WEBVTT - A Conversation with Jacob Goldstein 0:00:04.440 --> 0:00:12.760 Welcome to tech Stuff, a production from iHeartRadio. Hey there, 0:00:12.760 --> 0:00:15.840 and welcome to tech Stuff. I'm your host Jonathan Strickland. 0:00:15.840 --> 0:00:19.440 I'm an executive producer with iHeart Podcasts and how the 0:00:19.480 --> 0:00:22.680 tech are you? Folks. We have a very very special 0:00:22.720 --> 0:00:25.360 episode of tech Stuff because I have a very very 0:00:25.400 --> 0:00:29.440 special guest with me today, Jacob Goldstein. Now, if I 0:00:29.480 --> 0:00:32.760 were to go and run down his entire resume, it 0:00:32.800 --> 0:00:35.600 would be an entire episode just by itself. He's got 0:00:35.600 --> 0:00:40.680 a long and distinguished career in journalism, and among his 0:00:41.040 --> 0:00:44.360 many accomplishments happens to be the fact that he's the 0:00:44.360 --> 0:00:48.279 host of a podcast called What's Your Problem, which is 0:00:48.520 --> 0:00:52.559 a show that really dives into things like engineering and 0:00:52.640 --> 0:00:55.800 how engineers tackle problems, how do they define them, and 0:00:55.800 --> 0:00:59.960 then how do they create solutions. So Jacob, welcome to 0:01:00.080 --> 0:01:01.640 tech Stuff. Thank you for being here. 0:01:02.120 --> 0:01:04.320 Hi, Jonathan, thanks so much for having me. I'm delighted 0:01:04.360 --> 0:01:04.760 to be here. 0:01:05.240 --> 0:01:08.160 Yeah, I'm delighted you're here too. And before we really 0:01:08.200 --> 0:01:10.840 dive into a full discussion, we're going to talk a 0:01:10.880 --> 0:01:13.440 lot about engineering and a lot about AI in particular, 0:01:13.520 --> 0:01:16.440 because many of your episodes in this last season of 0:01:16.480 --> 0:01:18.759 your show have been on AI. I want to learn 0:01:18.800 --> 0:01:21.280 more about you, So tell us a bit about your 0:01:21.319 --> 0:01:24.920 background and how you came to become a podcaster on 0:01:25.080 --> 0:01:25.960 What's Your Problem. 0:01:26.480 --> 0:01:29.360 So, before I started What's Your Problem, I was one 0:01:29.360 --> 0:01:31.640 of the co hosts of a podcast called Planet Money, 0:01:32.040 --> 0:01:35.840 which is a show about economics. And before I had 0:01:35.840 --> 0:01:39.160 that job, I didn't know that much about economics. You know, 0:01:39.200 --> 0:01:41.600 I was an English major in college. I'd covered healthcare 0:01:41.600 --> 0:01:44.800 for the Wall Street Journal, and so getting there and 0:01:44.880 --> 0:01:49.680 covering economics for a while, to me, the big exciting 0:01:49.920 --> 0:01:54.760 idea at the heart of economics is the pie can 0:01:54.840 --> 0:01:57.960 get bigger, right, everybody can be better off. The world 0:01:58.040 --> 0:02:01.560 is not a zero sum game, and I think that 0:02:01.640 --> 0:02:07.640 is a very non intuitive, big exciting idea. And basically 0:02:08.280 --> 0:02:11.480 the way we all get better off in the long 0:02:11.560 --> 0:02:15.960 run is through technology. Right, It's through people figuring out 0:02:16.440 --> 0:02:20.160 more efficient ways to do things, ways so that you know, 0:02:20.200 --> 0:02:22.680 we do the same amount of work basically, but we 0:02:22.720 --> 0:02:25.840 get more stuff. You get more output for every hour 0:02:25.880 --> 0:02:30.639 of labor. And that is fundamentally you know, engineering and technology, 0:02:30.680 --> 0:02:33.000 as you said, And so I wanted to go deeper 0:02:33.040 --> 0:02:36.080 on like how that actually works, Like, you know, there 0:02:36.080 --> 0:02:38.040 are people whose job is I'm going to go to 0:02:38.080 --> 0:02:40.520 work and like figure out a better way to do something, 0:02:40.639 --> 0:02:42.520 and so that is what I'm trying to do on 0:02:42.560 --> 0:02:44.080 What's Your Problem? Those are the kind of people I 0:02:44.120 --> 0:02:44.920 talk to on the show. 0:02:45.520 --> 0:02:48.880 That's awesome. I hear what you're saying because it resonates 0:02:48.880 --> 0:02:50.320 a lot with a lot of the stuff we talk 0:02:50.360 --> 0:02:53.080 about here on tech stuff. It is to me one 0:02:53.120 --> 0:02:58.040 of the key funny components about technology is how everyone 0:02:58.120 --> 0:03:01.840 anticipates the next big technology development means that their jobs 0:03:01.840 --> 0:03:05.720 are going to be way less labor intensive, and then 0:03:05.880 --> 0:03:09.840 often it turns out like, well, sure each individual task 0:03:09.960 --> 0:03:12.840 is easier, but now you're doing way more tasks because 0:03:12.880 --> 0:03:16.160 everyone's more efficient. So if you remember, because I'm not 0:03:16.160 --> 0:03:17.600 going to put an age on you, Jacob, but I 0:03:17.600 --> 0:03:19.239 will say that I am of a certain age where 0:03:19.280 --> 0:03:23.400 I remember the concept of the paperless office, yes, and 0:03:23.440 --> 0:03:26.320 how we were going to get to this incredibly efficient thing, 0:03:26.360 --> 0:03:29.160 and that maybe, like maybe your workday would be reduced 0:03:29.160 --> 0:03:31.440 to maybe three hours a day. Turns out that that 0:03:31.680 --> 0:03:34.760 was perhaps a bit idealistic on the part of the 0:03:34.800 --> 0:03:36.880 workers and not the way it worked out. 0:03:37.480 --> 0:03:40.240 Yes, although I mean there is a lot less paper 0:03:40.400 --> 0:03:43.000 to start with us, right, Like I am old enough 0:03:43.080 --> 0:03:46.080 to remember, like when I started working, everybody had like 0:03:46.120 --> 0:03:49.240 a file drawer by their desk and like hanging files 0:03:49.280 --> 0:03:52.080 with papers in them, so there is less paper. I mean, 0:03:52.200 --> 0:03:54.920 the sort of less work thing is interesting, right, because 0:03:54.960 --> 0:03:56.839 on the one hand, people are like, oh, I hope 0:03:56.880 --> 0:03:59.840 technology you'll mean I have less work. By the same token, 0:03:59.840 --> 0:04:04.960 people are like, oh I hope technology doesn't take my job, right, Yes, 0:04:05.400 --> 0:04:09.520 In fact, the basic mechanism is like technology the happy 0:04:09.600 --> 0:04:13.160 story anyways, the story that we hope happens is technology 0:04:13.200 --> 0:04:16.440 makes us more productive, not so that we can work less, 0:04:16.480 --> 0:04:19.120 but so that our output can be greater, right, Like 0:04:19.440 --> 0:04:23.080 you know, I did not start working on podcasts and 0:04:23.200 --> 0:04:26.080 radio in the real to real tape era, but I 0:04:26.120 --> 0:04:28.880 know people who did, and like they talk about how 0:04:28.920 --> 0:04:32.440 long it took to literally cut tape by hand, and 0:04:32.839 --> 0:04:34.880 you can cut a lot more tape now that it's 0:04:34.920 --> 0:04:37.479 not actual tape. Right, that's a productivity game. 0:04:37.760 --> 0:04:41.640 Yes, totally. It's so funny to kind of see these 0:04:41.839 --> 0:04:44.800 changes over time and the different perceptions that go into 0:04:44.839 --> 0:04:47.279 whether we was it going to be like versus what 0:04:47.400 --> 0:04:50.040 it actually turns out to be. We're gonna as I said, 0:04:50.080 --> 0:04:53.320 talk about AI a lot, and to your point, one 0:04:53.360 --> 0:04:56.919 of the things that I hear repeated about AI in 0:04:57.000 --> 0:05:00.640 general and specifically within the realm of robotics and AI 0:05:01.160 --> 0:05:04.680 is that its ideal role is to tackle tasks that 0:05:04.760 --> 0:05:08.000 fall into the three d's, which are dirty, dangerous, and dull. 0:05:08.360 --> 0:05:11.440 That these are technologies that are best suited to take 0:05:11.480 --> 0:05:15.200 on jobs that are perhaps less desirable for humans for 0:05:15.279 --> 0:05:19.240 various reasons, whether they could potentially cause injury, or they're 0:05:19.240 --> 0:05:23.840 not very rewarding, that sort of thing, And I really 0:05:23.960 --> 0:05:28.680 like that concept. The fear everyone has, obviously is that 0:05:28.760 --> 0:05:33.359 it's tackling everything. It's non discriminate, it's not looking at 0:05:33.640 --> 0:05:37.440 just the three d's, it's looking at every possible option. 0:05:37.600 --> 0:05:40.000 And you're seeing a lot of discourse, at least I 0:05:40.040 --> 0:05:42.839 am online where I see a lot of people saying, 0:05:43.400 --> 0:05:47.680 why aren't we looking at how to automate c suite jobs? 0:05:47.800 --> 0:05:49.800 Because it seems to me like a lot of the 0:05:49.960 --> 0:05:52.880 duties that c suite executives have are ones that would 0:05:52.920 --> 0:05:56.240 be best suited for some of the AI tools we're 0:05:56.600 --> 0:05:59.760 talking about. Why are we talking about eliminating these lower 0:05:59.839 --> 0:06:02.640 level jobs when some of these upper level ones particularly 0:06:02.680 --> 0:06:06.400 when you have stories about c suite executives who perhaps 0:06:06.440 --> 0:06:09.640 had a less than stellar run at the top of 0:06:09.680 --> 0:06:12.200 the company ladder, like maybe the company didn't perform as 0:06:12.240 --> 0:06:14.680 well as it should have, and yet they still retire 0:06:14.960 --> 0:06:18.400 with these massive packages. So it's funny to see how 0:06:18.640 --> 0:06:23.080 the perceptions of things like automation and AI are shaping 0:06:23.960 --> 0:06:25.760 social discussions in that way. 0:06:26.240 --> 0:06:28.919 Well, I think certainly. I mean, the rise of large 0:06:29.000 --> 0:06:33.120 language models, you know lms like chat GPT have shifted 0:06:33.120 --> 0:06:35.400 that conversation some, right. I think if you go back 0:06:35.440 --> 0:06:38.920 five years, people thought about, you know, automating warehouse jobs. 0:06:38.960 --> 0:06:42.080 But what I've seen in the last say year or 0:06:42.120 --> 0:06:46.680 so since chat GPT, you know, stormed, our discourse is 0:06:46.800 --> 0:06:50.880 people are talking a lot more about journalists and lawyers 0:06:51.279 --> 0:06:55.120 being automated away, right, and plausibly to some extent, plausibly 0:06:55.160 --> 0:06:57.240 to some margin. I mean. The other thing is like, 0:06:57.400 --> 0:07:00.160 do you want a robot boss? When people talk about like, 0:07:00.200 --> 0:07:03.000 I haven't heard about people talking about automating the c suite, 0:07:03.000 --> 0:07:06.400 but like, sure, bad bosses are bad, and overpaying bad 0:07:06.440 --> 0:07:10.320 bosses is bad. I actually think the role of a 0:07:10.360 --> 0:07:14.160 good boss, of a good CEO is largely to be 0:07:14.200 --> 0:07:18.360 a human being, right, is not fundamentally about you know, 0:07:18.440 --> 0:07:21.040 assessing the data and making the best decision, although obviously 0:07:21.040 --> 0:07:24.680 that's important. It's to be there to you know, talk 0:07:24.720 --> 0:07:26.960 to people, essentially, to be in the room to tell 0:07:27.000 --> 0:07:29.120 people that things are going to be okay. And that 0:07:29.320 --> 0:07:33.240 seems like the set of domains that are less likely 0:07:33.280 --> 0:07:35.119 to be automated, certainly in the short run. 0:07:35.560 --> 0:07:39.120 I think you're absolutely right. I think it's based on again, perception, right. 0:07:39.200 --> 0:07:44.480 A lot of people don't have like FaceTime with CEOs, 0:07:45.040 --> 0:07:48.960 and so their perception is from a distance, and they're 0:07:49.000 --> 0:07:51.920 looking at the effects that the CEO decisions are having 0:07:52.080 --> 0:07:54.920 on a broad level, especially in the wake of something 0:07:55.000 --> 0:07:57.600 like a round of layoffs, for example, Whereas you and 0:07:57.640 --> 0:08:00.440 I have had the opportunity to speak face to face 0:08:00.720 --> 0:08:03.880 and you find out very quickly these CEOs are human beings, 0:08:03.960 --> 0:08:06.120 some more so than others. I mean, there are some. 0:08:06.720 --> 0:08:09.880 There are some CEOs out there who I suspect maybe 0:08:09.920 --> 0:08:11.680 at least part cyborg, but. 0:08:12.200 --> 0:08:13.239 Could be had happen. 0:08:14.560 --> 0:08:18.000 Yeah, yeah, most of them. I mean, I'm not sure 0:08:18.000 --> 0:08:22.120 if Elon Musk is even robotic, he might be alien. 0:08:22.200 --> 0:08:24.800 I'm not entirely certain. But there are a lot of 0:08:24.800 --> 0:08:27.160 them out there. When you just have a short conversation 0:08:27.280 --> 0:08:30.160 and you realize these aren't just talking points. For a 0:08:30.160 --> 0:08:34.199 lot of these leaders, they sincerely believe their mission statement, 0:08:34.280 --> 0:08:37.840 or they sincerely believe in the strategies that they're following, 0:08:37.880 --> 0:08:40.960 and they sincerely feel bad when they have to make 0:08:41.000 --> 0:08:44.600 decisions that lead to things like layoffs. But when you 0:08:44.679 --> 0:08:47.000 are at more of a distance, I think it's easier 0:08:47.040 --> 0:08:52.040 to kind of dehumanize the person. And it's understandable, right 0:08:52.120 --> 0:08:55.679 you see those big effects and you just think this, Meanwhile, 0:08:55.679 --> 0:09:00.400 the CEO is potentially making enormous amounts of money. I'm 0:09:00.440 --> 0:09:03.360 reminded of a former CEO I worked for, not a 0:09:03.400 --> 0:09:06.440 former CEO. He's still a CEO, but he's a find, 0:09:06.440 --> 0:09:09.440 my former boss, David Zaslov. And whenever I see any 0:09:09.480 --> 0:09:12.480 stories about him, I sit there and think, well, I've 0:09:12.520 --> 0:09:15.120 met the man, I've had conversations with him. I know 0:09:15.440 --> 0:09:17.800 a little bit more. I feel like I could give 0:09:17.800 --> 0:09:19.920 a bit more perspective to this. But at the same time, 0:09:20.240 --> 0:09:24.439 you're not entirely wrong with some of the conclusions you've drawn. 0:09:24.720 --> 0:09:26.920 Yeah, and I mean, you know, to some extent, like 0:09:27.600 --> 0:09:29.240 I mean the sub I feel like a lot of 0:09:29.280 --> 0:09:31.840 the subtext of what you're talking about is inequality. 0:09:31.920 --> 0:09:32.120 Right. 0:09:32.120 --> 0:09:34.760 It's the gas between what CEOs make and whatever the 0:09:34.840 --> 0:09:38.480 median worker at their company makes, and that indeed ballooned 0:09:38.520 --> 0:09:40.760 out a lot at the end of the twentieth century 0:09:40.800 --> 0:09:43.600 and has stayed quite wide obviously, and to some extent, 0:09:43.679 --> 0:09:47.719 that is an effect of technology, right, although it's complicated, 0:09:47.760 --> 0:09:51.199 there's a lot of it is norms. Right. Yeah, that's 0:09:51.240 --> 0:09:52.439 a pretty subtle one. 0:09:53.200 --> 0:09:56.360 Yeah, that's another thing that as communicators about things that 0:09:56.400 --> 0:09:58.920 are in the technological space, often we do need to 0:09:58.960 --> 0:10:02.040 take a step back away from just the technology and 0:10:02.160 --> 0:10:08.800 acknowledge these other components that impact the entire direction of tech. 0:10:08.840 --> 0:10:10.960 I mean, I'm sure as someone who has looked into 0:10:10.960 --> 0:10:14.640 Silicon Valley you have seen how things like social norms 0:10:14.640 --> 0:10:18.719 and politics and even things like living expenses in San 0:10:18.760 --> 0:10:22.040 Francisco have a big impact on these sorts of things, 0:10:22.480 --> 0:10:25.480 and people can get frustrated when I step back and 0:10:25.520 --> 0:10:28.560 talk about these elements. But my argument is that you 0:10:28.600 --> 0:10:31.120 can't really have a full understanding of technology unless you 0:10:31.160 --> 0:10:34.040 also take into account these other things that do have 0:10:34.080 --> 0:10:35.959 an impact. But one of the things I wanted to 0:10:35.960 --> 0:10:38.720 ask you about so what's your problem. You are talking 0:10:38.720 --> 0:10:42.400 with a lot of problem solvers, obviously, and I was curious, 0:10:42.760 --> 0:10:45.160 now that you've spoken with quite a few people who 0:10:45.480 --> 0:10:47.960 either come directly from engineering or have kind of an 0:10:48.000 --> 0:10:52.200 engineering perspective, what's your take on engineers, Because I know 0:10:52.240 --> 0:10:54.720 that's a general question and not everyone falls into the 0:10:54.760 --> 0:10:59.200 same bucket, But I have a love for engineers in 0:10:59.240 --> 0:11:00.680 the way that they approach things. 0:11:01.360 --> 0:11:04.800 Yes, same, And you know, I also love engineers. And 0:11:04.840 --> 0:11:07.400 as I mentioned, I was an English major. I am 0:11:07.440 --> 0:11:09.800 not an engineer at all. I'm not even good at 0:11:09.840 --> 0:11:14.439 fixing things around the house, although I try. But in college, 0:11:14.440 --> 0:11:18.840 I took one computer science class my last term of college. 0:11:19.240 --> 0:11:22.280 It was just the intro class. And on the first 0:11:22.360 --> 0:11:25.000 day it was, you know, big lecture class. The professors 0:11:25.040 --> 0:11:27.440 computer scientist was talking about his grading system and it 0:11:27.440 --> 0:11:28.959 was this weird thing was like a check and a 0:11:29.080 --> 0:11:31.800 check plus. And he said the top grade is a 0:11:31.800 --> 0:11:35.960 plus plus and that is for code that makes me weep, 0:11:37.400 --> 0:11:40.000 like cause it's so beautiful, you know, so erien. And 0:11:40.040 --> 0:11:42.400 that was like a revelation to me because you know, 0:11:42.600 --> 0:11:45.560 as an English major, as a non engineer, I always 0:11:45.600 --> 0:11:48.200 thought of engineering as like, oh, a thing works or 0:11:48.200 --> 0:11:50.360 it doesn't work. The building falls down or it doesn't 0:11:50.360 --> 0:11:53.679 fall down. But this idea that there is elegance and 0:11:53.800 --> 0:11:57.640 beauty in the construction of the thing itself was really 0:11:57.679 --> 0:12:00.520 exciting to me and remains really exciting to me. And 0:12:00.840 --> 0:12:04.440 engineers really are like that, you know, like they love 0:12:04.600 --> 0:12:08.760 building things and they find beauty in an elegant solution, 0:12:09.360 --> 0:12:11.600 the way other people find beauty in a song or 0:12:11.640 --> 0:12:12.840 a poem or a painting. 0:12:13.280 --> 0:12:15.679 I love that. I would say there's like a spectrum 0:12:15.760 --> 0:12:19.600 in engineering as well, where you have sort of the artists, 0:12:19.640 --> 0:12:23.920 who are the ones who are very carefully creating and 0:12:24.080 --> 0:12:27.960 refining their code, and then maybe you have on the 0:12:28.000 --> 0:12:32.440 punk rock side the hackers who didn't necessarily build the thing, 0:12:32.760 --> 0:12:34.880 but they really want to know how the thing works, 0:12:34.920 --> 0:12:37.320 and they will take the thing down to the very 0:12:38.160 --> 0:12:40.840 base level of the structure, and then they'll say, what 0:12:40.920 --> 0:12:43.560 if I rebuild it so it does something else? Like 0:12:44.000 --> 0:12:48.200 I just think that entire culture, from the artists to 0:12:48.280 --> 0:12:51.560 the punk rockers, who, hey, they're artists. I'm a punk 0:12:51.920 --> 0:12:54.280 rock kind of fan myself, but I always find that 0:12:54.360 --> 0:12:57.240 to be a wonderful way to have a conversation is 0:12:57.280 --> 0:12:59.880 to talk about people and about how they're approach to 0:13:00.160 --> 0:13:02.480 sort of stuff. And I also always say whenever I 0:13:02.480 --> 0:13:04.520 talk with engineers, I come away with the feeling that 0:13:05.080 --> 0:13:07.920 they view the world, as you can think of it, 0:13:07.960 --> 0:13:10.400 as either a set of problems or a set of challenges, 0:13:10.559 --> 0:13:15.480 and they're constantly thinking about solutions, which is nice because 0:13:15.559 --> 0:13:17.520 I am unfortunately one of those people who's far more 0:13:17.600 --> 0:13:19.440 likely to point out a problem but not have a 0:13:19.440 --> 0:13:22.720 solution for you, right. So to talk to someone who's 0:13:22.760 --> 0:13:26.520 already thinking ahead about how to solve the problem, not 0:13:26.640 --> 0:13:29.200 just that there is a problem, I always find that 0:13:29.320 --> 0:13:30.040 very inspiring. 0:13:30.559 --> 0:13:33.000 Yeah, it's nice. I guess I also have tended to 0:13:33.040 --> 0:13:35.760 be problem focused, you know. I sort of came up 0:13:35.760 --> 0:13:38.920 in my career as a journalist, which is essentially all 0:13:38.960 --> 0:13:42.200 about pointing out problems right to a significant degree if 0:13:42.280 --> 0:13:45.760 you read the paper. Basically, what's going on in many 0:13:45.840 --> 0:13:48.440 many stories is here is a thing that is bad, 0:13:50.000 --> 0:13:53.520 and so talking to people who are trying to make 0:13:53.559 --> 0:13:56.440 things better or fix things is great. And you know, 0:13:56.559 --> 0:13:59.720 to be clear, I don't want to be too Pollyanna 0:13:59.760 --> 0:14:03.240 issue here, Like there are plenty of engineers who build 0:14:03.280 --> 0:14:05.959 things that on net don't help the world right, Building 0:14:06.200 --> 0:14:08.719 new things is not always helpful to the world, and 0:14:08.920 --> 0:14:13.199 there are certainly engineers who become enamored of just building 0:14:13.240 --> 0:14:15.240 the thing and don't think about what it might mean. 0:14:15.559 --> 0:14:19.280 And frankly, you know, in choosing who to talk to 0:14:19.400 --> 0:14:21.840 for the show, I do try and talk to people 0:14:21.880 --> 0:14:25.120 who I think are some combination of well cognizant of 0:14:25.120 --> 0:14:27.760 what they're building and actually trying to do a good thing, 0:14:27.920 --> 0:14:30.840 and you know, aware of the fact that there might 0:14:30.880 --> 0:14:34.640 be unintended bad consequences of the thing that they're building. 0:14:35.320 --> 0:14:37.640 Maybe I don't always succeed, but it is a useful 0:14:38.200 --> 0:14:39.640 frame totally. 0:14:39.960 --> 0:14:43.280 I also tend to think about how when engineers build 0:14:43.320 --> 0:14:47.240 things that make sense to them, assuming that's they're building 0:14:47.240 --> 0:14:49.160 something that ultimately is supposed to be used by the 0:14:49.200 --> 0:14:51.960 general public, the great ones will take into account how 0:14:52.480 --> 0:14:57.000 a quote unquote normal person would approach the whatever it is. 0:14:57.280 --> 0:14:57.360 Like. 0:14:57.400 --> 0:15:01.040 I'm thinking of user interfaces in particular, and that making 0:15:01.080 --> 0:15:03.240 sure that the user interface is going to make sense 0:15:03.280 --> 0:15:06.520 to enore me as opposed to an engineer. And then 0:15:06.560 --> 0:15:10.560 there are other engineers or in fact entire companies where 0:15:10.560 --> 0:15:13.680 they will build things that work great. If you're an 0:15:13.720 --> 0:15:17.200 engineer they're fantastic if you're an engineer. If you're not 0:15:17.280 --> 0:15:21.040 an engineer, it may require a bit more work on 0:15:21.080 --> 0:15:24.360 your part. I'm looking specifically at Google and the Android 0:15:24.360 --> 0:15:27.680 operating System because I'm an Android user, but at the 0:15:27.680 --> 0:15:32.080 same time, I fully recognize that iOS is an operating 0:15:32.120 --> 0:15:35.880 system that is so intuitive. You can literally hand an 0:15:35.920 --> 0:15:38.760 iOS device to a child and they will have it 0:15:38.840 --> 0:15:41.320 figured out in no time. You can hand an Android 0:15:41.360 --> 0:15:44.120 device to someone and they will spend a lot of 0:15:44.120 --> 0:15:46.640 time asking questions about how to do things and how 0:15:46.640 --> 0:15:49.880 to access things. And it's not that the Android operating 0:15:49.880 --> 0:15:52.160 system is bad, it's not that it's worse than iOS. 0:15:52.320 --> 0:15:54.840 If you happen to be an engineer, Android is awesome, 0:15:55.320 --> 0:15:57.600 and if you're not, it's still awesome. But you have 0:15:57.680 --> 0:16:01.120 to put in work to get to realize that. And 0:16:01.160 --> 0:16:04.680 it really differentiates the two, right because Apple has always 0:16:04.760 --> 0:16:08.600 had a focus on how can we make this into 0:16:08.640 --> 0:16:12.600 a product that people realize they need, even if they 0:16:12.640 --> 0:16:15.600 never had that need before. And Google's like, how can 0:16:15.640 --> 0:16:17.960 we make this so that it's really powerful and that 0:16:18.040 --> 0:16:19.960 it does what we wanted to do? But it may 0:16:20.040 --> 0:16:22.680 require a little bit of work on the user's part. 0:16:22.720 --> 0:16:25.480 In order to have it work out. So that's sort 0:16:25.520 --> 0:16:28.680 of also a fascinating thing about engineering that I've really 0:16:29.240 --> 0:16:31.920 loved to look at and to talk about. I don't 0:16:31.960 --> 0:16:35.040 necessarily think one is better than the other, apart from 0:16:35.040 --> 0:16:37.160 the fact that one is just much easier for the 0:16:37.240 --> 0:16:40.480 general public to kind of glom onto. As much as 0:16:40.520 --> 0:16:43.280 I love Android, I would never say that it's more 0:16:43.360 --> 0:16:44.680 user friendly than iOS. 0:16:46.000 --> 0:16:48.720 Yeah, I mean there's a few ways of thing about that, right, 0:16:48.800 --> 0:16:51.120 Like somebody a long time ago told me there's a 0:16:51.160 --> 0:16:54.680 phrase people use sometimes the user is never wrong, right, 0:16:54.840 --> 0:16:57.080 which is which is an interesting framework. And I think 0:16:57.160 --> 0:16:58.800 they told it to me to be nice, because I 0:16:58.840 --> 0:17:01.240 was like trying to do something in a radio studio 0:17:01.280 --> 0:17:02.840 and I couldn't figure it out. And it was an 0:17:02.840 --> 0:17:05.440 engineer who was like a thoughtful guy. I said, no, no, 0:17:05.480 --> 0:17:08.000 you're not bad at this, it's just not set up well. 0:17:08.119 --> 0:17:10.480 I mean, the other way of thinking about it, in 0:17:10.560 --> 0:17:15.760 terms of the Android versus io ask question is as 0:17:15.800 --> 0:17:18.520 an optimization problem. Right, if you're an engineer, then the 0:17:18.560 --> 0:17:21.360 question is, well, are we optimizing for sort of the 0:17:21.880 --> 0:17:24.240 mobile operating system that can do the most things or 0:17:24.280 --> 0:17:26.560 be the most flexible, or the mobile operating system that 0:17:26.680 --> 0:17:28.920 is just like bulletproof. You can give it to anybody 0:17:28.920 --> 0:17:31.399 in any language and they will immediately understand what it 0:17:31.480 --> 0:17:34.960 is and you get different sort of solutions depending on 0:17:34.960 --> 0:17:36.000 what you're optimizing for. 0:17:36.359 --> 0:17:39.280 Yeah, that's true. Like you've identified whatever your goal is, 0:17:39.320 --> 0:17:43.040 so obviously the execution is going to be different. Well, 0:17:43.080 --> 0:17:45.040 I'm glad that you weren't told that the problem was 0:17:45.040 --> 0:17:49.680 between keyboard and chair, which is the other classic classic answer. 0:17:49.720 --> 0:17:50.960 Wait a minute, that's me. 0:17:51.560 --> 0:17:55.080 Yeah, I've received that particular one more than more than 0:17:55.119 --> 0:17:58.159 once in my lifetime. Well, now you have a reply. 0:17:58.520 --> 0:18:02.880 Now you have to reply yes, yeah, yes, that's true. 0:18:02.880 --> 0:18:06.160 The user is never wrong. We've got a lot more 0:18:06.240 --> 0:18:08.920 to talk about, Jacob and I, but before we get 0:18:08.920 --> 0:18:10.840 to the rest of our conversation, we need to take 0:18:10.880 --> 0:18:23.800 a quick break to thank our sponsors. Your show covers 0:18:24.080 --> 0:18:28.119 all realms of technology, not just AI, but because this 0:18:28.240 --> 0:18:32.720 past year has been undeniably AI centric when it comes 0:18:32.760 --> 0:18:35.600 to tech news, clearly a lot of your episodes do 0:18:35.880 --> 0:18:40.000 tackle AI, and as someone else who tries to communicate 0:18:40.040 --> 0:18:43.479 technology in a way that's really accessible and understandable, one 0:18:43.520 --> 0:18:46.760 of the things I frequently run into is that talking 0:18:46.800 --> 0:18:51.600 about AI in a responsible way is in itself challenging. 0:18:52.080 --> 0:18:54.920 But I'm curious to hear what your take is when 0:18:54.960 --> 0:18:57.440 it comes time for you to communicate about AI. How 0:18:57.480 --> 0:18:59.719 do you perceive that and how do you approach it? 0:19:00.200 --> 0:19:02.280 A thing in general that I try and do on 0:19:02.320 --> 0:19:06.320 my show and certainly with respect to AI, is to 0:19:06.440 --> 0:19:10.720 go narrow essentially right, Like I'm not going on my 0:19:10.760 --> 0:19:13.560 show and saying here is what AI is and here's 0:19:13.600 --> 0:19:15.960 what's going to happen. I'm talking to people who are 0:19:16.200 --> 0:19:21.200 typically building a company to do a specific thing with AI. 0:19:21.760 --> 0:19:25.280 Not quite always, but usually right. And so that to 0:19:25.400 --> 0:19:28.960 me is a helpful way to well, AA say something 0:19:29.040 --> 0:19:31.680 new because so many people are making so many broad 0:19:31.680 --> 0:19:36.080 statements about AI and b steer clear of you know, 0:19:36.640 --> 0:19:40.400 over generalization. How about you, I mean, what's your take? 0:19:41.080 --> 0:19:44.760 So my concern is I always want to avoid being 0:19:44.760 --> 0:19:49.560 reductive because AI is such a huge discipline and it 0:19:49.640 --> 0:19:52.919 involves so many different aspects, that is very easy to 0:19:52.960 --> 0:19:55.200 fall into the trap because I mean, clearly we see 0:19:55.200 --> 0:19:57.200 this in mainstream media all the time. Not that I 0:19:57.280 --> 0:20:00.440 blame them, but they're just they're taking some shortcuts where 0:20:00.440 --> 0:20:03.880 they'll use the term artificial intelligence. It almost implies that 0:20:04.040 --> 0:20:06.120 what they're talking about is the end all be all 0:20:06.160 --> 0:20:09.200 of artificial intelligence. And usually they're talking about generative AI. 0:20:09.440 --> 0:20:11.480 In the last year, I would say, like, that's been 0:20:11.760 --> 0:20:15.639 the biggest topic in artificial intelligence, but it's one topic, 0:20:15.880 --> 0:20:19.320 and that AI actually covers a lot more than that, 0:20:19.640 --> 0:20:22.639 and it falls into so many other buckets too, Like 0:20:23.040 --> 0:20:25.879 robotics obviously has a lot to do with AI. That 0:20:26.040 --> 0:20:30.120 not always you can have a fully remote controlled robot, 0:20:30.280 --> 0:20:35.040 but often there's some AI components there, things like assisted vision, 0:20:35.640 --> 0:20:38.880 brain computer interfaces. I mean, there's so many different things 0:20:38.920 --> 0:20:41.200 that don't have anything to do with generative AI. That's 0:20:41.240 --> 0:20:43.240 still at least touch on AI. 0:20:43.600 --> 0:20:46.640 Doing a Google search, yeah, like exactly. 0:20:46.800 --> 0:20:50.959 Yeah, anything that you're getting into, like automation, I mean 0:20:51.040 --> 0:20:51.600 you can get a. 0:20:51.600 --> 0:20:55.879 Lot recommending a show to you. Yeah, blind spot monitoring 0:20:55.920 --> 0:20:59.119 system saying there's a cardex too. Like these are all side. 0:20:59.200 --> 0:21:01.520 The Yeah, they're all they're all different aspects of AI. 0:21:01.680 --> 0:21:03.440 And like you could argue, well, sometimes you get a 0:21:03.440 --> 0:21:05.840 little fuzzy. I'm like, well, so is the word intelligence? 0:21:05.880 --> 0:21:08.159 Like intelligence itself is a fuzzy term. I thought that. 0:21:08.160 --> 0:21:11.960 Would be better off if the AI did not exist 0:21:12.480 --> 0:21:14.919 me too, like I think it's an unfortunate choice of 0:21:14.960 --> 0:21:17.000 words that is unhelpful. 0:21:17.119 --> 0:21:20.440 Ultimately, it's I think largely because everyone starts to jump 0:21:20.520 --> 0:21:23.400 to They jump to science fiction, they jump to Skynet, 0:21:23.560 --> 0:21:26.280 they jump to Terminator, they jump to this idea of 0:21:26.840 --> 0:21:30.320 something that appears to think the way humans do. And 0:21:30.400 --> 0:21:32.520 of course we don't even know if we ever reach 0:21:33.240 --> 0:21:36.320 strong AI or general AI, however you want to define it. 0:21:36.600 --> 0:21:39.359 We don't know if it's going to quote unquote think 0:21:39.440 --> 0:21:41.960 like a human, or even think at all. It may 0:21:42.080 --> 0:21:46.960 just be indistinguishable to us from the way humans think. 0:21:47.320 --> 0:21:50.240 And I think Touring would argue, well, that's good enough. 0:21:50.880 --> 0:21:53.520 It doesn't matter. If it's indistinguishable, then it might as 0:21:53.520 --> 0:21:53.840 well be. 0:21:54.080 --> 0:21:57.840 You quoted somebody on your show a while back as 0:21:57.920 --> 0:22:03.200 saying something like intelligence is whatever machines can't do yet, 0:22:03.880 --> 0:22:05.440 which I thought was pretty good. 0:22:06.119 --> 0:22:08.399 Yeah. I think that it's very similar to how a 0:22:08.400 --> 0:22:12.200 lot of philosophers define consciousness, right They say, like, ugh, 0:22:12.600 --> 0:22:13.120 we don't. 0:22:12.960 --> 0:22:13.840 Know what consciousness is. 0:22:14.200 --> 0:22:17.199 All we do. All we do is we define what 0:22:17.320 --> 0:22:20.000 consciousness isn't we haven't gotten to a point where we 0:22:20.000 --> 0:22:22.480 can say what consciousness is. We get we chip away 0:22:22.520 --> 0:22:25.720 So what we're doing is we've got the marble slab, yeah, 0:22:25.760 --> 0:22:28.520 and we're chipping at it, but we haven't yet seen 0:22:28.560 --> 0:22:32.000 the statue that's living under the slab. Yet we're still 0:22:32.040 --> 0:22:32.560 just chipping. 0:22:32.880 --> 0:22:35.840 I was already worried about intelligence. As you say consciousness, 0:22:35.880 --> 0:22:37.800 I like, I don't even know what to do with 0:22:37.880 --> 0:22:38.399 that one. 0:22:38.680 --> 0:22:41.199 Oh no, well, I mean it often goes hand in 0:22:41.240 --> 0:22:44.520 hand with AI, right, because people immediately assure that intelligence 0:22:44.680 --> 0:22:47.800 and self awareness go hand in hand with one another, 0:22:47.840 --> 0:22:50.119 and maybe it will. We don't know. That's the point. 0:22:50.200 --> 0:22:53.520 We don't know. But long story short, to answer the 0:22:53.600 --> 0:22:56.560 question of how I approach this, I usually go where 0:22:56.600 --> 0:22:59.760 I start from the broad foundation and then I go narrow. 0:22:59.840 --> 0:23:02.719 So I start with saying, first we need to acknowledge 0:23:02.800 --> 0:23:06.199 that artificial intelligence is a very very big field, and 0:23:06.240 --> 0:23:09.280 that this is one aspect of AI, and that we're 0:23:09.320 --> 0:23:11.160 not going to talk about the other aspects of AI, 0:23:11.240 --> 0:23:14.560 but we need to remember they exist, and that while 0:23:14.600 --> 0:23:16.600 the thing we're talking about is important, and while it 0:23:16.640 --> 0:23:19.600 has its own set of challenges and potential, you know, 0:23:19.680 --> 0:23:22.880 rewards and risks and all the things that go with it, 0:23:22.880 --> 0:23:25.679 it's one part. It's like if you were to say, 0:23:25.720 --> 0:23:28.359 you wouldn't hold up a remote control and say this 0:23:28.480 --> 0:23:31.400 is all of technology. Right, This is one thing that's 0:23:31.400 --> 0:23:34.920 a technological gadget, but it doesn't represent all of technology. 0:23:35.000 --> 0:23:37.760 First of all, you'd have to find it, right, that's. 0:23:37.560 --> 0:23:39.879 True, which you know, you know, you got to have 0:23:40.000 --> 0:23:42.040 some sort of method to figure out where it is. 0:23:42.200 --> 0:23:44.200 This is, by the way, is why three D television 0:23:44.280 --> 0:23:46.560 never became a thing. Who wants to look for glasses 0:23:46.600 --> 0:23:49.879 so that they can watch True Detective season four? Not me. 0:23:51.280 --> 0:23:54.080 So yeah, that's kind of my approach. And so I 0:23:54.080 --> 0:23:57.040 don't think that our approaches are that different. I think 0:23:57.119 --> 0:23:59.760 that we're pretty similar. And it's that I think we 0:24:00.080 --> 0:24:03.520 both feel there's a responsibility to make certain that we 0:24:03.640 --> 0:24:09.080 never overgeneralize or be reductive, because that feeds into a 0:24:09.280 --> 0:24:14.800 narrative that I think actually contributes to the old fud 0:24:14.880 --> 0:24:17.960 the fear, uncertainty, in doubt. And while there are things 0:24:18.000 --> 0:24:21.280 to certainly be concerned about and to be aware of, 0:24:21.720 --> 0:24:25.200 we don't want to rush into anything, you know, with 0:24:26.160 --> 0:24:29.359 a poor understanding of the situation. I think that's true 0:24:29.359 --> 0:24:33.280 whether you're you know, really enthusiastic and excited about AI, 0:24:33.720 --> 0:24:36.920 I think it's true if you are really concerned about AI. 0:24:37.119 --> 0:24:40.560 I think, you know, taking critical thinking and a really 0:24:40.680 --> 0:24:45.439 methodical approach is absolutely key if you want to avoid pitfalls. 0:24:46.040 --> 0:24:48.919 Sure, it seems hard to argue against critical thinking and 0:24:48.960 --> 0:24:51.199 a methodical approach, right, Who's going to take the other 0:24:51.280 --> 0:24:52.920 side of that one? 0:24:53.280 --> 0:24:55.160 I mean, Mark. 0:24:55.400 --> 0:24:57.520 Fair enough, everybody, ye, fair enough? 0:24:57.920 --> 0:24:59.200 Like Sam Altman maybe? 0:25:00.760 --> 0:25:03.400 I mean, so it is interesting to think about the 0:25:03.720 --> 0:25:06.560 Sam Altman, the head of Open AI. I mean, one 0:25:06.560 --> 0:25:09.399 of the really interesting things to me about AI. And 0:25:09.440 --> 0:25:13.240 that seems different in particular when you know, we're talking 0:25:13.280 --> 0:25:16.440 about engineers, Like, I feel like the extent to which 0:25:16.480 --> 0:25:20.119 the engineers working on AI are worried about AI is 0:25:20.200 --> 0:25:24.000 really interesting and different, right, I feel like the traditional 0:25:24.440 --> 0:25:27.080 kind of engineer stance is like, this thing is cool, 0:25:27.200 --> 0:25:30.879 let's build it right again. That's obviously reductive and somewhat unfair, 0:25:30.920 --> 0:25:37.760 but whatever. Whereas with AI, to some significant degree, many 0:25:37.800 --> 0:25:39.760 of the people who are most worried about it are 0:25:39.800 --> 0:25:42.000 the people who understand it the best. And you know, 0:25:42.040 --> 0:25:44.400 I've heard people are you like, oh, that's just marketing, 0:25:44.480 --> 0:25:46.760 and that doesn't seem true to me. First of all, 0:25:46.920 --> 0:25:48.720 why would you market a thing by saying we should 0:25:48.720 --> 0:25:50.560 be worried about it. And second of all, if you 0:25:50.640 --> 0:25:53.119 just look like open ai was started as a nonprofit 0:25:53.200 --> 0:25:55.000 and then they you know, needed more money, but they 0:25:55.000 --> 0:25:58.320 became this weird capped profit model, and then people left 0:25:58.359 --> 0:26:00.639 open ai to start anthropic, which is another one of 0:26:00.680 --> 0:26:03.399 the big ones, because they thought open ai wasn't worried enough. 0:26:03.520 --> 0:26:06.119 And then you had, you know, people calling like Elon 0:26:06.240 --> 0:26:08.720 Musk calling for a six month pause on AI development. 0:26:08.800 --> 0:26:11.240 And so I do think that people who know a 0:26:11.240 --> 0:26:14.040 lot of it about AI are in fact really worried 0:26:14.040 --> 0:26:17.520 about it, which is just interesting on its face, indifferent 0:26:17.560 --> 0:26:19.679 than the way technology often works. 0:26:20.000 --> 0:26:22.240 Yeah, there are a lot of conspiracy theories that pop 0:26:22.320 --> 0:26:24.400 up or fringe theories that pop up around this. By 0:26:24.400 --> 0:26:27.120 the way, like you have the ones who say, well, 0:26:27.359 --> 0:26:29.760 they say they're worried about AI because what they're trying 0:26:29.800 --> 0:26:34.240 to do is shape the discussions around regulations so that 0:26:34.440 --> 0:26:38.199 their own personal organization ends up benefiting from those regulations 0:26:38.240 --> 0:26:41.760 while those same regulations slow down smaller companies that are 0:26:41.760 --> 0:26:44.320 in the space. You had people saying, well, Elon Musk, yes, 0:26:44.359 --> 0:26:46.080 he was arguing that there needs to be a pause, 0:26:46.119 --> 0:26:48.280 but it's because he was launching his own AI company, 0:26:48.320 --> 0:26:50.280 and he wanted a chance to be able to catch up. 0:26:50.320 --> 0:26:52.679 Like You've got a lot of other fringe theories out there, 0:26:52.800 --> 0:26:55.400 and I understand that there may be, you know, some 0:26:55.560 --> 0:26:58.160 credibility to some of those who knows. But I think 0:26:58.240 --> 0:27:00.679 when it gets to a point where the board of 0:27:00.720 --> 0:27:05.880 directors of open Ai get together and decide seemingly spontaneously 0:27:06.400 --> 0:27:08.480 that they're going to get rid of the CEO and 0:27:08.560 --> 0:27:13.439 co founder of the company and then do so, I 0:27:13.480 --> 0:27:17.600 think that speaks to a genuine and sincere concern that 0:27:17.760 --> 0:27:21.640 perhaps the organization is moving in a direction that they 0:27:21.680 --> 0:27:26.840 feel as fundamentally counteractive to what they had intended. And 0:27:26.920 --> 0:27:30.560 of course we know they subsequently had to reverse that 0:27:30.640 --> 0:27:33.800 decision and step down from the board of directors because 0:27:33.800 --> 0:27:37.719 the overwhelming support within the organization was to that co 0:27:37.840 --> 0:27:43.520 founder and CEO seemingly embracing this new approach toward developing 0:27:43.560 --> 0:27:47.560 AI that was a departure from the original organization's intent. 0:27:48.119 --> 0:27:50.399 But to your point, the fact that the board of 0:27:50.400 --> 0:27:54.280 directors was willing to do such an extreme move, even 0:27:54.280 --> 0:27:55.600 though it was on a Friday at the end of 0:27:55.640 --> 0:27:58.320 a news cycle, even that they were willing to do 0:27:58.400 --> 0:28:02.040 that knowing that it would lead to them having to 0:28:02.119 --> 0:28:05.239 leave the organization. I think that speaks to me a 0:28:05.280 --> 0:28:09.160 genuine concern. You don't go and remove a co founder 0:28:09.200 --> 0:28:13.360 and CEO for a small reason. 0:28:14.280 --> 0:28:18.199 Yeah, I mean sure, I'm sure that some amount of 0:28:18.240 --> 0:28:20.959 the public worrying over AI by people in the field 0:28:21.280 --> 0:28:26.880 is some kind of self interested behavior. But I think overall, 0:28:27.400 --> 0:28:30.399 there are clearly a lot of people who know a 0:28:30.440 --> 0:28:33.320 lot who are even building these things, who are genuinely worried. 0:28:33.359 --> 0:28:34.879 Like that seems obviously true. 0:28:35.240 --> 0:28:40.080 Yeah, And honestly, like anytime someone's building a technology and 0:28:40.160 --> 0:28:43.160 they're they're bringing concerns up to me, that's a good thing. 0:28:43.360 --> 0:28:46.480 And it doesn't necessarily mean that the technology is ultimately 0:28:47.120 --> 0:28:50.520 harmful or not beneficial. But you know, I think it's 0:28:50.520 --> 0:28:54.160 a responsible person who does ask those questions. For one thing, 0:28:55.040 --> 0:28:57.280 it really can save you a lot of time and 0:28:57.400 --> 0:29:00.360 heartache further down the line, if you're tackling these kinds 0:29:00.360 --> 0:29:04.920 of things before they've escalated to a point where they're 0:29:04.960 --> 0:29:09.560 actively causing catastrophe. So I like seeing that. Whether it 0:29:09.640 --> 0:29:12.320 ends up being merited or not, well, that's just sort 0:29:12.360 --> 0:29:14.160 of a curse we have to bear, right, What if 0:29:14.640 --> 0:29:17.200 it turns out that it was never merited, we won't know. 0:29:17.640 --> 0:29:19.560 And if it turns out if it was merited, we 0:29:19.600 --> 0:29:21.880 still don't know. Because they asked the questions ahead of 0:29:21.880 --> 0:29:24.640 time and fix the problems before they became problems. And 0:29:24.880 --> 0:29:27.480 it's only if we take the other path that we 0:29:27.520 --> 0:29:30.520 find out for sure, like WHOA, we should have thought 0:29:30.520 --> 0:29:32.080 of this before we did it. 0:29:32.880 --> 0:29:35.880 I mean, you know, tools are complicated, right. People come 0:29:35.960 --> 0:29:38.760 up with new tools, and then other people use those 0:29:38.800 --> 0:29:42.080 tools in various ways, some of which enhance human well 0:29:42.120 --> 0:29:45.960 being and some of which cause new miseries. And plainly 0:29:46.160 --> 0:29:46.960 AI will do. 0:29:46.960 --> 0:29:50.200 Both, yes, and so it really becomes important that we 0:29:50.280 --> 0:29:53.240 are really good stewards of the technology and we're paying attention, 0:29:53.360 --> 0:29:55.680 that we're calling things out and we're addressing them as 0:29:55.720 --> 0:29:58.840 they come up. I don't think that we're that close 0:29:58.920 --> 0:30:03.040 yet to the doom day problem of the superhuman intelligent 0:30:03.080 --> 0:30:05.320 AI that's stuck in a box and then is convincing 0:30:05.320 --> 0:30:06.840 people to let it out of the box. I don't 0:30:06.840 --> 0:30:09.080 think we're close to that yet. I mean, even quote 0:30:09.120 --> 0:30:13.840 unquote dumb AI can do terrible things if it's poorly implemented. Right, 0:30:13.880 --> 0:30:19.400 We've seen that, We've seen accidents with autonomous cars, which 0:30:19.400 --> 0:30:24.800 show that AI can make bad choices sometimes because perhaps 0:30:24.840 --> 0:30:29.880 it encounters a scenario that no one anticipated, because as 0:30:29.920 --> 0:30:32.560 it turns out, reality has far more variables than we 0:30:32.560 --> 0:30:36.480 can account for when we're designing things and then something 0:30:36.840 --> 0:30:40.240 terrible happens. That doesn't mean that the technology itself is 0:30:41.400 --> 0:30:45.720 deeply flawed or bad, but it does highlight that we 0:30:46.040 --> 0:30:49.640 constantly have to be asking how can we make it better, 0:30:49.960 --> 0:30:51.880 and how can we make it safer, and how can 0:30:51.920 --> 0:30:55.040 we make it so that it's actually benefiting us and 0:30:55.120 --> 0:30:58.440 not just causing you know, maybe a little bit of 0:30:58.480 --> 0:31:02.080 benefit but a larger amount of harm. Yeah. 0:31:02.120 --> 0:31:05.200 I mean autonomous cars you mentioned are an interesting one, 0:31:05.560 --> 0:31:11.680 right because plainly there have been, you know, tragic crashes 0:31:11.960 --> 0:31:15.560 by autonomous cars. One question there is what are we 0:31:16.040 --> 0:31:20.400 benchmarking them against? Right, Like, there are tragic crashes with 0:31:20.520 --> 0:31:24.520 non autonomous cars every hour of every day, and so 0:31:24.760 --> 0:31:29.400 in a sort of if people were just mathematically rational optimizers, 0:31:29.440 --> 0:31:32.600 we would all say, okay, well, let's see if you know, 0:31:32.840 --> 0:31:37.120 over a million hours of driving, autonomous cars are safer 0:31:37.240 --> 0:31:41.200 or less safe than human drivers. That's clearly not what's happening. 0:31:41.240 --> 0:31:45.400 People clearly favor human drivers for some complicated set of 0:31:45.480 --> 0:31:50.960 human reasons. And we're not obviously benchmarking autonomous cars against humans, who, 0:31:51.000 --> 0:31:53.480 by the way, are terrible drivers. Like one thing about 0:31:53.560 --> 0:31:55.720 human beings. We're really bad at driving. 0:31:56.240 --> 0:31:58.840 Yeah, if you look at the stats in the United 0:31:58.880 --> 0:32:02.400 States for the number of fatalities and injuries that result 0:32:02.520 --> 0:32:07.200 from car accidents that are just human error caused car accidents, 0:32:07.400 --> 0:32:11.320 it's a staggering number. And when you think how much 0:32:11.400 --> 0:32:17.120 that could be reduced through autonomous cars, and you imagine, well, 0:32:17.360 --> 0:32:20.200 think of the ripple effect. It's not just the idea 0:32:20.280 --> 0:32:23.440 that those people who died would still be alive, which 0:32:23.800 --> 0:32:26.800 on its own is already a phenomenal thing to talk about. 0:32:27.080 --> 0:32:30.840 That means that the impact on those people's friends and 0:32:30.920 --> 0:32:33.680 families that would not have happened. It means that the 0:32:33.720 --> 0:32:37.120 impact on whatever their place of employment was that would 0:32:37.120 --> 0:32:40.160 not have happened. They would be contributing members of society. 0:32:40.280 --> 0:32:43.080 That would be a phenomenal change there. So when you 0:32:43.120 --> 0:32:47.360 start thinking about that, you realize the overall benefit is 0:32:47.440 --> 0:32:51.080 so huge that it only makes sense to really pursue 0:32:51.840 --> 0:32:56.040 autonomous vehicles. And as long as the data does show 0:32:56.080 --> 0:33:00.000 that in fact, they are better drivers per million miles 0:33:00.400 --> 0:33:03.360 than humans are, and that to me is something I 0:33:03.400 --> 0:33:05.680 try and keep in mind. You have to balance it out. 0:33:05.760 --> 0:33:08.200 I think it's the same thing as people who really 0:33:08.240 --> 0:33:11.240 flip out when they go on a flight they are 0:33:11.240 --> 0:33:14.520 not directly in control of the plane typically. I mean, 0:33:14.520 --> 0:33:16.280 if they're flipping out, whether they're the pilot, that's a 0:33:16.320 --> 0:33:19.120 whole different issue. But if you're going on a flight 0:33:19.160 --> 0:33:21.720 and you flip it out because you lack a sense 0:33:21.720 --> 0:33:24.040 of control, I feel like that's a very similar thing 0:33:24.120 --> 0:33:26.640 to how people feel when they're thinking about autonomous cars. 0:33:26.680 --> 0:33:30.000 It's that somehow the fact that someone's not in control 0:33:30.520 --> 0:33:34.680 brings up something very scary to a lot of people. 0:33:34.920 --> 0:33:38.600 It also raises other questions obviously, like accountability. Who do 0:33:38.640 --> 0:33:40.680 you hold accountable in these cases? I mean, there are 0:33:40.680 --> 0:33:43.440 a lot of questions that as a society we have 0:33:43.480 --> 0:33:46.720 to solve. It's not just the technology. But yeah, I 0:33:46.760 --> 0:33:49.520 agree with you that that gets complicated because it involves 0:33:49.560 --> 0:33:51.880 a lot of human feelings. And once you get to 0:33:51.960 --> 0:33:56.240 human feelings, the whole data and stats and everything kind 0:33:56.240 --> 0:34:00.440 of falls away. It's hard to convince someone who has 0:34:00.480 --> 0:34:06.240 a deep seated distrust of changing their mind just by 0:34:06.280 --> 0:34:09.239 showing them data, because they're always going to think of 0:34:09.320 --> 0:34:13.239 the things that fall outside the norm as being more 0:34:13.280 --> 0:34:16.479 important than the norm. Right, So if the accident rate 0:34:16.560 --> 0:34:20.120 per million miles is let's say, one tenth of what 0:34:20.239 --> 0:34:22.479 it would be for humans, they would still be looking 0:34:22.520 --> 0:34:25.439 at that tenth and not the nine tenths. Right. 0:34:25.960 --> 0:34:30.520 People don't think statistically, right they yeah, clearly, Like in general, 0:34:30.560 --> 0:34:34.680 statistics don't convince people of the way the world works. 0:34:34.760 --> 0:34:38.600 I do think, I mean I would have thought autonomous 0:34:38.680 --> 0:34:41.720 vehicles would have developed faster, right. They are the classic 0:34:41.760 --> 0:34:45.080 thing that's been five years away for fifteen years. Yeah, 0:34:45.120 --> 0:34:47.120 and they still feel five years away. Maybe they feel 0:34:47.160 --> 0:34:49.840 a little farther right, like five years ago they really 0:34:49.880 --> 0:34:51.640 felt five years away. It's like, okay, but this time 0:34:51.680 --> 0:34:53.839 we mean it. Look we got these, you know things 0:34:53.880 --> 0:34:56.719 for have around San Francisco. But I do think like 0:34:56.800 --> 0:34:59.120 that one which is AI by the way, right, it's 0:34:59.160 --> 0:35:04.800 basically computer vision is essential to that to autonomous cars. Uh, 0:35:05.080 --> 0:35:07.480 I feel like that one's gonna happen, don't you like? 0:35:08.200 --> 0:35:10.880 And yes, people be worried about it, but it's the 0:35:11.000 --> 0:35:13.839 kind of thing like you know, you drive, you ride 0:35:13.880 --> 0:35:15.680 in like a driverless train when you go to the 0:35:15.719 --> 0:35:18.520 airport and you take the train from Terminal A to 0:35:18.560 --> 0:35:22.920 Terminal C or whatever. And yes, obviously driving is more complicated, 0:35:22.920 --> 0:35:24.560 and obviously we're used to driving the car and not 0:35:24.640 --> 0:35:27.000 driving the train. It's not exactly the same, but like, 0:35:27.120 --> 0:35:30.719 you get used to it. People just get used to things, right, 0:35:30.880 --> 0:35:33.879 Like people didn't used to all walk around looking at 0:35:33.880 --> 0:35:36.440 their phones all the time, and now they do. And 0:35:37.040 --> 0:35:38.839 I'm old enough that it still seems a little weird 0:35:38.840 --> 0:35:41.359 to me. But I'm the weirdo for thinking it's weird, right, 0:35:41.400 --> 0:35:45.799 And I think that's gonna happen with driverless cars in Michael. 0:35:45.600 --> 0:35:48.160 Jacob, you just called me a weirdo, because I also 0:35:48.760 --> 0:35:51.680 I do too. Once in a while, I'll take my 0:35:51.719 --> 0:35:54.400 smartphone out and I'll just stop for a second and 0:35:54.440 --> 0:35:58.279 think I have a computer in my pocket. When I 0:35:58.400 --> 0:36:01.840 was a kid, a computer or my Apple to e 0:36:02.280 --> 0:36:06.319 was a fraction of of what I'm holding in my 0:36:06.480 --> 0:36:09.360 hand right now. I also have a device where I 0:36:09.360 --> 0:36:13.680 could contact pretty much anyone. I know. It was just 0:36:13.719 --> 0:36:16.400 a couple of like, like, it'll hit me for a second, 0:36:16.440 --> 0:36:18.880 it's most attentive. 0:36:18.560 --> 0:36:22.200 And no, maybe I should just check Twitter real quick, Like, yeah, 0:36:22.239 --> 0:36:25.160 it's a it's I mean, it's another tool that has 0:36:25.200 --> 0:36:28.200 like a complicated set of effects positive AGA. 0:36:28.440 --> 0:36:31.120 But you know, you had said, like, do do you 0:36:31.160 --> 0:36:33.160 think that autonomous cars are still going to be a thing? 0:36:33.200 --> 0:36:35.480 I absolutely do think they're going to be a thing. 0:36:35.560 --> 0:36:38.279 I think there there are companies out there that are 0:36:38.360 --> 0:36:43.600 so invested in it that it's going to happen. The 0:36:43.600 --> 0:36:48.239 timeline is really interesting. I would actually argue that some 0:36:48.440 --> 0:36:52.319 other issues in AI that are not related to autonomous 0:36:52.360 --> 0:36:57.239 cars could potentially keep that five years out going for 0:36:57.280 --> 0:37:01.480 a while. Because people have this concern about A I 0:37:01.480 --> 0:37:04.800 think they port that concern over to pretty much every 0:37:04.960 --> 0:37:08.840 kind of AI, whether it's warranted or not. Because the 0:37:09.200 --> 0:37:13.239 scary risks of generative AI, this idea that it's going 0:37:13.280 --> 0:37:15.560 to displace people out of their jobs and such, which 0:37:15.600 --> 0:37:19.440 it very may well do. They then kind of say like, well, 0:37:20.040 --> 0:37:24.800 that application of AI is is really seems very harmful 0:37:24.840 --> 0:37:28.560 to me. Then I think there's a tendency to kind 0:37:28.600 --> 0:37:30.520 of apply that, even if it's not the same sort 0:37:30.560 --> 0:37:34.239 of artificial intelligence to other implementations. And maybe I'm being 0:37:34.280 --> 0:37:38.200 a little too cynical with that, but because I've seen 0:37:38.400 --> 0:37:42.280 so much reporting go on where there isn't any effort 0:37:42.400 --> 0:37:46.200 made to distinguish between different types of artificial intelligence and 0:37:46.239 --> 0:37:49.520 what their purposes are and what their limitations are. It 0:37:49.920 --> 0:37:53.600 feels like we are conditioning the public, and by we, 0:37:53.719 --> 0:37:56.600 I mean like mass media conditioning the public to think 0:37:56.600 --> 0:37:59.800 of AI as all existing in this one single bucket. 0:38:00.360 --> 0:38:02.279 I feel like you consume a lot of really bad 0:38:02.400 --> 0:38:04.240 media based on what you've been saying. 0:38:05.000 --> 0:38:08.680 I mean, I'm reading articles all the time, and it's 0:38:08.719 --> 0:38:10.919 not that they're written poorly or that the people who 0:38:10.960 --> 0:38:14.839 write them are bad writers, but they are taking shortcuts, 0:38:15.680 --> 0:38:18.319 there's no getting around it, and those shortcuts I think 0:38:18.360 --> 0:38:23.160 are ultimately harmful. But then I also understand, especially if 0:38:23.160 --> 0:38:26.000 you're assigned to write a certain number of articles per week, 0:38:26.040 --> 0:38:27.759 you're probably not going to take the time to sit 0:38:27.800 --> 0:38:30.960 there and explain the intricacies of how this is different 0:38:31.000 --> 0:38:35.120 from every other implementation of artificial intelligence. But I certainly 0:38:35.120 --> 0:38:37.200 can take the time on my show, so I do. 0:38:39.840 --> 0:38:42.560 Jacob Goldstein of What's Your Problem has a lot more 0:38:42.600 --> 0:38:46.320 to say about tech and engineering and AI, But before 0:38:46.360 --> 0:38:59.239 we jump into that, let's take another quick break. Let's 0:38:59.280 --> 0:39:01.040 talk a little bit of about some of the episodes 0:39:01.080 --> 0:39:03.279 you've done on What's your Problem? Are there Are there 0:39:03.280 --> 0:39:06.360 any that kind of stand out as like a particularly 0:39:07.480 --> 0:39:13.480 fun or informative conversation, perhaps opening your eyes to something 0:39:13.520 --> 0:39:14.920 that you hadn't considered before. 0:39:15.520 --> 0:39:18.520 Yeah, yeah, a lot. Actually. You know, when you told 0:39:18.560 --> 0:39:21.840 me you wanted to talk about the AI shows that 0:39:21.880 --> 0:39:23.960 I've done, the interviews that I've done, I actually went 0:39:24.080 --> 0:39:27.080 back through the back catalog and you know, we listened 0:39:27.120 --> 0:39:29.640 to some shows and looked, and there really are a 0:39:29.680 --> 0:39:31.719 lot of them, as you said, in sort of some 0:39:31.760 --> 0:39:35.200 different domains, Like I've done a lot on AI and health, 0:39:35.480 --> 0:39:37.160 which is really interesting to me. I mean, one of 0:39:37.239 --> 0:39:40.200 the things that I try and find are places where 0:39:40.239 --> 0:39:43.319 it's like, oh, this is there's actually real stakes here, right. 0:39:43.360 --> 0:39:47.040 It's not just like, oh, making some kind of company 0:39:47.080 --> 0:39:49.279 I don't care about ten percent more profitable or whatever, 0:39:49.320 --> 0:39:51.279 which fine, like it's fine for people to do that, 0:39:51.320 --> 0:39:53.359 it's just not that interesting to me. Whereas with health, 0:39:53.400 --> 0:39:56.080 it's like, oh, if you can make it less likely 0:39:56.160 --> 0:39:59.799 for me and the people I love to die, I'm interested. 0:40:01.160 --> 0:40:05.160 When I just did recently, I interviewed this woman Succi Saria, 0:40:05.160 --> 0:40:07.239 who she's a professor at Johns Hopkins and She also 0:40:07.239 --> 0:40:10.880 has this company called Baesian Health, and her story is 0:40:10.920 --> 0:40:13.759 really interesting. She started out as a grad student. She 0:40:13.840 --> 0:40:16.200 was interested in AI and robots, and as she told 0:40:16.239 --> 0:40:17.400 me about it, she's like, you know, I was like 0:40:17.480 --> 0:40:19.440 trying to figure out how to make a robot juggle 0:40:19.560 --> 0:40:21.799 or whatever, just because it was fun. And she had 0:40:21.800 --> 0:40:24.279 this friend who was a doctor. She was a grad 0:40:24.320 --> 0:40:26.680 student at Stanford and her friend was a doctor at 0:40:26.719 --> 0:40:30.840 Stanford Hospital who was taken care of of premature babies 0:40:30.880 --> 0:40:34.240 in the neonatal intensive care unit. And this was about 0:40:34.239 --> 0:40:38.759 twelve years ago, and at this time hospitals were just 0:40:38.880 --> 0:40:42.080 starting to use electronic health records, which is kind of 0:40:42.120 --> 0:40:44.000 amazing that it was that late. Like we're talking like, 0:40:44.040 --> 0:40:45.759 you know, I don't know, twenty twelve or something. And 0:40:45.880 --> 0:40:47.880 it is one of the really interesting things to me 0:40:48.000 --> 0:40:51.080 about healthcare is in some ways it's super high tech, 0:40:51.120 --> 0:40:53.759 you know, like these crazy CT scanners and like, you know, 0:40:53.800 --> 0:40:56.879 everybody's got like bionic knees and amazing stuff. But when 0:40:56.880 --> 0:40:59.400 you get to actual like care at the bedside, like 0:40:59.560 --> 0:41:03.399 doctor treating patients in the hospital, it has remained rather 0:41:03.480 --> 0:41:06.600 old fashioned in many ways. Right, you know, twelve years 0:41:06.640 --> 0:41:10.480 ago it was still paper charts. Today, it's still doctors relying, 0:41:10.520 --> 0:41:12.720 you know, to a significant degree on evidence, but also 0:41:13.040 --> 0:41:17.720 to a significantry on essentially intuition. And so this computer scientist, 0:41:17.719 --> 0:41:20.560 who she's sorry it, basically decides, oh, I'm going to 0:41:20.600 --> 0:41:23.720 try and figure out how to use AI to make 0:41:24.000 --> 0:41:27.880 patient care in hospitals better. Like that's basically her big project. 0:41:27.960 --> 0:41:32.359 And she starts doing it with these premature babies twelve 0:41:32.400 --> 0:41:35.800 years ago and in fact figures out that by using 0:41:35.840 --> 0:41:38.840 this data that's now being captured in the electronic health record, 0:41:39.120 --> 0:41:42.640 she can build an AI model that can essentially better 0:41:42.719 --> 0:41:47.279 predict outcomes for these premature babies than the standard of care. 0:41:48.000 --> 0:41:51.880 But it's so early that it doesn't really go anywhere, right, Like, 0:41:51.920 --> 0:41:55.600 hospitals are just starting to use electronic health records, and 0:41:56.080 --> 0:41:58.960 a lot of doctors don't want to hear from some 0:41:59.200 --> 0:42:03.280 random computers scientists. They studied medicine for a long time 0:42:03.400 --> 0:42:05.480 and they've treated a lot of patients and they know 0:42:05.520 --> 0:42:08.760 what they're doing, and so it takes a long time, 0:42:09.520 --> 0:42:13.400 but she eventually starts this company, and more recently she 0:42:13.520 --> 0:42:16.480 decided to go after sepsis, which is this really common 0:42:16.480 --> 0:42:21.160 complication at hospitals. In hospitalized patients. It's basically a terrible 0:42:21.320 --> 0:42:25.399 infection to your body's response to infection, and you can 0:42:25.480 --> 0:42:28.759 die from it. Lots of people die from it. It's complicated, 0:42:28.800 --> 0:42:31.720 it's somewhat hard to diagnose. If you can diagnose it sooner, 0:42:33.040 --> 0:42:36.560 the patient has a much better chance of surviving, right, So, 0:42:37.560 --> 0:42:41.000 very high stakes and fundamentally, you know, if you think 0:42:41.040 --> 0:42:44.600 about what AI is today, people generally mean machine learning 0:42:44.640 --> 0:42:47.000 when they say that, right, as you know, and what 0:42:47.160 --> 0:42:50.799 machine learning is really good at doing is taking a 0:42:50.840 --> 0:42:55.080 lot of data and matching it to patterns, right saying, Oh, 0:42:55.120 --> 0:42:57.960 when you have all of this set of data like this, 0:42:58.480 --> 0:43:01.520 you tend to get this kind of out, which really 0:43:01.719 --> 0:43:04.360 is what a medical diagnosis is, right, Like, that's what 0:43:04.480 --> 0:43:06.800 a doctor is doing when they look at a patient 0:43:07.040 --> 0:43:12.320 who has some set of symptoms, age, everything, and they say, oh, 0:43:12.440 --> 0:43:15.760 this person might have sepsis. Let's do a test to see. 0:43:16.200 --> 0:43:19.720 And so she built this system and it basically works. 0:43:19.760 --> 0:43:23.439 They did some trials. But a really interesting thing she said, 0:43:23.480 --> 0:43:25.439 and it goes back, Jonathan, something you were talking about 0:43:25.480 --> 0:43:28.319 earlier in the conversation, is like she realized getting the 0:43:28.480 --> 0:43:31.120 AI to work. You know, it's not one hundred percent, 0:43:31.120 --> 0:43:35.120 but to usefully flag that a patient might have sepsis 0:43:35.280 --> 0:43:38.160 essentially what it does is like maybe half the problem, 0:43:38.200 --> 0:43:41.480 maybe not even What's really hard is getting super busy 0:43:41.520 --> 0:43:44.120 doctors who are getting a million alerts all the time 0:43:44.360 --> 0:43:46.680 to believe that this alert is worth paying attention to. 0:43:47.200 --> 0:43:49.319 And like you were talking about UI, she was like, 0:43:49.360 --> 0:43:51.480 it's totally a UI problem. Like the math was the 0:43:51.520 --> 0:43:54.000 easy part, Like you know, getting it so that instead 0:43:54.000 --> 0:43:57.600 of doctors having to spend one minute when this alert 0:43:57.600 --> 0:44:00.400 comes up, they can spend three seconds. Like that was 0:44:00.400 --> 0:44:03.600 actually a huge breakthrough for like more than the AI model. 0:44:03.719 --> 0:44:06.120 So like that's an example of an episode where there's 0:44:06.160 --> 0:44:09.520 like a cool aipiece, big stakes, but also like this 0:44:09.719 --> 0:44:13.880 interesting human UI kind of messy humanity piece. 0:44:14.239 --> 0:44:18.560 Yeah. I love talking with folks who who really tackle 0:44:18.719 --> 0:44:22.600 those sorts of challenges. I remember chatting with some roboticists 0:44:22.600 --> 0:44:27.839 who were focused not on the robotics side necessarily, not 0:44:27.880 --> 0:44:31.840 on how the robot actually functioned, but rather how to 0:44:31.920 --> 0:44:35.160 design the robots so that they could interact within a 0:44:35.239 --> 0:44:38.560 human environment in a way that did not disrupt that 0:44:38.719 --> 0:44:41.840 environment at all. And it turns out that's a really 0:44:42.000 --> 0:44:46.120 tough challenge right, like creating a robot that can navigate 0:44:46.160 --> 0:44:49.640 through a human environment still do useful things. So you 0:44:49.719 --> 0:44:53.160 have to design the robot so it can go through 0:44:53.160 --> 0:44:56.080 an environment that we have designed to make sense to us, 0:44:56.360 --> 0:44:59.600 which doesn't necessarily make sense to a robot, but then 0:44:59.640 --> 0:45:01.759 to all. So do it in a way where people 0:45:01.800 --> 0:45:05.160 aren't just stopping everything they're doing to watch the robot 0:45:05.480 --> 0:45:08.360 bump into a wall fourteen times before it finds the doorway. 0:45:08.600 --> 0:45:12.919 So yeah, I think that those conversations can be really 0:45:12.960 --> 0:45:16.359 fascinating because it does open up your eyes to other 0:45:16.560 --> 0:45:22.239 issues within technology that don't necessarily relate directly to how 0:45:22.280 --> 0:45:25.839 the tech functions, but rather how do we interact with that, 0:45:26.120 --> 0:45:29.520 what happens when you have the intersection of human experience 0:45:29.920 --> 0:45:34.160 and technology. Those are really really great. We need to 0:45:34.200 --> 0:45:37.000 take one more break to thank our sponsors, but we'll 0:45:37.000 --> 0:45:40.920 be back with more conversation about communicating technology to the 0:45:40.960 --> 0:45:53.600 general public. So another episode, I just wanted to call out. 0:45:53.640 --> 0:45:55.319 We don't have to talk about it, really, but I 0:45:55.360 --> 0:45:57.759 wanted to call out because you spoke with someone I 0:45:57.800 --> 0:46:00.600 had spoken with as well on a different show. Casner, 0:46:00.920 --> 0:46:03.880 the founder of a Panoai, which is a company that 0:46:04.080 --> 0:46:09.040 uses cameras that are co located, typically on cellular towers, 0:46:09.280 --> 0:46:13.680 to monitor for forest fires in remote places, and then 0:46:13.880 --> 0:46:16.880 it's using AI to look for signs of forest fires, 0:46:16.880 --> 0:46:20.239 which then it flags anything that it suspects as a 0:46:20.239 --> 0:46:23.880 forest fire. A human reviews the footage, so it's not 0:46:24.520 --> 0:46:28.560 just relying upon AI, and if the human determines, oh 0:46:28.560 --> 0:46:30.759 my gosh, yes, this does look like the beginnings of 0:46:30.800 --> 0:46:33.480 a forest fire, they can then send an alert to 0:46:34.120 --> 0:46:37.239 the authorities that would be responsible to respond to that 0:46:37.400 --> 0:46:41.839 and potentially cut off disasters before they could happen. When 0:46:41.880 --> 0:46:44.240 I spoke with her, it was at a time where 0:46:44.280 --> 0:46:48.280 there were the infamous Canadian forest fires that were really 0:46:48.440 --> 0:46:53.160 ravaging Canada, and so it was very clear that this 0:46:53.320 --> 0:46:58.279 sort of application of artificial intelligence had a potentially like 0:46:58.680 --> 0:47:03.080 a really beneficial implementation where it could save property in 0:47:03.160 --> 0:47:06.879 people and all sorts of benefits beyond that. You think 0:47:06.880 --> 0:47:09.719 about just even just cutting back the amount of air 0:47:09.760 --> 0:47:12.960 pollution that affected all of the Northeast. You know, all 0:47:12.960 --> 0:47:16.320 those folks who had to breathe smoky air for months 0:47:16.400 --> 0:47:20.440 because of this, Like you start again, I always talk 0:47:20.480 --> 0:47:22.439 about the ripple effect. You always want to look at 0:47:22.480 --> 0:47:26.360 how this is rippling outward because you start to realize, oh, 0:47:26.800 --> 0:47:31.240 this has even greater benefit than just the the ground 0:47:31.320 --> 0:47:34.239 zero point, right, It has all these other things that 0:47:34.320 --> 0:47:36.799 will end up benefiting people, most of which you won't 0:47:36.800 --> 0:47:40.480 even realize because you have prevented the bad thing so 0:47:40.520 --> 0:47:43.239 you don't experience the bad thing. And so I just 0:47:43.280 --> 0:47:45.040 wanted to call that out for listeners who might be 0:47:45.040 --> 0:47:47.719 looking to see where to start off, because you've got 0:47:47.800 --> 0:47:48.800 quite a few episodes. 0:47:48.920 --> 0:47:51.080 She's really interesting and you know, one of the things 0:47:51.080 --> 0:47:54.480 that was interesting to me about Sonya, the person who 0:47:54.480 --> 0:47:57.799 started this company, was she has this big idea that 0:47:58.000 --> 0:48:01.399 actually goes beyond wildfires. Is that's what they're doing now, 0:48:01.440 --> 0:48:05.200 that's their business now. But her big dream is is 0:48:05.239 --> 0:48:10.080 about data and adapting to climate change. Basically, right, there 0:48:10.080 --> 0:48:13.200 are more wildfires because of climate change. But she's like, look, 0:48:13.200 --> 0:48:15.480 we're going to be spending trillions of dollars over the 0:48:15.520 --> 0:48:18.920 next decades to mitigate the effects of climate change to 0:48:19.040 --> 0:48:21.440 you know, deal with seawater rise and flooding, and like, 0:48:21.760 --> 0:48:24.080 are you know flood maps are one hundred years old, 0:48:24.200 --> 0:48:27.840 and so if we can in different domains bring data 0:48:27.880 --> 0:48:30.840 to bear on like where should we prioritize the money. 0:48:30.840 --> 0:48:32.920 Where are there going to be more floods if we 0:48:32.960 --> 0:48:36.719 can bring technology to bear? And in her case, what 0:48:36.760 --> 0:48:38.920 that really means is data. Right, Like the sort of 0:48:38.960 --> 0:48:41.720 substrate of AI. The thing AI needs to be clever 0:48:41.960 --> 0:48:43.920 is a lot of data. If we can bring data 0:48:43.960 --> 0:48:47.400 to that, it'll just work better. For every million dollars 0:48:47.400 --> 0:48:50.000 we spend, for every billion dollars we spend, if we 0:48:50.080 --> 0:48:53.160 can be more smart about it, we will get better results. 0:48:53.480 --> 0:48:57.399 Right. It's it's like the difference between being proactive and reactive, right, 0:48:57.680 --> 0:49:00.640 being able to being able to plan for something and 0:49:01.480 --> 0:49:04.880 minimize its impact, as opposed to, Oh, now we have 0:49:04.920 --> 0:49:07.879 to clean up because this catastrophic event has happened, and 0:49:07.920 --> 0:49:10.759 how do we deal with that? And I think when 0:49:10.760 --> 0:49:13.440 we look back at some of those catastrophic events that 0:49:13.480 --> 0:49:16.680 have happened in our lifetimes, you can really see the 0:49:16.760 --> 0:49:22.000 benefit of mitigation versus reaction and cleaning up, you know, 0:49:22.080 --> 0:49:26.560 the ability to save lives and prevent damage. It's tremendous. 0:49:26.760 --> 0:49:32.200 So certainly there are plenty of artificial intelligence applications that 0:49:32.239 --> 0:49:36.160 would be incredibly helpful when put to the proper use. 0:49:36.680 --> 0:49:39.680 So again, I think if there's any lesson to take 0:49:39.719 --> 0:49:43.040 home with this conversation is use that critical thinking, try 0:49:43.200 --> 0:49:45.279 not to be reductive. I know that I can get 0:49:45.520 --> 0:49:49.160 really cynical about artificial intelligence, but again it's mostly because 0:49:49.200 --> 0:49:53.160 of the marketing language around it rather than the technology itself. 0:49:53.760 --> 0:49:53.960 Yeah. 0:49:54.000 --> 0:49:56.200 I think it's also because, like I see a lot 0:49:56.200 --> 0:50:00.279 of similarities in the AI evangelists that I saw with 0:50:00.520 --> 0:50:04.759 NFT evangelists. And that's low We all know how that. 0:50:05.480 --> 0:50:10.200 Yeah, you know, I think AI has more legs than NFTs. 0:50:10.360 --> 0:50:12.160 I feel like I'm not going out on a limb 0:50:12.160 --> 0:50:12.680 to say that. 0:50:12.719 --> 0:50:16.680 I certainly think it has more potential beneficial uses than 0:50:16.800 --> 0:50:20.399 n FTS. I think NFTs probably have some benefits too, 0:50:20.480 --> 0:50:23.040 but the problem is that no one was focusing on 0:50:23.080 --> 0:50:24.960 those when they were going crazy about them. 0:50:25.480 --> 0:50:27.600 I mean, one of the interesting things to me, when 0:50:28.080 --> 0:50:30.320 you know, when you think about what should we worry 0:50:30.360 --> 0:50:33.399 about with AI, there's sort of these two there's sort 0:50:33.400 --> 0:50:35.600 of like a Barbelle where like the thing you hear 0:50:35.600 --> 0:50:38.120 about most is it's going to take all our jobs, 0:50:38.160 --> 0:50:39.960 where a robot's going to kill us, all right, that's 0:50:40.000 --> 0:50:43.160 the like amazing there's an interesting other end of the 0:50:43.160 --> 0:50:45.200 spectrum that you know, some of the people have talked 0:50:45.239 --> 0:50:46.719 to on the show have talked about which is the 0:50:46.800 --> 0:50:50.440 risk of people over relying on AI, right, People worry that, 0:50:50.600 --> 0:50:54.640 you know, critical decisions are going to be made based 0:50:54.760 --> 0:50:57.880 on AI outputs that are not that good, that are 0:50:57.880 --> 0:51:00.840 not that robust, that are not that reliable. And you know, 0:51:00.880 --> 0:51:02.640 one of the people I talk to like runs a 0:51:02.680 --> 0:51:07.880 company basically like to stress test AI to catch eyes mistakes, 0:51:07.960 --> 0:51:11.000 and he talked about just like really dumb mistakes that 0:51:11.040 --> 0:51:12.799 he sees all the time, you know. Where he gave 0:51:12.840 --> 0:51:15.399 the example of like on a life insurance application if 0:51:15.400 --> 0:51:18.480 someone puts their year of birth instead of their age, right, 0:51:18.520 --> 0:51:21.640 so you put whatever, nineteen eighty four instead of forty, 0:51:22.160 --> 0:51:24.879 and I will actually think the person is one nine 0:51:24.960 --> 0:51:27.600 hundred and eighty four years old and will want to 0:51:27.680 --> 0:51:29.919 charge them a lot for their life insurance because boy, 0:51:29.920 --> 0:51:32.239 if you're that old, you're gonna have a lot of 0:51:32.239 --> 0:51:34.080 health risks. And I said to him, like, is it 0:51:34.239 --> 0:51:36.839 really that dumb, Like is it really or you being 0:51:36.920 --> 0:51:39.520 you know, is this hyperbole? He said, no, it's really 0:51:39.600 --> 0:51:43.080 that dumb. And so that is an interesting side of 0:51:43.080 --> 0:51:45.439 it to me, right, Like, oh, there's also a risk, 0:51:45.560 --> 0:51:47.640 there's a risk from AI being too smart. There's also 0:51:47.680 --> 0:51:50.479 a risk from a being not smart enough if people 0:51:50.520 --> 0:51:51.760 are over reliant. 0:51:51.400 --> 0:51:53.920 On it, which you would hope that people wouldn't fall 0:51:53.960 --> 0:51:56.319 into that trap. But at the same time, you just 0:51:56.400 --> 0:51:59.520 look back over the history of technology and how would 0:51:59.520 --> 0:52:04.960 we've had technology that helps remove certain tasks that we 0:52:05.080 --> 0:52:08.440 just let them go. So, for example, I can probably 0:52:08.560 --> 0:52:13.160 rattle off maybe half a dozen phone numbers of people 0:52:13.200 --> 0:52:15.840 that I know would love, but all the rest are 0:52:15.880 --> 0:52:19.440 just buried in my phone contacts because I don't need 0:52:19.440 --> 0:52:21.799 to have them stored in my own brain. I have 0:52:21.960 --> 0:52:23.680 offloaded that to technology. 0:52:24.040 --> 0:52:26.600 Half does it is a lot? Are those from twenty 0:52:26.640 --> 0:52:28.680 years ago? I don't. I haven't member a city, but 0:52:28.840 --> 0:52:30.399 he's phone number in a long time. 0:52:30.480 --> 0:52:32.560 My parents haven't changed their phone numbers since I was 0:52:32.600 --> 0:52:36.440 a child, so that one I remember, honestly, if I'm 0:52:36.440 --> 0:52:39.319 being really honest, it's probably more like three, but I. 0:52:39.400 --> 0:52:41.040 Might be down to two at this point when my 0:52:41.120 --> 0:52:43.319 mom got rid of her landline a few years ago. 0:52:43.920 --> 0:52:45.960 I definitely don't know my mom's cell number. 0:52:46.320 --> 0:52:48.799 Yeah, I know my parents' landline number because they still 0:52:48.840 --> 0:52:51.879 have it. I couldn't tell you they're cell numbers either, 0:52:52.239 --> 0:52:54.799 but that's kind of a simple example, and you know, 0:52:54.960 --> 0:52:57.239 obviously it's going to be a lot more complicated when 0:52:57.239 --> 0:53:01.320 you're talking about offloading, you know, potentially decisions to AI. 0:53:01.840 --> 0:53:03.960 But I think the argument I can make is that 0:53:04.000 --> 0:53:08.319 there's precedent, So I think that concern is well warranted, right. 0:53:08.400 --> 0:53:12.360 I think it also gets back to that concern about 0:53:12.680 --> 0:53:15.640 autonomous cars. Everyone worries that the autonomous car they're getting 0:53:15.640 --> 0:53:18.160 into is the one that's being driven by a crazy robot. 0:53:19.160 --> 0:53:21.359 So it's odd also to think of a world where 0:53:21.360 --> 0:53:23.879 people might be nervous to get into an autonomous car, 0:53:23.920 --> 0:53:27.040 but they might be willing to have an AI complete 0:53:27.080 --> 0:53:31.200 their taxes for them. For example. It's a weird world 0:53:31.239 --> 0:53:31.680 we live in. 0:53:32.080 --> 0:53:34.279 Hey, I do in taxes? Is interesting? Are you do 0:53:34.320 --> 0:53:36.960 you have an AI accountant? I mean, I like my accountant, but. 0:53:37.920 --> 0:53:41.759 My account's pretty good and I'm ninety seven percent sure 0:53:41.800 --> 0:53:46.400 she's human, So I think I'm in the clear on 0:53:46.600 --> 0:53:49.839 that one. But I could easily see that being a thing, 0:53:49.960 --> 0:53:52.240 especially for something like the United States, where the tax 0:53:52.280 --> 0:53:55.680 code gets complicated enough where people like you and I 0:53:55.760 --> 0:53:58.319 we feel the need to go out and reach out 0:53:58.360 --> 0:54:01.200 to a professional because handling it yourself as daunting. 0:54:01.400 --> 0:54:04.960 You could imagine like a happy story is like the 0:54:05.040 --> 0:54:07.480 AI does a lot of the work. An accountant can 0:54:07.480 --> 0:54:09.960 have more clients and charge each of them less, and 0:54:10.040 --> 0:54:12.239 like go over the work of the AI. Right, Like 0:54:12.280 --> 0:54:14.480 there's yes, and this it's sort of mundane, right. The 0:54:14.520 --> 0:54:17.080 reason people don't talk about outcomes like that is because 0:54:17.160 --> 0:54:20.640 it's boring. But there are a lot of boring incremental gains. 0:54:20.680 --> 0:54:22.960 If I could pay my accountant half as much and 0:54:22.960 --> 0:54:25.680 my accountant could have twice as many clients and do work, 0:54:25.719 --> 0:54:27.799 that's maybe a little better or at least as good. 0:54:28.320 --> 0:54:30.600 Everybody wins. I mean, I suppose at the margin there's 0:54:30.760 --> 0:54:33.640 need for fewer accountants in that world, but like that's 0:54:33.680 --> 0:54:35.719 okay with me, right, Like those people who would have 0:54:35.719 --> 0:54:38.040 been accountants can go and like, you know, work on 0:54:38.120 --> 0:54:39.400 AI healthcare or something. 0:54:40.080 --> 0:54:42.719 Yeah. I like the people who argue that instead of 0:54:42.719 --> 0:54:46.359 calling it artificial intelligence, maybe call it augmented intelligence, where 0:54:46.360 --> 0:54:50.480 the goal is to augment our abilities to get things done. 0:54:51.120 --> 0:54:53.200 And I think it would be a lot easier to 0:54:53.239 --> 0:54:56.400 do that if we heard fewer stories like a CEO 0:54:56.560 --> 0:55:00.440 suggesting that eight thousand unfilled jobs will ultimately be filled 0:55:00.440 --> 0:55:03.360 by AI and not humans. If we heard fewer stories 0:55:03.440 --> 0:55:07.640 like that and more stories about no, we we implemented 0:55:07.680 --> 0:55:13.399 this so that people could respond to customer concerns at 0:55:13.440 --> 0:55:17.320 a rate that's five times faster than before, which means 0:55:17.680 --> 0:55:21.160 they can resolve your issue and you're you're spending less 0:55:21.160 --> 0:55:24.320 time frustrated and sitting on hold. Like I think that's 0:55:25.120 --> 0:55:27.399 the direction that everyone wants it to go, and they're 0:55:27.400 --> 0:55:29.799 just worried that's going to go in the direction of, Hey, 0:55:30.680 --> 0:55:33.759 those coworkers you used to like they're all replaced by 0:55:34.320 --> 0:55:37.879 algorithms now, Like that's that's where we need to really go. 0:55:38.000 --> 0:55:43.080 Yes, I mean, technological unemployment is complicated, right, Like people 0:55:43.120 --> 0:55:46.240 have been certainly afraid of it for hundreds of years. 0:55:46.280 --> 0:55:49.120 Now today let's talk about some let's talk about the 0:55:49.239 --> 0:55:52.640 Dutch weavers high yeah right, I mean, you know, unemployment 0:55:52.680 --> 0:55:55.760 is below four percent today, wages are going up. People 0:55:56.000 --> 0:55:59.000 get angry when you point that out. But it's true, 0:55:59.600 --> 0:56:03.280 and it's possible that AI will be bad for workers, 0:56:03.840 --> 0:56:06.880 but we don't know yet, Like that's one like I 0:56:07.000 --> 0:56:09.560 just don't know, and I don't think anybody knows the 0:56:09.600 --> 0:56:10.319 answer to that one. 0:56:10.400 --> 0:56:15.520 Yeah, yeah, And then this lays the scariness. Well, Jacob, 0:56:15.719 --> 0:56:18.640 thank you so much for joining the show. This has 0:56:18.680 --> 0:56:22.040 been a really fun conversation. I've really enjoyed it. I'm 0:56:22.040 --> 0:56:25.520 sure my listeners have too. And just to remind everybody, 0:56:25.680 --> 0:56:28.600 your podcast is What's Your Problem. You have these kinds 0:56:28.600 --> 0:56:32.319 of conversations with decision makers and the people who are 0:56:32.360 --> 0:56:35.919 actually creating the systems we've been talking about, and who 0:56:35.960 --> 0:56:40.839 are actively tackling these questions and determining how to address them. 0:56:41.360 --> 0:56:45.359 So I highly recommend to my listeners you check out 0:56:45.400 --> 0:56:49.080 What's Your Problem. You've got so many different episodes, I'm 0:56:49.120 --> 0:56:51.600 sure like there's going to be one on there that's 0:56:51.640 --> 0:56:53.799 going to speak to every single person who listens to 0:56:53.840 --> 0:56:54.279 my show. 0:56:54.800 --> 0:56:56.920 Thank you so much. That's such a kind generous thing 0:56:57.000 --> 0:56:59.279 to say, and thanks for having me. It was great. 0:57:00.640 --> 0:57:02.560 I hope you all enjoyed this conversation I had with 0:57:02.640 --> 0:57:04.680 Jacob Goldstein. It was a pleasure having him on the show. 0:57:05.080 --> 0:57:07.160 I know this was a long one. We literally could 0:57:07.200 --> 0:57:11.640 have gone another hour easy, so I had to use 0:57:11.680 --> 0:57:14.560 some restraint there. I hope all of you out there 0:57:14.640 --> 0:57:16.919 are well. I'm looking forward to having a lot more 0:57:17.000 --> 0:57:20.200 interviews in the future. In fact, I've got a couple 0:57:20.240 --> 0:57:22.760 that I'm working on right now to kind of line up, 0:57:23.040 --> 0:57:26.080 So that's really exciting for me. I love having another 0:57:26.120 --> 0:57:29.240 point of view come into the conversation. I hope you 0:57:29.280 --> 0:57:32.680 do too, and I will talk to you again really soon. 0:57:38.800 --> 0:57:43.480 Tech Stuff is an iHeartRadio production. For more podcasts from iHeartRadio, 0:57:43.800 --> 0:57:47.520 visit the iHeartRadio app, Apple Podcasts, or wherever you listen 0:57:47.560 --> 0:57:48.600 to your favorite shows.