WEBVTT - The Story: How to be a (Rational!) Techno-Enthusiast w/ Nicholas Thompson 0:00:00.200 --> 0:00:02.400 Thanks for tuni in to techt Stuff. If you don't 0:00:02.400 --> 0:00:05.440 recognize my voice, my name is Oz Valoshian and I'm 0:00:05.440 --> 0:00:08.760 here because the inimitable Jonathan Strickland has passed the baton 0:00:08.800 --> 0:00:11.760 to Cara Price and myself to host tech Stuff. The 0:00:11.800 --> 0:00:14.640 show will remain your home for all things tech, and 0:00:14.800 --> 0:00:17.520 all the old episodes will remain available in this feed. 0:00:17.960 --> 0:00:23.000 Thanks for listening. Welcome to tech Stuff. I'm oz Vaaloshian 0:00:23.239 --> 0:00:26.880 and I'm Cara Price. So it's Wednesday, and starting on 0:00:26.920 --> 0:00:30.520 today's Tech Stuff, and every Wednesday going forward, we're going 0:00:30.560 --> 0:00:32.640 to bring you an in depth conversation with one of 0:00:32.680 --> 0:00:37.920 the brightest and farthest seeing minds in all of technology. 0:00:39.560 --> 0:00:42.919 For me personally, hosting this podcast with you is kind 0:00:42.960 --> 0:00:45.400 of a dream come true. Parts I love spending time 0:00:45.440 --> 0:00:49.600 with you, but also because I love getting the opportunity 0:00:49.640 --> 0:00:53.240 to sit down with people who are in many cases 0:00:53.280 --> 0:00:56.880 building the future and asking them what they're looking at, 0:00:57.000 --> 0:00:59.360 how they're building, what they're scared of, and what they're 0:00:59.400 --> 0:01:01.880 excited about, and then bring that back. 0:01:02.240 --> 0:01:04.200 And it is my dream to have you do all 0:01:04.240 --> 0:01:05.959 of the work and to respond to it. 0:01:06.560 --> 0:01:09.760 So thank you so for our first Wednesday episode. Of 0:01:09.760 --> 0:01:12.280 tech stuff. There was no one I wanted to reach 0:01:12.319 --> 0:01:13.880 out to more than Nicholas Thompson. 0:01:14.720 --> 0:01:17.160 I really like Nick Thompson. I remember when he was 0:01:17.480 --> 0:01:18.040 at Wired. 0:01:18.280 --> 0:01:20.600 Yeah, he was the editorn chief of Wired, and he's 0:01:20.640 --> 0:01:24.000 been a long time chronicler of tech. In fact, I 0:01:24.080 --> 0:01:27.240 really like this thing he does on LinkedIn almost every 0:01:27.240 --> 0:01:30.560 single day, which is a kind of selfie video called 0:01:30.600 --> 0:01:33.360 the most Interesting Thing in Tech This Week. 0:01:33.480 --> 0:01:38.880 It's podcast series and actually we found ourselves mentioned for 0:01:38.920 --> 0:01:40.440 a very particular reason. 0:01:40.560 --> 0:01:42.800 Hosted by Karra Price and Oswalsh. 0:01:42.560 --> 0:01:46.000 And so look at one day. Back in twenty nineteen, 0:01:46.480 --> 0:01:50.280 one of Nicholas Thompson's most Interesting Things in Tech was 0:01:50.400 --> 0:01:53.400 our very own podcast that we hosted together, Sleepwalkers. 0:01:53.640 --> 0:01:56.280 But what I like about it is it's real reporting, 0:01:56.760 --> 0:02:05.160 an analysis, but there's realism about the complicated trade offs. 0:02:05.400 --> 0:02:07.720 They're both optimistic and pessimistic. 0:02:08.240 --> 0:02:10.120 Yeah, that was a very wild moment for us. I mean, 0:02:10.160 --> 0:02:12.639 you and I do like getting press hits. My mother's 0:02:12.639 --> 0:02:14.400 a publicist. I know what a big deal it is. 0:02:14.760 --> 0:02:19.080 Certainly, when Nicholas put this video up on LinkedIn, obviously, 0:02:19.160 --> 0:02:20.800 the first thing I did was to get his email 0:02:20.840 --> 0:02:22.760 address and right to him and ask him to have 0:02:22.840 --> 0:02:27.160 a coffee, which he agreed to do, and subsequently he 0:02:27.280 --> 0:02:31.919 actually a wide magazine syndicated our podcast Sleepwalkers as a column, 0:02:31.919 --> 0:02:34.200 which was just very very very cool and very exciting. 0:02:34.639 --> 0:02:36.679 And I think in some ways it is part of 0:02:36.760 --> 0:02:39.120 the reason we're we're back in the seat a few 0:02:39.200 --> 0:02:42.400 years later, because he contributed to giving us the confidence 0:02:42.639 --> 0:02:45.359 and maybe even the credibility to be hosting tech stuff today. 0:02:45.760 --> 0:02:49.680 Absolutely, I think for us to have the real deal 0:02:50.240 --> 0:02:52.520 put his stamp of approval on things, I think was 0:02:52.600 --> 0:02:55.160 very exciting for us. But you know, we can't assume 0:02:55.200 --> 0:02:58.680 that everyone knows who Nicholas Thompson is, so what does 0:02:58.720 --> 0:02:59.320 he do now? 0:02:59.760 --> 0:03:03.400 So Nicholas Thompson went on to become the CEO of 0:03:03.480 --> 0:03:06.480 the Atlantic, So yeah, he's kind of bounced around throughout 0:03:06.520 --> 0:03:09.680 a long career in journalism. And Nicholas has written about politics, 0:03:09.800 --> 0:03:13.560 about the law, and of course technology. He's been a writer, 0:03:13.760 --> 0:03:17.200 He's been an editor and an author of books. He 0:03:17.240 --> 0:03:20.640 wrote The Hawk and the Dove, Paul Knitzer, George Kennan, 0:03:20.639 --> 0:03:23.080 and the History of the Cold War. And because he 0:03:23.160 --> 0:03:26.440 never ever stops, he's writing a new book called Running 0:03:26.480 --> 0:03:29.320 for Your Life on middle age marathons and the quest 0:03:29.400 --> 0:03:30.320 for peak performance. 0:03:31.120 --> 0:03:32.960 I just think about what I do in a day, 0:03:33.400 --> 0:03:35.040 what he does in a day. But you know, I'm 0:03:35.120 --> 0:03:37.000 very excited to hear from him. I think he I 0:03:37.040 --> 0:03:40.560 think we also gravitate towards him as a person because 0:03:40.600 --> 0:03:44.440 he is incredibly multifaceted. 0:03:44.080 --> 0:03:48.080 Highly highly, highly energetic. That's disturbingly energetic. In fact, he 0:03:48.200 --> 0:03:50.840 really is just a ball of kind of optimistic energy. 0:03:51.120 --> 0:03:53.520 And we had a lot to cover together. We talked 0:03:53.520 --> 0:03:56.200 about the deal he struck with Open Ai in twenty 0:03:56.240 --> 0:04:01.440 twenty four as Atlantic CEO, which included licensing the magazine's 0:04:01.560 --> 0:04:02.840 archive to train AI. 0:04:03.480 --> 0:04:04.200 That was drama. 0:04:04.360 --> 0:04:06.960 That was drama. So I asked him about that, and 0:04:07.320 --> 0:04:09.080 there are a few other kind of big questions, but 0:04:09.120 --> 0:04:11.680 I started the conversation asking about running and how he 0:04:11.800 --> 0:04:15.080 used tech to beat his best ever marathon time went 0:04:15.120 --> 0:04:15.960 into his forties. 0:04:17.400 --> 0:04:21.359 I had to somehow convince myself at a subconscious level 0:04:21.360 --> 0:04:23.360 that I could go faster than I thought I could. 0:04:23.800 --> 0:04:25.760 And the funny thing about running, which I didn't quite 0:04:25.839 --> 0:04:31.599 understand then, is what slows you down often isn't physiological pain. 0:04:32.480 --> 0:04:36.479 It's your body creating an illusion of physiological pain. Because 0:04:36.480 --> 0:04:39.760 it's worried that you'll lose homeostasis if you continue a 0:04:39.800 --> 0:04:41.599 pace for a certain period of time. And if you 0:04:41.600 --> 0:04:43.800 can convince your mind that you can do more, well, 0:04:43.800 --> 0:04:45.200 then you can do more. But what do you use 0:04:45.240 --> 0:04:46.880 to convince your mind? Or you have to use your mind? 0:04:47.800 --> 0:04:51.320 So I started using an arm heart rate monitor, and 0:04:51.320 --> 0:04:53.680 so I actually had very accurate readings of my heart 0:04:53.720 --> 0:04:57.280 rate as opposed to the highly inaccurate readings that we 0:04:57.400 --> 0:05:01.240 normally have, and that allowed me to both sort of 0:05:01.279 --> 0:05:03.920 titrate the effort during workouts and during races, but also 0:05:03.920 --> 0:05:05.640 to have confidence right when you're running a race and 0:05:05.680 --> 0:05:07.560 you're running at a fast pace and your heart rate 0:05:07.640 --> 0:05:10.440 is oh look, my heart rate is only one thirty five, right, 0:05:10.480 --> 0:05:14.159 I'm okay, I can go harder. That is extremely useful. Now, 0:05:14.200 --> 0:05:16.560 of course I use AI. I upload everything I've eaten. 0:05:16.600 --> 0:05:18.919 I asked for nutritional advice or to get it. Oh yeah, 0:05:19.120 --> 0:05:20.440 you know. This is what I had for breakfast, This 0:05:20.480 --> 0:05:21.880 is what I had for lunch. This is the workout 0:05:21.880 --> 0:05:24.520 I ran yesterday. What would you recommend I do between 0:05:24.520 --> 0:05:25.960 now and my next workout on Friday. 0:05:26.240 --> 0:05:31.040 It's great, how much of a paradigm for human machine interaction. 0:05:31.120 --> 0:05:33.720 Do you think this kind of experience, your experience with 0:05:34.160 --> 0:05:36.440 optimization through running is huge. 0:05:36.480 --> 0:05:39.280 I mean, if you think about AI, it's very good 0:05:39.720 --> 0:05:44.919 at tasks where it's better than the best human available 0:05:45.360 --> 0:05:48.960 and the answer doesn't have to be one hundred percent accurate, 0:05:49.080 --> 0:05:53.240 where a fast answer involving all of the inputs can 0:05:53.240 --> 0:05:56.279 be ninety five percent accurate and it's good. Right, And 0:05:56.800 --> 0:06:01.320 that's the case for what should I eat for dinner tonight? Right? Right? 0:06:01.440 --> 0:06:03.720 Like even if it tells me I need a little 0:06:03.720 --> 0:06:07.120 extra protein and maybe I don't need electric protein, who cares? 0:06:07.320 --> 0:06:09.360 But it still knows a lot more about nutrition than 0:06:09.360 --> 0:06:11.880 I know about nutrition, and it can analyze the content 0:06:11.920 --> 0:06:13.520 of the foods I've eaten in a much better way 0:06:13.720 --> 0:06:16.839 and it's an extremely useful tool. 0:06:17.080 --> 0:06:20.120 Yeah, because I mean, you're not a professional runner, although 0:06:20.160 --> 0:06:21.960 you were kind of in the elite or sub a 0:06:22.000 --> 0:06:23.040 leaite category. 0:06:22.760 --> 0:06:25.240 Sub sub elite, which is better if you folks like 0:06:25.640 --> 0:06:29.960 sub elite sounds ridiculous. Elite, you know, excellent, Like I've 0:06:30.000 --> 0:06:33.960 never I've never I'm never the elite elite, well not 0:06:34.000 --> 0:06:36.960 in running PHAs maybe in other ways you are, but 0:06:36.960 --> 0:06:37.960 but elite elite. 0:06:37.760 --> 0:06:40.599 Runners are Also, I mean you've you've talked about a runner, 0:06:40.640 --> 0:06:42.960 you know who had a digital twin, I mean talk 0:06:43.000 --> 0:06:43.400 about that. 0:06:43.800 --> 0:06:49.080 Yeah, so Deslynden, who is a wonderful runner. She set 0:06:49.120 --> 0:06:51.760 the world record for women in the fifty k run, 0:06:52.480 --> 0:06:55.599 and she's a force of nature, and so she had 0:06:55.680 --> 0:06:57.200 TCS built a digital. 0:06:56.960 --> 0:06:57.680 Twin of her heart. 0:06:58.080 --> 0:07:01.800 And it's still early days to see how useful that 0:07:01.800 --> 0:07:05.039 can be. Right now, it sort of just explains in 0:07:05.360 --> 0:07:08.159 much finer detail how she recovers from a workout and 0:07:08.160 --> 0:07:11.280 how she benefits from her workout. But you can imagine 0:07:11.840 --> 0:07:14.920 if I had a digital twin of my heart, I'm 0:07:14.960 --> 0:07:17.400 sure I could optimize workouts in a way that I 0:07:17.440 --> 0:07:17.880 can't now. 0:07:18.240 --> 0:07:20.320 Well, I guess you look at F one. There's only 0:07:20.320 --> 0:07:23.000 a sudden number of hours the cars are allowed to 0:07:23.040 --> 0:07:24.800 be on the track. So that's why simulation is so 0:07:24.800 --> 0:07:27.440 important for F one. Similarly, in running, I mean, you 0:07:27.960 --> 0:07:30.360 don't want to be running way too much, right, So 0:07:30.360 --> 0:07:33.040 in a sense, this idea of simulating training allows you 0:07:33.080 --> 0:07:35.480 to do way more training than you could do, right 0:07:35.480 --> 0:07:36.040 It was a real point. 0:07:36.040 --> 0:07:38.200 Well, I mean they're also there's specific things like so 0:07:38.440 --> 0:07:40.240 right now I run ultras and I'm trying to run 0:07:40.280 --> 0:07:45.160 a fast fifty miler, and the problem is in ultra training, 0:07:45.280 --> 0:07:47.680 you never run fifty miles in a workout, and so 0:07:47.840 --> 0:07:50.200 you can't actually test your body and see whether you're 0:07:50.200 --> 0:07:52.760 gonna make yourself puke. If you take in five hundred 0:07:52.760 --> 0:07:55.200 calories an hour for five hours, and if you could 0:07:55.280 --> 0:07:57.280 do that through a digital twin, and you can say, Okay, 0:07:57.280 --> 0:07:59.400 here's how my digestive system works. Here's the rate at 0:07:59.400 --> 0:08:02.160 which I burn calories. Here's how fast I'll be running. 0:08:02.480 --> 0:08:04.720 What is the optimal number of carbohydrates that I can 0:08:04.760 --> 0:08:09.160 take in without throwing up? That would be phenomenal. 0:08:10.080 --> 0:08:13.239 Running is one of your key passions. And then there's writing. 0:08:13.800 --> 0:08:16.160 I think I read that you've put your interviews that 0:08:16.200 --> 0:08:18.800 you're doing for your book through an NLM to kind 0:08:18.840 --> 0:08:22.360 of have new connections and themes suggested or. 0:08:22.480 --> 0:08:24.560 Yeah, so this is a really interesting process. I try 0:08:24.600 --> 0:08:28.000 to use large language models in every way possible as 0:08:28.040 --> 0:08:30.640 I write this book about running, with the exception of 0:08:31.480 --> 0:08:33.520 writing any words. Not one word in the book will 0:08:33.559 --> 0:08:34.920 be written by A but I try to use it 0:08:34.920 --> 0:08:36.480 for everything else to see where it's good and where 0:08:36.520 --> 0:08:39.080 it's not good, and also to accelerate the process, because 0:08:39.120 --> 0:08:40.679 when you write a story at The Atlantic or the 0:08:40.679 --> 0:08:42.760 New Yorkery, of this team of editors behind you right 0:08:42.880 --> 0:08:45.880 helping you all the time, you're writing a book, it's 0:08:45.880 --> 0:08:49.319 a much more solo project. And so the way the 0:08:49.320 --> 0:08:51.240 book is structured is it's partly about my life. It's 0:08:51.240 --> 0:08:53.480 probably about my father, and then it's partly about different 0:08:53.520 --> 0:08:56.600 runners who I've encountered or competed with along the way. 0:08:57.120 --> 0:08:59.600 And some of them are people that I've interviewed episodically 0:08:59.600 --> 0:09:03.560 over your time period. And so one of the characters, 0:09:03.600 --> 0:09:05.600 for example, is this one Bobby gibb first would run 0:09:05.640 --> 0:09:07.480 the Boston Marathon, mother of a friend of mine in 0:09:07.559 --> 0:09:10.600 high school. I have all these interviews. Yeah, and so 0:09:10.800 --> 0:09:12.680 the most useful task is I wrote a section on 0:09:12.679 --> 0:09:14.360 her in the book and say it's three thousand words, 0:09:14.920 --> 0:09:17.800 and then I've fed all the interviews into a large 0:09:17.840 --> 0:09:20.200 language model, and I said, here's the section I've written. 0:09:20.440 --> 0:09:23.480 Here are all the interviews. Is there anything I've written 0:09:23.480 --> 0:09:25.959 that is inaccurate based on what you've said? Are there 0:09:25.960 --> 0:09:27.000 any quotes from. 0:09:26.880 --> 0:09:30.000 Her faccurate or inaccurate characterization? Both? 0:09:30.120 --> 0:09:33.680 Yeah, it's less good on factually inaccurate, But like is 0:09:33.679 --> 0:09:35.800 anything I've said kind of unfair? Which is a test 0:09:35.800 --> 0:09:38.240 you should do as a journalist anyway, But is anything 0:09:38.280 --> 0:09:40.920 I've said unfair? And are there any quotes that she's 0:09:41.360 --> 0:09:44.240 given me that are better than the quotes I've included? 0:09:44.600 --> 0:09:47.040 And in fact it said yes, you should include this, 0:09:47.080 --> 0:09:48.520 and you should include that, And then I went back 0:09:48.559 --> 0:09:50.760 and I said, okay, great Now I have also at 0:09:50.800 --> 0:09:53.719 different points I've said, you know, here's the whole manuscript, 0:09:53.800 --> 0:09:56.160 you know, with the privacy protections on, so it's not 0:09:56.200 --> 0:09:58.760 fed back in where should I add this? And it's 0:09:59.520 --> 0:10:02.680 less good at that, But there are specific narrow tasks 0:10:02.679 --> 0:10:05.920 where it's amazing. It's just like having a very smart 0:10:06.160 --> 0:10:07.440 research assistant right there. 0:10:08.120 --> 0:10:11.439 It's interesting added that little caveat with the privacy setting 0:10:11.520 --> 0:10:14.280 zone because I think you had another experience with a 0:10:14.320 --> 0:10:16.480 book you've already written that had to do an AI 0:10:16.559 --> 0:10:17.480 that was less possive. 0:10:17.559 --> 0:10:22.439 Right. Well, yeah, so there's this big debate about my job. 0:10:22.520 --> 0:10:25.280 See you of the Atlantic, how are we licensing the 0:10:25.280 --> 0:10:28.160 Atlantics data to AI models? And there is a direct 0:10:28.200 --> 0:10:32.319 process whereby they in the last few years during which 0:10:32.360 --> 0:10:34.000 they've trained their model, have come to our site have 0:10:34.040 --> 0:10:37.520 scraped it have uploaded it, and that is something that 0:10:37.559 --> 0:10:39.400 we have some control over, right, and we can say 0:10:39.440 --> 0:10:41.120 don't do that. We can say we're going to license it, 0:10:41.120 --> 0:10:42.520 we can say you can sue you for doing it. 0:10:42.600 --> 0:10:46.080 But what is so interesting is that a huge percentage 0:10:46.679 --> 0:10:49.559 of the Atlantic content in these models doesn't come from 0:10:49.600 --> 0:10:54.000 reading the Atlantic. It comes from, well, the Atlantic's website 0:10:54.120 --> 0:10:56.520 was already captured as part of this process by which 0:10:56.520 --> 0:11:00.120 somebody captured the whole open web, or someone copy and 0:11:00.200 --> 0:11:02.040 pasted an article and to read it or put it 0:11:02.080 --> 0:11:04.480 on instant paper or pot And the same thing happens 0:11:04.480 --> 0:11:08.960 with books. So my book published by McMillan, The Hawk 0:11:09.000 --> 0:11:11.520 and the Dove, you know, on Paulmitz and George kennon 0:11:11.520 --> 0:11:14.040 the history of the Cold War, I never licensed it 0:11:14.080 --> 0:11:16.560 to an AI model, but you know, it went out 0:11:16.600 --> 0:11:19.600 to libraries and then this code or you know whatever 0:11:19.640 --> 0:11:22.360 hacked the stream of all bookstually, like, there are all 0:11:22.440 --> 0:11:24.680 these data sets that include the words in my book 0:11:24.720 --> 0:11:26.920 that have been fed into all these large language models. 0:11:27.040 --> 0:11:30.120 And that feels weird having realized that, what do you do? 0:11:30.520 --> 0:11:34.360 Well, it's complicated it because there's a there's a question 0:11:34.440 --> 0:11:37.200 of whether the AI companies argue that what they've done 0:11:37.240 --> 0:11:39.720 is fair use. They just taken data and they transform it, right, 0:11:39.720 --> 0:11:41.600 because you can't write the Hawk and the Dove with 0:11:41.679 --> 0:11:43.840 one of these things. It's transformative. It's just like they 0:11:43.880 --> 0:11:47.080 went into the library and read and then not only that, 0:11:47.240 --> 0:11:49.560 because it's from a data set and they didn't know 0:11:49.600 --> 0:11:53.280 the contents, it's kind of a secondary copyright violation. It's 0:11:53.320 --> 0:11:55.360 a little bit different if they had come and taken 0:11:55.679 --> 0:11:57.800 Hawk and the Dove, photographed it and fed it in. 0:11:57.920 --> 0:11:59.600 Yeah, they found a well in the street rather thant 0:11:59.600 --> 0:12:01.360 taking out of book gives correct, right. 0:12:01.600 --> 0:12:03.280 Or they went to a thrift store and they bought 0:12:03.360 --> 0:12:05.600 a big bucket of things and there were some wallets 0:12:05.600 --> 0:12:08.520 in there, right. So you don't have a lot of options. 0:12:08.600 --> 0:12:13.959 The option that I am most supportive of is being 0:12:14.000 --> 0:12:16.160 pursued by a company called pro Rada, and what they 0:12:16.200 --> 0:12:18.880 are doing is they are building a kind of reverse 0:12:18.960 --> 0:12:22.959 AI tool that will evaluate the answer given by a 0:12:23.040 --> 0:12:26.760 large language model way the sources that went into it, 0:12:27.000 --> 0:12:30.160 and then overlay payments process. So it's a little bit 0:12:30.280 --> 0:12:33.200 like askap right, And the idea is if an answer 0:12:33.360 --> 0:12:36.920 like open AI answers a question and their answer based 0:12:36.960 --> 0:12:41.239 on the work that pro Rada has done derives from. 0:12:41.160 --> 0:12:41.960 The Hawk and the Dove. 0:12:42.640 --> 0:12:44.920 You know, one percent derives from the Hawk and the Dove, 0:12:45.080 --> 0:12:48.520 and you know open Ai makes one penny off of it, 0:12:48.760 --> 0:12:50.600 then I should be given some fraction of one percent 0:12:50.600 --> 0:12:52.600 of the one penny, right, And that's the pro Rada 0:12:52.640 --> 0:12:53.199 is trying to. 0:12:53.200 --> 0:12:54.840 Develop that as a business model. 0:12:54.840 --> 0:12:57.120 I'm on the board of Them's a full disclosure, but 0:12:57.440 --> 0:12:59.480 there are a couple of companies out there that are 0:12:59.520 --> 0:13:03.560 trying to all this compensation issue because a lot of 0:13:03.640 --> 0:13:07.880 value has been created from copyrighted materials to which the 0:13:07.880 --> 0:13:10.319 copyright holders were being given nothing. There's not been a 0:13:10.480 --> 0:13:12.600 fair exchange of value, and that's a problem you. 0:13:12.600 --> 0:13:13.040 Have to solve. 0:13:13.960 --> 0:13:16.040 So you'll now see you at the Atlantic. But you're 0:13:16.080 --> 0:13:19.080 previously editor of Wired, and I have to ask you, 0:13:19.520 --> 0:13:21.360 I mean, what's your advice, honestly to us as we 0:13:21.400 --> 0:13:23.880 start this new podcast as to how to how to 0:13:23.920 --> 0:13:26.520 approach this set of stories and problems. 0:13:26.880 --> 0:13:29.319 I'm drawn to. I call it tech enthusiasm, where I 0:13:30.120 --> 0:13:33.200 love tech right and I think two things are true. 0:13:33.200 --> 0:13:35.280 One tech is amazing and two in the long archive time, 0:13:35.320 --> 0:13:38.440 technology makes the world better for people. That said, I 0:13:38.640 --> 0:13:40.840 was never a perfect fit with the sort of the 0:13:40.880 --> 0:13:46.680 pure optimism of early wires. It was also as the 0:13:46.679 --> 0:13:50.880 the position of pure optimism kind of fit the tech 0:13:50.880 --> 0:13:53.679 industry when they were the underdogs, but once they were 0:13:53.679 --> 0:13:56.080 the dominant forces in the world, it was a less 0:13:56.240 --> 0:13:59.959 appropriate response any case, So you should choose your own 0:14:00.040 --> 0:14:02.920 you of how you come to tech, right and true. 0:14:04.640 --> 0:14:08.880 But like my view of tech is I'm constantly trying 0:14:08.880 --> 0:14:11.280 to learn about it. I'm constantly trying to understand it, 0:14:11.480 --> 0:14:14.600 and every now and then I stop and I'm like, wait, 0:14:15.280 --> 0:14:18.439 this is horrifying. But then I'm like, you know what, 0:14:18.600 --> 0:14:21.120 like I'm enthusiastic. I'm just gonna keep going. We're gonna 0:14:21.160 --> 0:14:23.280 keep looking at this stuff. Because one of the risks 0:14:23.320 --> 0:14:25.800 in AI is that you look ahead and you're like, God, 0:14:25.800 --> 0:14:29.240 this is I'm terrified of this. And then you say, 0:14:29.280 --> 0:14:30.680 you know what I'm gonna do. I'm gonna like just 0:14:31.040 --> 0:14:33.120 I'm gonna be like King Canute, I'm gonna say the 0:14:33.120 --> 0:14:36.320 AI is not happening. Like on principle, I refuse and 0:14:36.320 --> 0:14:38.080 I'm not going to use AI because I don't like 0:14:38.160 --> 0:14:40.920 what it's doing to the world. Well, that's not the answer, 0:14:41.000 --> 0:14:43.240 because it's not going to go away just because you 0:14:43.320 --> 0:14:46.800 find it scary. You're just gonna miss the moment where 0:14:46.800 --> 0:14:48.480 you can shape it in a way that maybe makes 0:14:48.520 --> 0:14:50.080 it less scary. 0:14:51.120 --> 0:14:54.160 When we come back the Atlantic's decision to partner with 0:14:54.200 --> 0:15:07.680 Open AI and how that decision was received in the newsroom, 0:15:07.800 --> 0:15:10.440 I also really enjoy your newsletter, which is another content 0:15:10.440 --> 0:15:11.360 output that we haven't had it. 0:15:11.480 --> 0:15:14.040 I didn't even mention that, Yeah, that's a fun one. 0:15:14.640 --> 0:15:18.280 In June, you picked up an essay by Leopold Aschenbrenner 0:15:18.760 --> 0:15:22.560 called Situational Awareness, which I was both, you know, quite 0:15:22.600 --> 0:15:25.520 curious about it and also slightly put off by the tone. 0:15:26.240 --> 0:15:29.600 Oh I mean, I mean, I don't think I've ever 0:15:29.640 --> 0:15:34.000 read an essay that has the ratio of insight to 0:15:34.440 --> 0:15:37.880 alienation in that essay from like it ends up like 0:15:37.960 --> 0:15:39.680 ten times as insightful as a lot that you read 0:15:39.680 --> 0:15:42.400 and like ten times as alienating. Right, And so you know, 0:15:42.520 --> 0:15:45.640 begins with this sense that you know, because the author 0:15:45.640 --> 0:15:47.680 got really high grades at Columbia, you should trust him 0:15:47.720 --> 0:15:50.000 to like see the entire future, and he sees the 0:15:50.000 --> 0:15:52.200 future nobody else does, and you're like, okay. 0:15:51.960 --> 0:15:53.320 He's also an open AI researcher. 0:15:53.320 --> 0:15:54.960 There, right, So we worked at open A left, left, 0:15:55.000 --> 0:15:58.800 open AI, and he's he is, in fact, exceptionally bright, right. 0:15:58.800 --> 0:16:01.000 And I've spent time talking to him, exceptionally fun to 0:16:01.040 --> 0:16:02.640 talk to. And when you talk to him, it's a 0:16:02.640 --> 0:16:04.400 little easier than when you read the essay. But in 0:16:04.480 --> 0:16:09.320 case the essay says, hey, everybody, wake up, this is 0:16:09.360 --> 0:16:13.200 what happens if there's exponential AI. Here's how the improvement 0:16:13.240 --> 0:16:15.480 curve works, and here's what it will be able to 0:16:15.480 --> 0:16:19.280 do soon. And now we've watched AI go from being 0:16:19.320 --> 0:16:21.880 as smart as a toddler to being as smart as 0:16:21.920 --> 0:16:23.640 a high school student to being as smart as a 0:16:23.680 --> 0:16:27.920 pH d student. Now let's just extrapolate to when it's 0:16:27.960 --> 0:16:30.640 smarter than a Nobel Prize winner, and then when it's 0:16:30.640 --> 0:16:33.680 building a model by itself that we have no control over. 0:16:33.840 --> 0:16:36.560 And maybe he's wrong. Lots of people challenge his assumptions, 0:16:36.600 --> 0:16:40.320 and the question of whether AI will scale exponentially is 0:16:40.400 --> 0:16:43.040 hotly debated. The part that I found dangerous, that I 0:16:43.080 --> 0:16:47.600 think probably contributed to a series of mistakes that we 0:16:47.640 --> 0:16:50.440 are making right now, is the view. Okay, if you 0:16:50.480 --> 0:16:55.200 play it out, then whoever controls AI will control the world, 0:16:55.320 --> 0:16:56.960 and so therefore we need to really make sure that 0:16:56.960 --> 0:16:58.600 it's controlled in the United States and not in China. 0:16:58.680 --> 0:17:01.080 So therefore we need to have a very sagonistic relationship 0:17:01.120 --> 0:17:02.840 to China. We need to make sure they can't hack 0:17:02.880 --> 0:17:04.480 in and get our systems, and we need to set 0:17:04.480 --> 0:17:06.879 our foreign policy to prevent them from getting AI. And 0:17:06.920 --> 0:17:09.359 you can see not that Leopold Ashenbrener is responsible for, 0:17:09.440 --> 0:17:12.600 you know, the Biden administration Chips Act and anti China 0:17:12.800 --> 0:17:17.560 policies and technology, but he contributed to a conversation that 0:17:17.640 --> 0:17:20.840 I think has led to a set of policies that 0:17:21.560 --> 0:17:24.639 most people think are good, very aggressive policies by the 0:17:24.720 --> 0:17:27.680 United States to try to slow down China's AA industry, 0:17:27.960 --> 0:17:29.480 but that I think are bad. 0:17:30.320 --> 0:17:32.360 You credit that situation on the lne as I say, 0:17:32.359 --> 0:17:34.119 with a real policy. 0:17:33.760 --> 0:17:38.399 Shift, Like Leopold Asherbrener was probably in high school right 0:17:38.400 --> 0:17:40.919 when Trump started like going after Walway. But if you 0:17:40.960 --> 0:17:45.600 look at the conversation this summer, why did SB ten 0:17:45.680 --> 0:17:49.080 forty seven in California, which was strict AI regulation, Why 0:17:49.240 --> 0:17:53.360 was that knocked back by God Gavin Newsom, governor of California. 0:17:54.160 --> 0:17:57.399 I think in no small part because of a fear 0:17:57.680 --> 0:18:01.960 that if we regulate the AI industry, China will get ahead, right, 0:18:02.359 --> 0:18:05.880 And I do think that the sort of China Hawk 0:18:06.320 --> 0:18:11.120 element of the AI industry contribute it to the defeat 0:18:11.200 --> 0:18:15.400 of the most systematic attempt at regulation. And I do 0:18:15.480 --> 0:18:19.320 think that Ashrabnner's essay played a small role in that. 0:18:20.880 --> 0:18:24.399 Which I guess segues to You're no longer just an editor. 0:18:24.400 --> 0:18:27.080 You're also a CEO, right, I am, yes, And I'm 0:18:27.119 --> 0:18:30.159 wondering how do you think differently about AI as an 0:18:30.240 --> 0:18:31.439 editor versus a CEO. 0:18:32.520 --> 0:18:35.560 Well, it's a CEO you have all these other hard questions. Right, 0:18:35.600 --> 0:18:37.800 So as an editor and a writer, you're just like 0:18:38.520 --> 0:18:41.400 finding things that are interesting, You're sticking your curiosity. You're 0:18:42.480 --> 0:18:46.080 as a CEO, I have to think about how it 0:18:46.080 --> 0:18:49.439 will literally change my company and prepare for that. Right, So, 0:18:50.880 --> 0:18:53.000 what will it make easier? What will it make harder? 0:18:53.400 --> 0:18:55.960 How will it change jobs in the future. If you 0:18:56.040 --> 0:18:58.320 assume that ai wei should be powerful, should you be 0:18:58.520 --> 0:19:00.920 hiring a different kind of person? Right, So you'd be 0:19:01.000 --> 0:19:04.760 hiring somebody who's more flexible about what they do as 0:19:04.760 --> 0:19:07.199 many different skills as opposed to very narrow skills. Right, 0:19:07.240 --> 0:19:10.440 So you make those decisions. That's one category of decisions. 0:19:10.480 --> 0:19:13.719 Probably the most pressing is you have to anticipate how 0:19:13.720 --> 0:19:15.639 it will change your field and then how you operate 0:19:15.680 --> 0:19:18.480 in it. So how will it change the production of media? 0:19:18.600 --> 0:19:20.440 But one of the things that's doing is it's changing 0:19:20.480 --> 0:19:23.160 how search engines work. Right, We're going from search engines 0:19:23.200 --> 0:19:26.080 to answer engines. Answer engines don't drive traffic. We get 0:19:26.080 --> 0:19:28.439 the plurality of our readers from search engines. So if 0:19:28.480 --> 0:19:30.680 search engines go away and we move to answer engines, 0:19:30.840 --> 0:19:33.199