.569 --> 00:07:16.020 HOSN: And then, and then, oh my gosh, what’s next? 00:07:16.020 --> 00:07:17.680 And so, it’s a range of emotions … 00:07:17.680 --> 00:07:18.680 HUIZINGA: Right … 00:07:18.680 --> 00:07:20.240 HOSN: … but I would say the dominating one is enthusiasm. 00:07:20.240 --> 00:07:21.240 HUIZINGA: Yeah. 00:07:21.240 --> 00:07:27.990 Well, I’ve heard other people on your teams use words like surprise, sometimes even shock 00:07:27.990 --> 00:07:28.990 … 00:07:28.990 --> 00:07:31.599 HOSN: Yeah, yeah, there are a lot of “wow” factors. 00:07:31.599 --> 00:07:39.560 Every day, every day, I wake up, I read like my three favorite AI tweets or things like 00:07:39.560 --> 00:07:40.560 that, and I’m like, “Oh my gosh. 00:07:40.560 --> 00:07:46.000 I wouldn’t have imagined that this model could do this thing,” so [LAUGHS] … um, 00:07:46.000 --> 00:07:47.580 but it’s exciting. 00:07:47.580 --> 00:07:53.069 HUIZINGA: We may have to get those accounts in the show notes so that we can follow along 00:07:53.069 --> 00:07:55.550 with your surprise and amazement in the mornings! 00:07:55.550 --> 00:07:57.050 HOSN: [LAUGHS] Yes! 00:07:57.050 --> 00:08:01.950 HUIZINGA: Well, listen, when we talk about measuring the success of an AI system, we 00:08:01.950 --> 00:08:05.750 often use the common convention of what we call benchmarks. 00:08:05.750 --> 00:08:10.940 But I want to zoom out from AI systems for a minute and ask how you might measure the 00:08:10.940 --> 00:08:14.509 success of an AI lab, which is what you’re working in. 00:08:14.509 --> 00:08:19.729 What are your benchmarks or key performance indicators—we call them KPIs—for the work 00:08:19.729 --> 00:08:21.370 going on at AI Frontiers? 00:08:21.370 --> 00:08:27.289 HOSN: Yeah, so I’m going to start by something that may sound surprising maybe to some, but 00:08:27.289 --> 00:08:28.919 I think it’s the culture first. 00:08:28.919 --> 00:08:40.570 It’s the culture of endless curiosity, of enthusiasm coupled with a bit of skepticism, 00:08:40.570 --> 00:08:46.760 to be honest, to ask the questions, the right questions, and this drive to push further. 00:08:46.760 --> 00:08:54.110 So I would say one KPI of success for me, personally, is, you know, can we maintain 00:08:54.110 --> 00:09:03.380 this culture of enthusiasm coupled with skepticism so we can ask hard questions and an envelope 00:09:03.380 --> 00:09:06.870 of enthusiasm and drive for everyone? 00:09:06.870 --> 00:09:08.420 So that’s one. 00:09:08.420 --> 00:09:15.580 I would say the other three are … one is around how much can we push scientifically 00:09:15.580 --> 00:09:18.560 as a community, right? 00:09:18.560 --> 00:09:24.540 This is a team of people that are getting together with a mission to push the boundaries 00:09:24.540 --> 00:09:27.990 of our understanding of artificial intelligence. 00:09:27.990 --> 00:09:30.670 So are we pushing that scientific boundaries? 00:09:30.670 --> 00:09:36.230 Are we creating insights, not just for the scientific community, but also for Microsoft 00:09:36.230 --> 00:09:40.970 and the world, so that we know how to derive value from these discoveries, right? 00:09:40.970 --> 00:09:44.890 At the end of the day, it is awesome to push scientifically. 00:09:44.890 --> 00:09:51.720 It’s even more awesome if you take this and translate it into something a human being 00:09:51.720 --> 00:09:52.720 can use … 00:09:52.720 --> 00:09:53.720 HUIZINGA: Yeah … 00:09:53.720 --> 00:09:56.570 HOSN: … or an enterprise can use. 00:09:56.570 --> 00:10:02.000 And I think … that’s kind of my KPIs of success. 00:10:02.000 --> 00:10:08.890 Culture first, pushing on the scientific boundaries, creating insights for the scientific community 00:10:08.890 --> 00:10:13.950 as well as for Microsoft so we can derive value for us as a society, right. 00:10:13.950 --> 00:10:15.300 HUIZINGA: Yeah. 00:10:15.300 --> 00:10:19.940 Well, continuing on this idea of success, and you’ve alluded to this already in terms 00:10:19.940 --> 00:10:27.410 of characteristics of curiosity and so on, part of your job, as you put it, was “enabling 00:10:27.410 --> 00:10:28.890 brilliant minds to find success.” 00:10:28.890 --> 00:10:35.310 So talk a little bit about the personal qualities of these brilliant minds and how you help 00:10:35.310 --> 00:10:36.990 them find success. 00:10:36.990 --> 00:10:45.070 HOSN: Yeah, you know, everybody I work with brings different aspects of brilliance to 00:10:45.070 --> 00:10:47.080 the table—every day. 00:10:47.080 --> 00:10:57.070 So in our community of engineers, PMs, researchers, everybody is present with their ideas and 00:10:57.070 --> 00:10:58.250 their strengths. 00:10:58.250 --> 00:11:03.779 And they’re pulling together to push harder and faster on our key priorities. 00:11:03.779 --> 00:11:09.940 And I find folks working in AI these days, you know, to have a renewed fire. 00:11:09.940 --> 00:11:13.630 It’s really amazing to see. 00:11:13.630 --> 00:11:20.350 And I talk a lot about curiosity, but, you know, I cannot emphasize how much this is 00:11:20.350 --> 00:11:28.010 driving a lot of our community to explore new paths that they hadn’t thought about 00:11:28.010 --> 00:11:31.400 prior to this GenAI coming along. 00:11:31.400 --> 00:11:41.560 And so everybody is showing up, present, asking these questions and trying to solve new scenarios, 00:11:41.560 --> 00:11:44.139 new problems that are emerging. 00:11:44.139 --> 00:11:51.160 And from my perspective, you know, as you mentioned, I just try to unblock, basically. 00:11:51.160 --> 00:11:56.300 My team and I are here to [LAUGHTER] … well, two things I would say. 00:11:56.300 --> 00:11:58.410 First is bring the outside-in perspective. 00:11:58.410 --> 00:12:04.590 That’s so important because science is amazing, but unless you can derive value from it, it 00:12:04.590 --> 00:12:08.450 remains an awesome paper and an awesome equation, right. 00:12:08.450 --> 00:12:11.740 So asking, who can use this? 00:12:11.740 --> 00:12:17.250 What are the scenarios it could, you know, light up? 00:12:17.250 --> 00:12:18.660 How can we derive value? 00:12:18.660 --> 00:12:24.930 So those are the questions that my team and I can contribute to, and we are trying to 00:12:24.930 --> 00:12:30.889 participate from ideation all the way to basically delivering on key milestones. 00:12:30.889 --> 00:12:32.949 And that last mile is so important. 00:12:32.949 --> 00:12:36.620 Like, once you know what you want to do, how do you structure? 00:12:36.620 --> 00:12:43.730 How do you have an operational strategy that is amenable to these times, which is fast, 00:12:43.730 --> 00:12:45.310 fast, fast, and faster? 00:12:45.310 --> 00:12:48.250 So that’s, kind of, what we’re trying to do here. 00:12:48.250 --> 00:12:50.170 HUIZINGA: Yeah, yeah. 00:12:50.170 --> 00:12:55.620 Well, two things came to my mind in terms of what kinds of people would end up working 00:12:55.620 --> 00:12:56.620 in this area. 00:12:56.620 --> 00:13:00.220 And one would be agility, or agile. 00:13:00.220 --> 00:13:06.949 And that would, to me, represent in a researcher that the person would be able to spin or pivot 00:13:06.949 --> 00:13:09.140 if something didn’t work out. 00:13:09.140 --> 00:13:12.060 And the other one is sort of a risk-reward mentality. 00:13:12.060 --> 00:13:18.060 It’s like, where are you willing to push to get that reward versus what might keep 00:13:18.060 --> 00:13:19.490 you from even trying? 00:13:19.490 --> 00:13:29.010 HOSN: Yeah, so definitely in this AI Frontiers community, I’m finding a lot of adaptability. 00:13:29.010 --> 00:13:35.750 So people willing to try, failing fast when they fail, and pivoting. 00:13:35.750 --> 00:13:43.699 And you have to, nowadays, in this atmosphere that we are living in. 00:13:43.699 --> 00:13:49.089 And because we have the privilege of working in research—and it’s really an honor and 00:13:49.089 --> 00:13:53.870 a privilege, and I’m not saying it just lightly—but it is the place where you can 00:13:53.870 --> 00:13:55.490 take risks, Gretchen. 00:13:55.490 --> 00:14:01.430 It is the place where failing is totally fine because you’re learning and you’re pivoting 00:14:01.430 --> 00:14:06.769 in a way that allows you to progress on the next thing you tackle. 00:14:06.769 --> 00:14:17.019 So I feel like most of the people I work with in this community, AI Frontiers, we are risk 00:14:17.019 --> 00:14:18.259 takers. 00:14:18.259 --> 00:14:22.690 We want to push, and it’s OK to fail, and it’s OK to adapt. 00:14:22.690 --> 00:14:28.350 So, I think, as an aggregate, that’s kind of the spirit I’m seeing. 00:14:28.350 --> 00:14:32.990 HUIZINGA: In the past, Rafah, you’ve stressed the importance of both teams and timing. 00:14:32.990 --> 00:14:36.700 And so we’ve been talking about the teams and the minds and the kinds of qualities in 00:14:36.700 --> 00:14:38.160 those people. 00:14:38.160 --> 00:14:41.269 But what about the “when” of research? 00:14:41.269 --> 00:14:45.020 How does timing impact what gets done in your world? 00:14:45.020 --> 00:14:50.350 HOSN: Well, in this new era, Gretchen, everything is yesterday! 00:14:50.350 --> 00:14:54.210 [LAUGHS] I mean, it is true. 00:14:54.210 --> 00:15:00.370 AI research is moving at such speeds that I feel like we need to get accustomed to a 00:15:00.370 --> 00:15:02.900 timing of now. 00:15:02.900 --> 00:15:05.259 And if it’s not now, it’s yesterday. 00:15:05.259 --> 00:15:16.310 So the timing is important, but the leeway has shrunk so significantly that I feel like 00:15:16.310 --> 00:15:24.300 we have to really just be present in the moment and just move as fast as we can because everybody 00:15:24.300 --> 00:15:27.850 else is moving at the highest speed. 00:15:27.850 --> 00:15:30.260 So timing is “now,” is what I would say. 00:15:30.260 --> 00:15:36.570 HUIZINGA: On that note, with so many innovations in AI coming out every day, every minute, 00:15:36.570 --> 00:15:40.649 what you’ve just expressed is that research horizons are shorter than ever. 00:15:40.649 --> 00:15:46.839 But as one of your team members noted in a recent panel, it still takes a lot of time 00:15:46.839 --> 00:15:51.740 to translate a research artifact, maybe a noteworthy finding or a published paper or 00:15:51.740 --> 00:15:56.680 an equation, an algorithm, into a useful product for humans. 00:15:56.680 --> 00:16:01.060 So how are you then dealing with these newly compressed timelines of “it needs to be 00:16:01.060 --> 00:16:05.950 done yesterday to keep up,” and how has the traditional research-to-product pipeline 00:16:05.950 --> 00:16:06.950 changed? 00:16:06.950 --> 00:16:09.640 HOSN: Yeah, it’s an awesome question. 00:16:09.640 --> 00:16:18.870 It is so true that going from research to a production-quality algorithm or capability 00:16:18.870 --> 00:16:19.870 takes time. 00:16:19.870 --> 00:16:27.990 But what I’m seeing is that the research-to-capabilities is accelerating, meaning if you look at the 00:16:27.990 --> 00:16:37.160 world today in generative AI and its surrounding, folks even in research are creating assets 00:16:37.160 --> 00:16:40.520 as they are creating their research. 00:16:40.520 --> 00:16:45.220 And so they are thinking as well, how do I showcase this? 00:16:45.220 --> 00:16:48.160 And of course, these assets are not production ready. 00:16:48.160 --> 00:16:49.690 But here’s the kicker. 00:16:49.690 --> 00:16:59.639 I think that the product teams are also adapting to this generative AI era, and they are changing 00:16:59.639 --> 00:17:02.020 to meet this disruptive moment. 00:17:02.020 --> 00:17:09.890 They are changing the way they think, and they are accelerating the way they productize 00:17:09.890 --> 00:17:16.829 and look at hardening and securing the assets so that they can put them in the hands of 00:17:16.829 --> 00:17:22.260 even a limited set of users just to get a feel of what it means to have them in the 00:17:22.260 --> 00:17:29.820 hands of end users and quickly iterating so that they can further harden and further improve 00:17:29.820 --> 00:17:33.179 the design until it’s production ready. 00:17:33.179 --> 00:17:41.220 And I feel like our product partners are meeting the moments, meaning they also are really 00:17:41.220 --> 00:17:46.200 adapting their processes such that they can get these assets and put them in the hands 00:17:46.200 --> 00:17:49.679 of users and test them out before they actually release them. 00:17:49.679 --> 00:17:50.679 HUIZINGA: Right. 00:17:50.679 --> 00:17:56.200 Let’s drill in a little bit more on that and talk about the traditional research-to-product 00:17:56.200 --> 00:18:01.130 pipeline, where you would have a researcher working on something and then an RSDE. 00:18:01.130 --> 00:18:02.970 What does RSDE stand for? 00:18:02.970 --> 00:18:05.710 HOSN: A research software development engineer. 00:18:05.710 --> 00:18:07.000 It’s a mouthful. 00:18:07.000 --> 00:18:08.000 HUIZINGA: Right. 00:18:08.000 --> 00:18:11.270 And then to the PM, or program manager, and then to the engineer. 00:18:11.270 --> 00:18:15.750 And you’ve said this provocative statement: now everyone is a PM! 00:18:15.750 --> 00:18:17.210 HOSN: Everyone is a PM! 00:18:17.210 --> 00:18:18.210 [LAUGHTER] 00:18:18.210 --> 00:18:19.540 HUIZINGA: What do you mean by that? 00:18:19.540 --> 00:18:27.669 HOSN: I just, I just feel like if we are to meet the moment, we need to be thinking outside-in, 00:18:27.669 --> 00:18:31.740 inside-out simultaneously. 00:18:31.740 --> 00:18:39.850 And I believe that the spirit of program management, which is looking at the design from a user-centric 00:18:39.850 --> 00:18:48.940 perspective, is embedded as we are ideating, as we are trying to explore new methodologies, 00:18:48.940 --> 00:18:50.950 new algorithms, new assets. 00:18:50.950 --> 00:18:59.850 And so what has changed is that in the old days, we had the luxury of creating something, 00:18:59.850 --> 00:19:05.490 going and piloting for three months until we know whether it works or not, and then 00:19:05.490 --> 00:19:07.210 taking one year to productize! 00:19:07.210 --> 00:19:09.039 That … that, that doesn’t work anymore. 00:19:09.039 --> 00:19:10.039 [LAUGHTER] 00:19:10.039 --> 00:19:11.039 HUIZINGA: Right. 00:19:11.039 --> 00:19:12.039 HOSN: Because guess what? 00:19:12.039 --> 00:19:17.610 In three months, this innovation is, like, topped by four other innovations, be it at 00:19:17.610 --> 00:19:20.010 Microsoft or elsewhere. 00:19:20.010 --> 00:19:27.210 So that speed is really shifting the mindset and the, and the spirit of people. 00:19:27.210 --> 00:19:34.169 I have colleagues and friends, researchers, that are asking me, oh, scenarios, users … I 00:19:34.169 --> 00:19:36.320 mean it’s amazing to see. 00:19:36.320 --> 00:19:41.350 So, yes, everybody has gotten a little PM in them now. 00:19:41.350 --> 00:19:42.350 [LAUGHTER] 00:19:42.350 --> 00:19:47.980 HUIZINGA: Yeah, I did a podcast with Shamsi Iqbal and Jina Suh. 00:19:47.980 --> 00:19:52.940 And Shamsi was talking about this concept, this old concept, of the researcher being 00:19:52.940 --> 00:19:55.900 in their lab and saying, well, I’ve done this work; now go see what you want to do 00:19:55.900 --> 00:19:56.900 with it. 00:19:56.900 --> 00:20:00.480 I don’t think you have that affordance anymore as a researcher. 00:20:00.480 --> 00:20:01.480 HOSN: No … 00:20:01.480 --> 00:20:04.980 HUIZINGA: You’ve got to work much more tightly with other team members and think like a PM. 00:20:04.980 --> 00:20:05.980 HOSN: Totally. 00:20:05.980 --> 00:20:11.020 HUIZINGA: So let’s talk about how the big general idea behind AI Frontiers is giving 00:20:11.020 --> 00:20:14.250 birth to smaller, more specific ideas. 00:20:14.250 --> 00:20:18.799 What are some of the research directions and projects that you could tell us about that 00:20:18.799 --> 00:20:20.670 illustrate this vision here? 00:20:20.670 --> 00:20:27.320 HOSN: Yeah, and I’m sure you’ve heard some of it come from Ece Kamar as she spoke 00:20:27.320 --> 00:20:30.900 on this community that we have. 00:20:30.900 --> 00:20:38.000 In AI Frontiers, we’re exploring, I would say, three major areas of research. 00:20:38.000 --> 00:20:40.530 And I want you to imagine a stack. 00:20:40.530 --> 00:20:46.429 At the bottom of the stack, we’re asking ourselves questions around, what are some 00:20:46.429 --> 00:20:51.790 new architectures we can be thinking about for these foundational models? 00:20:51.790 --> 00:20:53.169 How do we create them? 00:20:53.169 --> 00:20:57.940 What kind of data we need to train them, to pre-train them. 00:20:57.940 --> 00:21:03.150 And then on top of that stack, which starts with a foundation model, we’re asking ourselves, 00:21:03.150 --> 00:21:06.390 OK great, you have a pretrained model. 00:21:06.390 --> 00:21:11.660 In a lot of cases, when you’re creating especially small models, you need to fine-tune 00:21:11.660 --> 00:21:12.660 them. 00:21:12.660 --> 00:21:19.960 So what is this methodology and data generation pipeline that we’re going to use to fine-tune 00:21:19.960 --> 00:21:28.170 these models and specialize them for both across domains and across skill set? 00:21:28.170 --> 00:21:34.900 And on top of that—so now we’re on the third layer—we have a final layer that encapsulates 00:21:34.900 --> 00:21:43.130 these models and orchestrates among them to allow them the ability to do, you know, complex 00:21:43.130 --> 00:21:44.130 tasks. 00:21:44.130 --> 00:21:49.470 And we don’t want to stop there because for us it’s … we don’t want to have 00:21:49.470 --> 00:21:53.690 an agent that just does things and doesn’t learn. 00:21:53.690 --> 00:21:59.410 So that learnability, that learning on the job, like we do as humans, is something we’re 00:21:59.410 --> 00:22:00.559 asking ourselves, as well. 00:22:00.559 --> 00:22:03.380 Like, how do we encapsulate these models? 00:22:03.380 --> 00:22:05.380 We orchestrate among them. 00:22:05.380 --> 00:22:12.210 And we allow these encapsulated things, we call them agents, to learn on the job so that 00:22:12.210 --> 00:22:14.350 they can accomplish more complex tasks. 00:22:14.350 --> 00:22:16.080 So those are the three things. 00:22:16.080 --> 00:22:21.000 And then cutting across these three layers, imagine there’s a thing that cuts across 00:22:21.000 --> 00:22:28.929 them, is doing everything in a way that allows us to rigorously evaluate and to ensure that 00:22:28.929 --> 00:22:32.690 we’re doing things in a safe and responsible way. 00:22:32.690 --> 00:22:35.370 So those are the main things that we’re working on. 00:22:35.370 --> 00:22:36.370 Does that make sense? 00:22:36.370 --> 00:22:39.350 HUIZINGA: That’s … yes, it does. 00:22:39.350 --> 00:22:44.250 And I imagine, you know, if you go to the website and you see those, kind of, three 00:22:44.250 --> 00:22:50.529 main areas, I imagine that even under there, there are specific projects on, you know, 00:22:50.529 --> 00:22:52.600 how then do we iterate? 00:22:52.600 --> 00:22:53.600 How then do we explore? 00:22:53.600 --> 00:22:54.600 HOSN: That’s right. 00:22:54.600 --> 00:22:58.200 That’s a good plug for people to visit the AI Frontiers website! 00:22:58.200 --> 00:23:00.029 Thank you, Gretchen! 00:23:00.029 --> 00:23:01.029 [LAUGHS] 00:23:01.029 --> 00:23:05.760 HUIZINGA: Well, I’ve been intrigued for a while by this idea of what you’ve called 00:23:05.760 --> 00:23:11.090 bi-directional enrichment, which represents both how research informs product but also 00:23:11.090 --> 00:23:16.380 how product informs research, but you’ve recently talked about how this idea has expanded 00:23:16.380 --> 00:23:21.320 to embrace what you call multi-directional enrichment and co-innovation. 00:23:21.320 --> 00:23:24.330 So what do you mean by that, and what does it look like for you? 00:23:24.330 --> 00:23:33.429 HOSN: So we talked just moments ago how the time has shrunk tremendously in artificial 00:23:33.429 --> 00:23:38.260 intelligence and the speed at which innovations are coming out. 00:23:38.260 --> 00:23:44.990 So what does that mean when you are sitting in research and you’re trying to derive 00:23:44.990 --> 00:23:47.870 value for Microsoft, for example? 00:23:47.870 --> 00:23:57.950 It means that now, rather than going on a journey to try out you know different things, 00:23:57.950 --> 00:24:04.970 what you want is for product to come on a co-innovation journey with you. 00:24:04.970 --> 00:24:11.850 And not every team has the capability or the time or the resources to do it. 00:24:11.850 --> 00:24:20.210 But sometimes product teams have applied scientists that are asking themselves very similar questions. 00:24:20.210 --> 00:24:26.970 And so now we have this huge synergistic effect by which, you know, researchers can come and 00:24:26.970 --> 00:24:35.720 explore their research but anchor them in a real-world scenario that the product team 00:24:35.720 --> 00:24:38.040 is, you know, asking themselves about. 00:24:38.040 --> 00:24:41.049 And that’s what I mean by co-innovation. 00:24:41.049 --> 00:24:46.429 And we look for co-innovation, so these are product teams or applied scientists in product 00:24:46.429 --> 00:24:49.470 teams that are not looking at something I can ship tomorrow. 00:24:49.470 --> 00:24:52.520 Because that’s not … that’s not frontiers. 00:24:52.520 --> 00:24:55.690 That’s feature-function that they can deliver right now to their customers. 00:24:55.690 --> 00:25:01.659 When we co-innovate, we have to co-innovate on a bit of a longer timespan. 00:25:01.659 --> 00:25:03.539 Now it’s no longer years, right? 00:25:03.539 --> 00:25:09.190 With generative AI, everything is months, but nonetheless, this is not next week. 00:25:09.190 --> 00:25:11.220 This is in a few months. 00:25:11.220 --> 00:25:16.539 And so … but this is really, really great because, again, I keep saying this and I have 00:25:16.539 --> 00:25:23.590 maybe a huge bias, but I do believe that research, without it being anchored in real-world scenario, 00:25:23.590 --> 00:25:28.799 just doesn’t have the same effect. 00:25:28.799 --> 00:25:30.270 So I have a bias for that. 00:25:30.270 --> 00:25:32.120 It’s my PM hat, what can I say? 00:25:32.120 --> 00:25:33.770 I love real-world scenarios! 00:25:33.770 --> 00:25:34.770 [LAUGHTER] 00:25:34.770 --> 00:25:38.590 HUIZINGA: What you just referred to is an interesting flow. 00:25:38.590 --> 00:25:45.000 I’ve noticed in my years doing this podcast that some people that started in research 00:25:45.000 --> 00:25:50.940 ended up over in product—and we’ll call them embedded researchers, if you will—and 00:25:50.940 --> 00:25:55.580 then some people that were in a product scenario come back over to research. 00:25:55.580 --> 00:26:03.649 And so, there’s this flow, multi-directional, bi-directional, and also where they’re placed 00:26:03.649 --> 00:26:05.590 within the company. 00:26:05.590 --> 00:26:13.130 How do you see that flow and the value of that flow between these organizations? 00:26:13.130 --> 00:26:23.330 HOSN: Yeah, you know, like, I think that the flow is important because that’s how cross-pollination 00:26:23.330 --> 00:26:24.330 happens. 00:26:24.330 --> 00:26:26.590 And you talked about brilliant minds. 00:26:26.590 --> 00:26:31.029 In product teams, there are brilliant minds, as well, right. 00:26:31.029 --> 00:26:39.380 And although their focus area is more around the product they live and breathe every day, 00:26:39.380 --> 00:26:47.270 this is enriching to researchers and continues to be enriching because when you deploy research 00:26:47.270 --> 00:26:53.350 capabilities in a real-world setting, there are surprising new research questions that 00:26:53.350 --> 00:26:55.080 come up, not just engineering. 00:26:55.080 --> 00:27:00.169 A lot of times people think of research, OK, yeah, you scale it, you harden it, you secure 00:27:00.169 --> 00:27:01.830 it, and it’s good to go. 00:27:01.830 --> 00:27:03.809 But that’s not always the case. 00:27:03.809 --> 00:27:11.100 In a lot of cases, because of the interactivity that happens with real-world scenarios, it 00:27:11.100 --> 00:27:13.810 opened up brand-new paths for research. 00:27:13.810 --> 00:27:18.980 And so I think that flow continues to happen even now. 00:27:18.980 --> 00:27:20.809 It’s just compressed. 00:27:20.809 --> 00:27:25.350 It’s just that researchers are no longer thinking six years. 00:27:25.350 --> 00:27:27.200 Researchers are thinking three months. 00:27:27.200 --> 00:27:30.200 Like, what am I going to do in three months? 00:27:30.200 --> 00:27:34.490 Because in three months, there will be a hundred other researchers that are coming up with 00:27:34.490 --> 00:27:37.320 innovation on the same question. 00:27:37.320 --> 00:27:39.510 So I think the flow still exists. 00:27:39.510 --> 00:27:42.890 I think that time has shrunk. 00:27:42.890 --> 00:27:50.890 And I think the mobility from researchers and research going to product and vice versa 00:27:50.890 --> 00:27:55.120 is enriching for the people that do it because you gain different perspectives. 00:27:55.120 --> 00:28:00.690 HUIZINGA: Well, and let’s push in even there a little bit. 00:28:00.690 --> 00:28:04.630 Researchers like everyone else can get comfortable looking at things through a particular lens. 00:28:04.630 --> 00:28:07.970 I would say that’s a human trait, not just a research trait … 00:28:07.970 --> 00:28:08.970 HOSN: Absolutely. 00:28:08.970 --> 00:28:11.500 HUIZINGA: … until a disruption challenges their status quo. 00:28:11.500 --> 00:28:16.909 So you’ve talked about LLMs, which we’ve called large language models, as being a good 00:28:16.909 --> 00:28:22.850 forcing function for researchers to think differently, even about the questions they’re 00:28:22.850 --> 00:28:23.850 asking. 00:28:23.850 --> 00:28:24.899 Can you elaborate on that a little bit? 00:28:24.899 --> 00:28:35.130 HOSN: Yeah, yeah, so, you know, the large language models and this disruption that we 00:28:35.130 --> 00:28:42.230 are living in at the moment is lighting fire underneath a lot of people’s intellect, 00:28:42.230 --> 00:28:45.320 I’m going to say. 00:28:45.320 --> 00:28:54.559 And so I think that people have to adapt quickly to change. 00:28:54.559 --> 00:28:56.350 And this is key. 00:28:56.350 --> 00:29:03.190 Adaptability, I believe, is just a key ingredient in doing research nowadays. 00:29:03.190 --> 00:29:04.669 Why? 00:29:04.669 --> 00:29:10.809 Because a lot of people are thinking directionally the same. 00:29:10.809 --> 00:29:17.600 And so, you know, if you’re not the first, you’re going to have to adapt to what came 00:29:17.600 --> 00:29:18.600 out. 00:29:18.600 --> 00:29:21.460 And then you have to think of, how do I differentiate? 00:29:21.460 --> 00:29:25.590 So the second point I would say is differentiation. 00:29:25.590 --> 00:29:31.570 And this mindset of, you know, how do I adapt to what just came out? 00:29:31.570 --> 00:29:33.929 How do I differentiate? 00:29:33.929 --> 00:29:39.100 And then—Rafah’s bias—how do I anchor in real-world scenario? 00:29:39.100 --> 00:29:41.380 This is the home run. 00:29:41.380 --> 00:29:47.320 And I would say you package all of this and focus, focus, focus … and you get a gold 00:29:47.320 --> 00:29:48.320 mine. 00:29:48.320 --> 00:29:55.120 HUIZINGA: I’m hearing “yes, and …” in this response in the sense of not everyone’s 00:29:55.120 --> 00:29:58.500 going to be first, but then, what else? 00:29:58.500 --> 00:29:59.880 This is back to the frontiers. 00:29:59.880 --> 00:30:01.100 It’s like, how do I differentiate? 00:30:01.100 --> 00:30:02.660 Yes, that’s awesome. 00:30:02.660 --> 00:30:03.669 And we’ve got this … 00:30:03.669 --> 00:30:04.669 HOSN: Exactly. 00:30:04.669 --> 00:30:11.230 And how do I build on what has just been discovered and give it a little bit of an edge or push 00:30:11.230 --> 00:30:15.539 it a little further or take it in a brand-new direction? 00:30:15.539 --> 00:30:22.169 I mean, so many different possibilities, but it does take adaptability, like a flexibility 00:30:22.169 --> 00:30:24.140 in the mindset, I would say. 00:30:24.140 --> 00:30:25.640 HUIZINGA: Yeah. 00:30:25.640 --> 00:30:31.640 Well, let’s go back to what you alluded to earlier, this idea of responsible AI. 00:30:31.640 --> 00:30:34.220 This is a big deal at Microsoft. 00:30:34.220 --> 00:30:37.980 And researchers are very thoughtful about the question of what could possibly go wrong 00:30:37.980 --> 00:30:39.910 if we got everything right. 00:30:39.910 --> 00:30:45.320 But how does that translate practically, and what concrete steps are you taking at what 00:30:45.320 --> 00:30:48.539 I’ll call the “frontier of responsibility?” 00:30:48.539 --> 00:30:55.679 HOSN: Yeah, and as I mentioned, you know, being at the frontiers is amazing. 00:30:55.679 --> 00:30:57.820 It also holds a big responsibility. 00:30:57.820 --> 00:31:06.780 We have so many different, I would say, checks and balances that we use, in model training 00:31:06.780 --> 00:31:15.899 and fine-tuning, to ensure that we are on top of all the regulatory, the policymaker 00:31:15.899 --> 00:31:24.850 suggestions, and we are abiding by Microsoft values first and foremost and responsibility 00:31:24.850 --> 00:31:27.790 in creating these innovations. 00:31:27.790 --> 00:31:37.210 So practically and tactically, what happens is that there are processes for how you actually 00:31:37.210 --> 00:31:41.130 even release any type of model. 00:31:41.130 --> 00:31:43.441 And this is just research. 00:31:43.441 --> 00:31:49.340 And when it goes to product, they have their own compliance, you know, a stricter even 00:31:49.340 --> 00:31:53.740 compliance, I would say, process that they go through. 00:31:53.740 --> 00:32:00.230 So we try, and I try particularly, to partner with our privacy champions, with our legal 00:32:00.230 --> 00:32:06.669 champions, with our people that are looking at this from a responsible AI perspective, 00:32:06.669 --> 00:32:12.150 so that we bring them in early on, and we say, hey, we’re thinking of doing this. 00:32:12.150 --> 00:32:16.210 And they tell us, well, you know, if you’re thinking about it this way, you might want 00:32:16.210 --> 00:32:17.480 to consider this. 00:32:17.480 --> 00:32:22.820 So we’re trying to bring them in as early as possible so that also we don’t go all 00:32:22.820 --> 00:32:26.440 the way and then we discover we did something wrong, so we have to backtrack. 00:32:26.440 --> 00:32:34.450 So I would say, you know, having these partners and colleagues come in early in the game just 00:32:34.450 --> 00:32:37.399 saves everybody a lot of time. 00:32:37.399 --> 00:32:44.190 And all this responsible AI for us, it’s ingrained with how we work, meaning we bring 00:32:44.190 --> 00:32:50.230 our champions early on and then we have them advise us as we move along the journey to 00:32:50.230 --> 00:32:51.230 create these innovations. 00:32:51.230 --> 00:32:55.380 So by the time we’re done, we know we’re good, right. 00:32:55.380 --> 00:33:00.740 And even by the time we’re done, we recheck everything, we run a lot of evaluation benchmarks, 00:33:00.740 --> 00:33:05.750 and, you know, we do the right thing per policies at Microsoft. 00:33:05.750 --> 00:33:07.890 So we take it very, very seriously. 00:33:07.890 --> 00:33:13.529 HUIZINGA: Well, let’s go back to this idea of research horizons for a second and anchor 00:33:13.529 --> 00:33:16.519 it in the way that we approach research. 00:33:16.519 --> 00:33:21.879 So many ideas are basically iterative steps on existing work, and they make a lot of sense 00:33:21.879 --> 00:33:26.480 … this is the next step … but then there are those out-of-the-box ideas that feel like 00:33:26.480 --> 00:33:31.429 maybe bigger swings—some might even call them outrageous—and in organizations like 00:33:31.429 --> 00:33:34.600 Microsoft Research, they might get the green light, too. 00:33:34.600 --> 00:33:42.370 Where do you find this idea of the outrageous or maybe longer-term idea finding a home or 00:33:42.370 --> 00:33:46.889 a place in an organization like Microsoft Research, and have you ever worked on something 00:33:46.889 --> 00:33:48.760 that felt outrageous to you? 00:33:48.760 --> 00:33:52.529 HOSN: Umm, you know, we like outrageous! 00:33:52.529 --> 00:33:55.570 That’s why we’re in research, right? 00:33:55.570 --> 00:34:00.659 So outrageous is good. 00:34:00.659 --> 00:34:07.950 I haven’t, to be honest, worked on an outrageous, but I am confident I will be. 00:34:07.950 --> 00:34:15.710 So … [LAUGHTER] I just have this belief that in AI Frontiers, we are going to have 00:34:15.710 --> 00:34:22.359 outrageous ideas, and we’re going to work on them, and we’re going to make bets that 00:34:22.359 --> 00:34:29.030 basically are hard to make in other parts of the company because we have the privilege 00:34:29.030 --> 00:34:33.480 of taking them and pursuing them. 00:34:33.480 --> 00:34:38.589 And, yes, they may fail, but if we have a breakthrough, it will be a significant breakthrough. 00:34:38.589 --> 00:34:41.879 So, so I think that outrageous is good. 00:34:41.879 --> 00:34:43.700 We need to think big. 00:34:43.700 --> 00:34:47.099 We need to take big leaps, big ideas. 00:34:47.099 --> 00:34:51.600 We also need to know how to fail gracefully and pivot fast! 00:34:51.600 --> 00:34:53.240 HUIZINGA: Hmmm. 00:34:53.240 --> 00:34:54.240 Mmm. 00:34:54.240 --> 00:34:59.800 You know, it strikes me, and I’m laughing to myself, it strikes me, even as we’re 00:34:59.800 --> 00:35:06.010 talking, that the idea that you work in AI Frontiers, that’s outrageous to most people 00:35:06.010 --> 00:35:08.220 and, and it’s normal to you. 00:35:08.220 --> 00:35:13.250 So maybe this idea of, “I haven’t worked on anything outrageous” is like, no, you 00:35:13.250 --> 00:35:16.650 live in outrageous, it just doesn’t seem like it! 00:35:16.650 --> 00:35:17.650 [LAUGHTER] 00:35:17.650 --> 00:35:18.650 HOSN: Maybe. 00:35:18.650 --> 00:35:20.520 It’s my day-to-day job, so, yes, I guess you’re right. 00:35:20.520 --> 00:35:21.520 HUIZINGA: Right. 00:35:21.520 --> 00:35:25.560 I mean, yeah, you say, we love outrageous, and that’s where it is right now. 00:35:25.560 --> 00:35:32.599 Every day that I follow, sort of, AI Twitter also and find myself going, seriously? 00:35:32.599 --> 00:35:34.750 That happened yesterday? 00:35:34.750 --> 00:35:35.750 What next? 00:35:35.750 --> 00:35:37.329 HOSN: Yeah, in two hours, there’ll be yet another thing. 00:35:37.329 --> 00:35:41.880 So, yeah, I guess I am living in outrageous, and I love it! 00:35:41.880 --> 00:35:43.470 It’s amazing! 00:35:43.470 --> 00:35:44.470 [LAUGHS] 00:35:44.470 --> 00:35:49.380 HUIZINGA: Yeah, maybe the idea of outrageous is just changed. 00:35:49.380 --> 00:35:51.110 HOSN: You know, you’re so right. 00:35:51.110 --> 00:35:54.180 I think that it’s become the norm. 00:35:54.180 --> 00:36:08.859 And it is, once we anchor in generative AI, and we push further on this idea, maybe we 00:36:08.859 --> 00:36:14.070 will go back in a cycle where outrageous is outrageous, but today it’s our life. 00:36:14.070 --> 00:36:15.500 It’s where we live. 00:36:15.500 --> 00:36:17.500 It’s what we breathe every day. 00:36:17.500 --> 00:36:20.579 So it’s become a norm. 00:36:20.579 --> 00:36:21.880 HUIZINGA: Yeah. 00:36:21.880 --> 00:36:28.250 Well, as we close, Rafah, I want to ask a question anchored on the big idea behind AI 00:36:28.250 --> 00:36:30.310 Frontiers. 00:36:30.310 --> 00:36:35.450 What do you believe might be true in say 10 to 15 years, and what should we be doing about 00:36:35.450 --> 00:36:36.450 it now? 00:36:36.450 --> 00:36:40.450 In other words, how does what we believe about the future influence how we conceptualize 00:36:40.450 --> 00:36:42.400 and execute on ideas today? 00:36:42.400 --> 00:36:47.200 HOSN: Yeah, you know, it’s … I can’t even predict what I’m going to be doing 00:36:47.200 --> 00:36:48.200 tomorrow! 00:36:48.200 --> 00:36:51.959 But … [LAUGHTER] here’s, here’s what I think. 00:36:51.959 --> 00:37:01.880 I think that we are truly approaching a moment in human history where a lot of unsurmountable 00:37:01.880 --> 00:37:11.340 problems, like very hard-to-tackle diseases that have been so hard, I think we are approaching 00:37:11.340 --> 00:37:19.200 a moment, you know, soon, I hope it’s even sooner than 10 years, where generative AI 00:37:19.200 --> 00:37:25.660 and innovations on top of it could lead to a lot of resolution for things that today 00:37:25.660 --> 00:37:29.200 … that cause unsurmountable pain and suffering. 00:37:29.200 --> 00:37:37.890 I’m very hopeful that with what we are creating that we can, you know, take inefficiencies 00:37:37.890 --> 00:37:45.640 out of so many different things that we see today that take time so that we liberate ourselves 00:37:45.640 --> 00:37:49.710 to think about the “what next” societally, right? 00:37:49.710 --> 00:37:58.220 I think what we need to be doing right now, to be honest, to influence the future is think 00:37:58.220 --> 00:37:59.260 about our curricula. 00:37:59.260 --> 00:38:01.950 What are we going to teach our kids? 00:38:01.950 --> 00:38:04.349 What are they going to work in? 00:38:04.349 --> 00:38:10.310 This is where I’m hoping that we pour some of our creativity, education system. 00:38:10.310 --> 00:38:13.680 How are we preparing the next generation? 00:38:13.680 --> 00:38:18.859 What are the paths that we are going to forge for them, knowing what we know today, knowing 00:38:18.859 --> 00:38:21.830 what this technology can bring forth? 00:38:21.830 --> 00:38:25.089 So my hope is that we put some brain power into that. 00:38:25.089 --> 00:38:29.720 HUIZINGA: Rafah Hosn, it’s always a pleasure to talk to you. 00:38:29.720 --> 00:38:32.790 A sincere pleasure, a delight. 00:38:32.790 --> 00:38:34.890 Thanks for joining us today on Ideas. 00:38:34.890 --> 00:38:35.890 [MUSIC PLAYS] 00:38:35.890 --> 00:38:37.430 HOSN: Thank you so much for having me, Gretchen. 00:38:37.430 --> 00:38:38.020 [MUSIC FADES]