WEBVTT - Hollywood’s Silicon Valley Obsession - Week in Tech 0:00:12.130 --> 0:00:14.360 Kyle, Julia, welcome. I want to read a quick quote 0:00:14.520 --> 0:00:18.280 from a press release from the Department of Justice. OpenAI 0:00:18.320 --> 0:00:22.880 personnel discovered a surreptitious audio recording device inserted into a 0:00:23.000 --> 0:00:25.700 USB port on the back of a television in an 0:00:25.790 --> 0:00:29.170 executive conference room inside one of their buildings in San Francisco. 0:00:29.750 --> 0:00:33.050 OpenAI determined the device, which had been concealed in toilet paper, 0:00:33.370 --> 0:00:36.670 was installed on or about September 28th and that its 0:00:36.750 --> 0:00:39.410 contents were encrypted. What's going on here? 0:00:39.870 --> 0:00:41.669 Some very bad espionage. 0:00:43.030 --> 0:00:45.920 Well, Aaron Sorkin needed some material for the next movie, 0:00:46.060 --> 0:00:46.400 I think. 0:00:46.700 --> 0:00:48.820 He should probably go to the Wall Street Journal article 0:00:48.880 --> 0:00:52.680 for better material, which is potential spies using dating apps, 0:00:52.760 --> 0:00:56.000 going to the playground with fake kids as props to 0:00:56.040 --> 0:00:57.840 pick up secrets. 0:00:58.160 --> 0:01:00.660 I don't know. Clearly something's in the air in San Francisco. 0:01:01.160 --> 0:01:02.980 Just den of spies out here. 0:01:03.420 --> 0:01:06.880 My favorite detail about this story was the OpenAI spy 0:01:07.020 --> 0:01:10.680 was also caught using OpenAI products to plan his heist 0:01:10.740 --> 0:01:14.080 and ask the internal systems which conference rooms had hiding 0:01:14.100 --> 0:01:15.060 places and stuff like that. 0:01:15.160 --> 0:01:17.830 Yeah, they can still see your wife's chats, dude. 0:01:18.690 --> 0:01:21.590 I love how OpenAI turned this announcement, by the way, 0:01:21.690 --> 0:01:24.450 into an advertisement for their own products. And we used 0:01:24.510 --> 0:01:27.230 our own technology to figure out what this was and 0:01:27.270 --> 0:01:28.610 what it was used for. Yeah. 0:01:29.450 --> 0:01:32.690 Chat, what is this USB stick on my TV? Is 0:01:32.770 --> 0:01:36.000 it listening to me? Yes, you're right to be suspicious. 0:01:42.880 --> 0:01:45.440 Welcome to Tech Stuff. I'm Oz Veloshian, and this is 0:01:45.480 --> 0:01:47.700 The Week in Tech, where I'm joined by the world's 0:01:47.740 --> 0:01:50.860 most plugged-in reporters to break down what's really happening in 0:01:50.900 --> 0:01:54.360 tech right now. Today, we're joined by Reid Arbogati, tech 0:01:54.420 --> 0:01:57.780 editor at Semaphore. And Kyle Chaker, staff writer at The 0:01:57.820 --> 0:01:58.180 New Yorker. 0:01:58.220 --> 0:01:59.390 Welcome. Great to be here. 0:01:59.560 --> 0:02:00.670 Hey, happy to be here. 0:02:01.330 --> 0:02:04.770 And a special first-time welcome to Julia Black, the tech 0:02:04.870 --> 0:02:07.450 and money correspondent at Vanity Fair. Welcome, Julia. 0:02:07.690 --> 0:02:08.410 Excited to be here. 0:02:09.130 --> 0:02:11.490 So, Julia, I'm an avid reader of A Billion to One, 0:02:11.650 --> 0:02:14.540 your newsletter at Vanity Fair. I'm curious, for those who 0:02:14.580 --> 0:02:16.740 don't know the newsletter, how would you describe it and 0:02:16.760 --> 0:02:17.239 what you cover? 0:02:17.620 --> 0:02:21.280 This is a newsletter about how money... moves through the 0:02:21.320 --> 0:02:25.109 world and changes lives. It's also about the individuals and 0:02:25.150 --> 0:02:26.970 their characters. You know, at first it was kind of, 0:02:26.990 --> 0:02:31.410 is this a tech newsletter specifically? And really it's more 0:02:31.450 --> 0:02:33.970 that it almost comes down to money than tech specifically. 0:02:34.030 --> 0:02:36.430 Of course, so much of that money is concentrated in tech. 0:02:36.970 --> 0:02:38.989 I think the kind of title is playing on a 0:02:39.030 --> 0:02:43.950 few things, a lot of billionaires involved and their impact 0:02:44.050 --> 0:02:48.489 is sort of this like billion to one ratio. Unfortunately, 0:02:48.690 --> 0:02:51.260 money is now allowing people to buy that much more 0:02:51.480 --> 0:02:52.880 power and influence than the individual. 0:02:53.320 --> 0:02:55.100 The title of your column, Billion to One, is a 0:02:55.100 --> 0:02:59.560 particularly fitting title for this week's column, which includes some 0:03:00.090 --> 0:03:04.030 movie reviews, including Artificial and The Social Reckoning. Just to 0:03:04.050 --> 0:03:06.170 kind of ground the conversation, can you give the logline 0:03:06.190 --> 0:03:07.049 of both of those movies? 0:03:07.430 --> 0:03:07.690 Sure. 0:03:08.010 --> 0:03:12.070 So Artificial, in an actually very savvy move, I think, 0:03:12.160 --> 0:03:14.160 if you watch the trailer, this is framed as sort 0:03:14.180 --> 0:03:16.320 of a Sam Altman biopic, you know, get to know 0:03:16.360 --> 0:03:20.080 the evil genius behind OpenAI, because that's the household name. 0:03:20.700 --> 0:03:21.520 Currently for that company. 0:03:21.580 --> 0:03:25.320 It's actually a film about Ilya Sitskever, who is, of course, 0:03:25.460 --> 0:03:27.100 OpenAI's former. 0:03:27.040 --> 0:03:28.980 Chief scientific officer. 0:03:29.360 --> 0:03:32.960 And he is framed as sort of a tragic hero 0:03:33.280 --> 0:03:38.380 in this origin story and eventual chaotic mess that becomes 0:03:38.870 --> 0:03:42.570 the world of OpenAI. Sam Altman is certainly a character 0:03:42.590 --> 0:03:44.410 in the film, but by no means the main character. 0:03:44.450 --> 0:03:47.990 It's actually fascinating that Andrew Garfield, who plays Sam Altman, 0:03:48.050 --> 0:03:53.090 will be in the nomination category as supporting actor, whereas Yura, 0:03:53.190 --> 0:03:58.170 who plays Ilya, will be considered for best actor. So 0:03:58.590 --> 0:04:02.050 that's that film, which I loved. On the other side 0:04:02.090 --> 0:04:05.950 of things, we have The Social Reckoning, which is, I guess, 0:04:06.150 --> 0:04:09.570 a sequel to The Social Network, which I think hopefully 0:04:09.590 --> 0:04:11.490 we could all agree is, at the very least, a 0:04:11.530 --> 0:04:14.590 pretty good film. I'm still scratching my head on why 0:04:14.660 --> 0:04:20.169 anyone thought this follow-up needed to exist um it firms. 0:04:19.830 --> 0:04:22.850 Itself sort of around i guess january 6th. 0:04:22.770 --> 0:04:26.530 Becomes the moment it's about the story of francis hogan 0:04:26.550 --> 0:04:30.330 the whistleblower at facebook and yeah it just feels kind 0:04:30.360 --> 0:04:32.810 of like it's rehashing old news at this point. 0:04:33.460 --> 0:04:36.360 Yeah i mean not having seen either of them um 0:04:36.720 --> 0:04:39.480 i mean the kind of drama of like who sam 0:04:39.560 --> 0:04:41.200 is even if he's not the main character but the 0:04:41.260 --> 0:04:45.770 kind of raging debates within open AI about safety versus, 0:04:46.050 --> 0:04:49.620 you know, commercialization and the acquisition of raw power feel 0:04:49.690 --> 0:04:53.380 like that feels like the story of today. Whereas like 0:04:53.880 --> 0:04:56.440 the story of Mark Zuckerberg a few years ago, before 0:04:56.480 --> 0:05:00.540 he became like, hence Julius Caesar, Mark Zuckerberg doesn't feel 0:05:01.020 --> 0:05:03.540 that interesting. In fact, much more captivating for me this 0:05:03.580 --> 0:05:07.420 weekend was the weird, weird, weird Mark Zuckerberg and Alexander 0:05:07.460 --> 0:05:12.080 Wang viral clips on X where Mark Zuckerberg reluctantly drinks 0:05:12.140 --> 0:05:13.800 a pint of Guinness. Anyone see these? 0:05:13.820 --> 0:05:15.400 Yeah, of course. 0:05:17.070 --> 0:05:19.330 I was just thinking about how I'd rather see Mark 0:05:19.350 --> 0:05:22.790 Zuckerberg now, like a movie about the kind of rebranding 0:05:22.830 --> 0:05:26.540 of Mark Zuckerberg and the broification and the like feeding 0:05:26.570 --> 0:05:29.140 his cows in Hawaii particular nuts. 0:05:29.120 --> 0:05:32.310 Or whatever he's doing. Like that's kind of more entertaining. 0:05:32.510 --> 0:05:35.190 I would watch a movie about Mark Zuckerberg's stylist. I 0:05:35.230 --> 0:05:37.090 don't think I would watch a movie about Mark Zuckerberg. 0:05:37.350 --> 0:05:41.219 You know, I mean, what makes art great? What makes artificial, 0:05:41.260 --> 0:05:45.320 for example, feel almost like a Greek tragedy is human stakes. 0:05:45.480 --> 0:05:50.060 And is anyone still feeling like Zuckerberg is human? I mean, 0:05:50.080 --> 0:05:53.250 I think I noted in my column, you know, like 0:05:53.710 --> 0:05:57.309 it's fascinating to watch a soul erode in real time, 0:05:57.550 --> 0:06:00.290 but Once the soul's gone, you're just kind of watching 0:06:00.330 --> 0:06:00.710 a body. 0:06:01.589 --> 0:06:04.530 How do you juxtapose the performances of Jeremy Strong and 0:06:05.070 --> 0:06:05.730 Andrew Garfield? 0:06:06.410 --> 0:06:09.500 Jeremy Strong nails it. It's very accurate and true to life, 0:06:09.800 --> 0:06:10.140 I feel. 0:06:10.260 --> 0:06:13.800 But again, because there's no pathos for this character, because 0:06:14.279 --> 0:06:18.219 there's nothing relatable there. It's someone we see on TV 0:06:18.320 --> 0:06:21.260 all the time. I don't think there's much compelling about it. 0:06:22.110 --> 0:06:24.050 And he really committed to this role. He did his 0:06:24.150 --> 0:06:25.210 usual method acting. 0:06:25.230 --> 0:06:28.360 But... for what? 0:06:29.320 --> 0:06:31.180 If you could have saved some time, it sounds like Julia. 0:06:32.260 --> 0:06:32.539 Yeah. 0:06:33.120 --> 0:06:36.430 I had Alex Stamos on my podcast and he had 0:06:36.670 --> 0:06:38.050 been to the premiere. 0:06:37.589 --> 0:06:40.169 And met- Wait, just say who Alex Stamos is, Reid. 0:06:40.250 --> 0:06:45.110 He's the former chief security officer of Facebook. He gave 0:06:45.150 --> 0:06:47.840 it two thumbs down. And his theory, I liked his theory. 0:06:47.880 --> 0:06:52.970 He's like, I think Aaron Sorkin was projecting- like Hollywood 0:06:53.050 --> 0:06:56.390 film studio execs onto Mark Zuckerberg. He's like, if anybody 0:06:56.570 --> 0:06:59.010 actually acted like that in the tech industry, no one 0:06:59.029 --> 0:07:01.130 would work for them. And I thought that was a great, 0:07:01.150 --> 0:07:02.510 I thought it was a great line. 0:07:02.890 --> 0:07:03.560 Wait, what does that mean? 0:07:03.940 --> 0:07:06.280 I haven't seen the movie, but like essentially he acts 0:07:06.320 --> 0:07:10.360 like a madman around his people. And it's like, that's 0:07:10.400 --> 0:07:13.520 how Hollywood film execs act. And it's not how tech 0:07:13.580 --> 0:07:17.680 execs act. So it's like Hollywood people sort of imagine 0:07:17.700 --> 0:07:22.030 like how would a Mark Zuckerberg act in meetings? you know, 0:07:22.090 --> 0:07:25.250 they just think of the bad men in their lives, essentially. 0:07:25.970 --> 0:07:26.650 That was his point. 0:07:26.910 --> 0:07:29.050 And you really buy that he's so much more human 0:07:29.090 --> 0:07:30.830 than that in meetings. I don't know. 0:07:31.800 --> 0:07:33.340 I mean, I don't know. This is a person who, 0:07:33.500 --> 0:07:35.000 who spent a lot of time with it, right? 0:07:35.820 --> 0:07:37.740 Well, he decided to work for him. So there's that. 0:07:38.060 --> 0:07:40.560 Look, I covered, I actually covered Facebook for the wall 0:07:40.580 --> 0:07:43.200 street journal too, and had a bunch of meetings with Zuckerberg. 0:07:43.310 --> 0:07:46.410 I mean, he's definitely, he has definitely has a little 0:07:46.490 --> 0:07:50.490 stiff personality, but I'll have to see the movie. I mean, Like, 0:07:50.810 --> 0:07:53.550 you know, behind closed doors, like people act differently than 0:07:53.590 --> 0:07:57.630 they do like in congressional hearings. Right. I mean, usually. 0:07:57.830 --> 0:08:00.150 Well, I will know that that's most of the film 0:08:00.350 --> 0:08:04.050 is him preparing for this hearing. And yeah, it's not 0:08:04.250 --> 0:08:07.230 a peak at his more human cattle farmer. 0:08:07.250 --> 0:08:14.160 Exactly. But apparently even his sipping beer is inhuman. I 0:08:14.240 --> 0:08:18.760 loved the Twitter exchange between Jessica Lesson, who's the information 0:08:19.110 --> 0:08:23.030 founder and knows Zuckerberg well. She was lamenting the fact 0:08:23.070 --> 0:08:25.570 that Jason Gay at the Wall Street Journal had written 0:08:25.610 --> 0:08:29.530 about the beer sipping incident over the weekend. And she's like, 0:08:29.590 --> 0:08:32.370 is this really what people at the Wall Street Journal 0:08:32.390 --> 0:08:35.200 should spend their time on? watching and Jason gave responding. 0:08:35.220 --> 0:08:37.360 He's like, well, usually I spend my time watching the 0:08:37.380 --> 0:08:39.420 New York jets. So yes, this is a better use 0:08:39.440 --> 0:08:39.940 of my time. 0:08:41.120 --> 0:08:43.880 It feels like we're analyzing them like politicians more and more. 0:08:44.059 --> 0:08:47.810 Like we would analyze how Obama drank a beer. Like, 0:08:48.030 --> 0:08:52.809 I don't know. And I was also just curious, Julia, 0:08:52.830 --> 0:08:56.150 you've seen them to what degree is realism, the point 0:08:56.650 --> 0:08:59.410 of these or the like accomplishment of these. 0:09:00.370 --> 0:09:04.330 Yeah, that's a good question. I know a bit about 0:09:04.370 --> 0:09:07.089 the backstory of how both films were made. And I 0:09:07.170 --> 0:09:12.310 do believe there were pretty strong attempts to accurately portray events. 0:09:12.710 --> 0:09:14.930 I'm going to flex on your behalf, Julie, because your 0:09:14.990 --> 0:09:17.470 name was on the whiteboard in the background of the 0:09:17.510 --> 0:09:18.750 artificial movie. 0:09:18.950 --> 0:09:20.790 I had to show that off. I mean, I couldn't 0:09:20.820 --> 0:09:21.220 believe it. 0:09:21.320 --> 0:09:25.120 Definitely my claim to fame and my moment in history. 0:09:26.640 --> 0:09:28.760 But I think that the reason that happened is because 0:09:28.840 --> 0:09:33.180 Luca Guadagnino, the director, really consulted with primary sources. 0:09:33.360 --> 0:09:35.200 He worked with people who worked at OpenAI. 0:09:36.070 --> 0:09:38.829 Simon rich who wrote the script uh actually at the 0:09:38.850 --> 0:09:42.590 new yorker there's this great little mini documentary um that 0:09:42.630 --> 0:09:45.670 simon participated in because he went to i believe high 0:09:45.710 --> 0:09:48.610 school with someone pretty high up at open ai so 0:09:48.630 --> 0:09:51.429 i think he came to the script with a certain 0:09:51.450 --> 0:09:55.290 amount of background knowledge and yeah these sources were on 0:09:55.390 --> 0:09:58.670 set they were consulting they were writing on the whiteboards 0:09:58.710 --> 0:10:02.550 you know like the film does get into pretty technical territory, 0:10:02.610 --> 0:10:06.450 which I wonder if that will alienate more mainstream viewers, 0:10:06.490 --> 0:10:08.689 you know, to how much does the average person really 0:10:08.710 --> 0:10:12.960 care about back propagation? I'm not sure, but I loved it. 0:10:13.600 --> 0:10:17.880 And then Roger McNamee is this former Facebook investor who 0:10:17.920 --> 0:10:20.620 became kind of an activist against the company. He is 0:10:20.700 --> 0:10:23.739 represented by a Bill Burr character in the film. I 0:10:23.760 --> 0:10:25.980 know that he consulted quite a lot with Aaron Sorkin. 0:10:26.020 --> 0:10:27.939 Well, there's a lot of interesting insight. I mean, Wendy 0:10:28.000 --> 0:10:31.480 Schmidt is obviously an Alex Gibney partner and financier who's 0:10:31.500 --> 0:10:33.940 doing the Elon doc. I feel like there's all kinds 0:10:34.000 --> 0:10:37.699 of weird like undercurrents of like different people in different, 0:10:38.000 --> 0:10:41.630 you know, financing positions on different cultural artifacts about who's 0:10:41.690 --> 0:10:44.170 up and who's down in the tech world. 0:10:44.429 --> 0:10:44.770 Yes. 0:10:45.370 --> 0:10:47.470 Just before we go to the break, like why, like, 0:10:48.360 --> 0:10:51.100 Why is this happening? Why are all these films coming 0:10:51.160 --> 0:10:52.740 out seemingly all at the same time? 0:10:53.120 --> 0:10:56.320 I mean, on behalf of Luca Guadagnino, kudos to that man. 0:10:56.400 --> 0:10:59.500 He could not have timed this any better. The fact 0:10:59.520 --> 0:11:03.310 that Artificial is coming out as this horror movie about 0:11:03.370 --> 0:11:06.010 something that people are suddenly feeling so deeply in their gut, 0:11:06.250 --> 0:11:09.870 since the Jacob Coxon revelations and all this activism coming 0:11:09.929 --> 0:11:14.420 out of Frontier Lab employees and former employees, that's just 0:11:14.440 --> 0:11:17.219 going to be terrific timing for them. I think that's 0:11:17.240 --> 0:11:20.380 a coincidence. I've heard conspiracy theories about, you know, this 0:11:20.420 --> 0:11:25.260 is all an EA-funded psy-op, but I don't believe that's true. 0:11:25.620 --> 0:11:27.470 I watched the credits to make sure there was no 0:11:27.520 --> 0:11:29.010 funding ties there. 0:11:29.950 --> 0:11:31.809 I think it's a little bit of a disservice. Like, 0:11:31.910 --> 0:11:34.530 I think these movies, like, I mean, Hollywood has to 0:11:34.550 --> 0:11:36.690 be Hollywood, right? And this stuff has to be, like, 0:11:37.090 --> 0:11:40.370 portrayed in a certain way to get people to watch it. But, like, 0:11:41.290 --> 0:11:44.160 reality is not that. And so when you're, like, a 0:11:44.220 --> 0:11:47.350 reporter and you're trying to write, like, reality. It's like, 0:11:47.790 --> 0:11:49.790 no one's like, I think people are like, well, why 0:11:49.809 --> 0:11:52.010 would I pay attention to that? Like, I've got this 0:11:52.050 --> 0:11:56.089 Hollywood version over here. That's like high drama constantly. And like, 0:11:56.530 --> 0:11:59.490 you know, these people acting like maniacs. And it's like, 0:11:59.690 --> 0:12:01.630 when you talk to people, like, it's so funny. Like 0:12:01.650 --> 0:12:04.290 when you talk to people, like the drama that happens 0:12:04.350 --> 0:12:06.349 in these, in these, there'll be like, Oh my God, 0:12:06.390 --> 0:12:09.490 that person is such a sociopath. And you're like, you're 0:12:09.530 --> 0:12:12.210 just digging like, but why like explain to me like 0:12:12.370 --> 0:12:14.739 what happened? And they're like, well, in the first meeting, 0:12:15.070 --> 0:12:18.610 they didn't know my name. And I'm like, well, it's 0:12:18.650 --> 0:12:21.050 like stuff like that. And I'm like, this is just 0:12:21.150 --> 0:12:23.920 normal stuff that happens in life. I feel like it's 0:12:24.000 --> 0:12:27.020 almost like the drama that happens in these, in these 0:12:27.100 --> 0:12:30.980 companies is like, it's not interesting enough for Hollywood. And 0:12:31.040 --> 0:12:32.900 so they have to just like sort of invent it. 0:12:33.000 --> 0:12:37.440 And it's, that's everything though. That's, that's Hollywood. These are entertainment. 0:12:37.559 --> 0:12:40.160 I feel like the problem is this is too fast. 0:12:40.660 --> 0:12:44.990 Like we're, we're transforming the news into, into like full 0:12:45.020 --> 0:12:47.809 budget feature films so quickly now that it just feels 0:12:47.870 --> 0:12:50.870 like there's no, there's no time to like process the 0:12:50.890 --> 0:12:54.650 reality as you're saying, it's just about translating it into this, 0:12:54.850 --> 0:12:57.070 this popular entertainment as quickly as you can. 0:12:57.150 --> 0:12:57.340 Yeah. 0:12:57.530 --> 0:13:00.699 Rip from the headline feature films in the theaters, like 0:13:00.760 --> 0:13:04.459 shouldn't exist. I totally agree. Like this should happen years later. 0:13:04.620 --> 0:13:06.440 Like do the rip from the headlines. I don't know. 0:13:06.480 --> 0:13:08.100 Like after school special. 0:13:08.140 --> 0:13:09.820 Julia, I think there's a strong disagree here. 0:13:10.100 --> 0:13:12.780 Yes, I think I have to politely disagree with both 0:13:12.880 --> 0:13:13.699 of your points. 0:13:13.720 --> 0:13:15.850 Don't be polite. 0:13:16.070 --> 0:13:20.449 First of all, the Facebook, the Social Reckoning film is 0:13:20.710 --> 0:13:24.170 a few years dated and it's totally dull for that reason. 0:13:24.210 --> 0:13:25.830 You know, it feels like old news. 0:13:25.890 --> 0:13:29.470 Like everyone's kind of come to terms with the divisiveness 0:13:29.510 --> 0:13:32.500 of Facebook and everybody shrugged. I mean, society's had plenty 0:13:32.540 --> 0:13:34.940 of chance to do something about it and we've decided 0:13:34.980 --> 0:13:39.170 not to for whatever reason. Whereas the open AI drama 0:13:39.220 --> 0:13:42.300 is unfolding in real time. Again, I think like the 0:13:42.360 --> 0:13:44.720 stakes are hitting people in their gut in this very 0:13:44.760 --> 0:13:48.960 emotional way right now, which to me feels the reality 0:13:49.000 --> 0:13:52.540 feels plenty dramatic. I mean, I remember being at my 0:13:52.620 --> 0:13:56.400 in-laws over Thanksgiving weekend, the week of the open AI poo. 0:13:56.800 --> 0:13:58.059 And that was just straight reporting. 0:13:58.100 --> 0:14:01.330 I was at the information and you know, I'm sitting 0:14:01.370 --> 0:14:03.490 at the Thanksgiving table and I have my laptop under 0:14:03.510 --> 0:14:04.569 the table and everyone's like, Julia, what. 0:14:04.510 --> 0:14:05.030 Are you doing? 0:14:05.410 --> 0:14:08.710 It was so dramatic living through it. And I, I 0:14:08.730 --> 0:14:10.870 think that the film reflects that sense. 0:14:10.929 --> 0:14:11.410 So I don't know. 0:14:11.650 --> 0:14:13.609 That's why I got into this job because it feels 0:14:14.070 --> 0:14:16.220 emotional and dramatic to me and exciting, but. 0:14:16.800 --> 0:14:18.140 I think the stakes are high. 0:14:18.160 --> 0:14:20.340 We're going to go to the break now, but I 0:14:20.360 --> 0:14:22.000 want to give my sort of final thought here, which 0:14:22.140 --> 0:14:24.720 I think we are all now in the open AI 0:14:24.900 --> 0:14:29.730 boardroom saga. Everyone feels like an open AI board member 0:14:29.770 --> 0:14:31.830 trying to ride the reins right now. So I think 0:14:31.850 --> 0:14:34.790 they really seem to really hit the zeitgeist. I agree 0:14:34.810 --> 0:14:37.270 with you, Julia. When we come back, why one of 0:14:37.310 --> 0:14:41.640 the buzziest and most controversial AI startups is filing for bankruptcy. 0:14:42.160 --> 0:14:59.340 Stay with us. Welcome back. Kyle, we've talked about Character.ai 0:14:59.380 --> 0:15:02.460 a few times on this show, mostly due to lawsuits 0:15:02.660 --> 0:15:05.180 around the safety of its chatbots, but now it's filed 0:15:05.200 --> 0:15:08.000 for bankruptcy. Just before we get into that, for those 0:15:08.040 --> 0:15:10.640 who don't know, what is it and what's driven this 0:15:10.990 --> 0:15:11.640 fall from grace? 0:15:12.490 --> 0:15:15.510 So this is a AI startup, like as you were saying, 0:15:15.590 --> 0:15:19.950 one of the busiest kind of emerging non-Google, non-meta, like 0:15:20.090 --> 0:15:24.510 independent products, AI companies. And it was started in 2021 0:15:24.510 --> 0:15:27.850 by some ex-employees of Google who like left their AI 0:15:28.470 --> 0:15:29.610 departments to do this. 0:15:30.350 --> 0:15:33.170 And not just any former Google, like senior Google employees. 0:15:33.200 --> 0:15:36.800 Yes, very senior, very important people. So it's like talent 0:15:36.820 --> 0:15:38.920 that went missing in a way to do this startup, 0:15:39.000 --> 0:15:44.510 which I have Basically, when you go to character.ai or 0:15:44.590 --> 0:15:47.290 went to at the beginning, what you saw was a 0:15:47.350 --> 0:15:49.950 menu of different chatbots that you could just open up 0:15:50.010 --> 0:15:54.390 and start chatting with. And they were mostly fictional characters 0:15:54.630 --> 0:15:57.660 or these kind of like fictional personality archetypes that you 0:15:57.680 --> 0:16:01.480 could chat with. And the whole presentation for character.ai was 0:16:02.020 --> 0:16:05.320 geared toward young users. It was kind of approachable. It 0:16:05.360 --> 0:16:09.280 was friendly. It was, you know, characters from Harry Potter 0:16:09.420 --> 0:16:13.030 or Marvel or whatever. But the problem was that when 0:16:13.070 --> 0:16:15.930 you started chatting with them, they had very few safeguards. 0:16:16.530 --> 0:16:18.600 So there was, you know, very quickly, there were a 0:16:18.620 --> 0:16:23.120 lot of problems with encouraging violence or the chatbots getting 0:16:23.160 --> 0:16:27.960 into erotic roleplay, let's say. And so the company faced 0:16:28.020 --> 0:16:32.900 all of these lawsuits and problems over the years. And now, 0:16:33.260 --> 0:16:37.270 I mean, at this point, their talent from character.ai was 0:16:37.350 --> 0:16:41.320 poached again, by Google in this kind of like acqui-hire 0:16:41.340 --> 0:16:44.200 AI company move that we've seen a lot lately. 0:16:44.220 --> 0:16:45.640 By Google and by Disney, right? 0:16:46.180 --> 0:16:46.420 Yes. 0:16:46.600 --> 0:16:49.620 And then now after Google, that was what, a year 0:16:49.660 --> 0:16:52.500 or two ago, and now by Disney, which wanted to 0:16:52.540 --> 0:16:55.100 license its technology and use it. But now I think 0:16:55.180 --> 0:16:58.100 it might be a little bit too toxic for those 0:16:58.140 --> 0:17:00.460 deals to keep going. And I don't know, I'm not 0:17:00.640 --> 0:17:04.070 particularly surprised. I think the postmortems from the company itself 0:17:04.119 --> 0:17:07.379 kind of said that there was no route toward profitability. 0:17:07.440 --> 0:17:11.250 They were spending much, much more on the compute and, 0:17:11.710 --> 0:17:14.010 you know, backend than they were making from their users 0:17:14.090 --> 0:17:17.770 and advertising that they put in chats. But I don't 0:17:17.790 --> 0:17:20.370 know what I, the kind of take that I wanted 0:17:20.410 --> 0:17:22.810 to get out of talking about this was, this is 0:17:22.869 --> 0:17:28.700 like a terrifying, horrible product that had no reason to exist. 0:17:29.280 --> 0:17:32.500 Like I used it myself to write about it. It 0:17:32.540 --> 0:17:36.330 was not safe. It was like, not good for any 0:17:36.369 --> 0:17:39.130 kind of young user. Like we think Instagram is toxic. 0:17:39.190 --> 0:17:43.440 This was like above and beyond. So I can't say 0:17:43.480 --> 0:17:44.869 I'm particularly sad about it. 0:17:45.260 --> 0:17:45.500 Yeah. 0:17:46.340 --> 0:17:49.359 I mean, I, I'm obviously not surprised that they, that 0:17:49.400 --> 0:17:52.140 they filed for bankruptcy. I mean, the company was basically 0:17:52.180 --> 0:17:54.680 acquired and, you know, the shell of the company gets 0:17:54.720 --> 0:17:56.950 kind of left to, for people to pick over it. 0:17:57.000 --> 0:18:00.480 But I mean, yeah, the lawsuit, they didn't have proper 0:18:00.520 --> 0:18:03.740 safety controls and they were basically getting decimated by these lawsuits. 0:18:04.369 --> 0:18:07.510 Noam Shazier, who was like the person you're talking about 0:18:07.570 --> 0:18:12.209 founded it brilliant engineer at Google. But he was long gone. 0:18:12.390 --> 0:18:14.770 And then I think the big question that was like, 0:18:15.210 --> 0:18:19.189 with all these open weight models, like, they're essentially are 0:18:19.250 --> 0:18:22.129 even fewer controls. Like this is this is the totally 0:18:22.150 --> 0:18:23.910 the Wild West. I mean, you can download this stuff 0:18:24.050 --> 0:18:27.790 on your laptop and run, you know, locally, these models 0:18:27.869 --> 0:18:31.149 and have any kind of chat that you want. So like, 0:18:31.960 --> 0:18:36.090 The problem isn't really solved, right? And I think we're 0:18:36.109 --> 0:18:37.670 going to have to grapple with that at some point 0:18:38.190 --> 0:18:38.890 as a society. 0:18:39.590 --> 0:18:41.729 I guess to me, there's two parts to this. One 0:18:41.790 --> 0:18:44.609 is like, can you afford to burn a lot of 0:18:44.690 --> 0:18:48.109 compute trying to make a consumer product? And the other 0:18:48.210 --> 0:18:50.980 is the lawsuits and safety issue. Or are they kind 0:18:51.020 --> 0:18:54.260 of two sides of the same coin? 404 Media said, quote, 0:18:54.340 --> 0:18:57.340 character AI's bankruptcy is one of the first dominoes to 0:18:57.420 --> 0:19:00.659 fall in the long-foreseen AI bubble. Do you agree with 0:19:00.700 --> 0:19:00.940 the 404 take? 0:19:01.950 --> 0:19:04.890 Yeah, I mean, I actually what you said before was 0:19:04.930 --> 0:19:08.410 really interesting. I think there's not enough compute right now 0:19:08.490 --> 0:19:11.030 for consumer AI. I think this is why like all 0:19:11.070 --> 0:19:15.899 these products are either made for enterprise or pivot toward 0:19:15.960 --> 0:19:19.679 enterprise because consumers don't want to pay for stuff. And 0:19:19.740 --> 0:19:23.280 so and this is a very expensive and so there 0:19:23.320 --> 0:19:25.740 just isn't there's there aren't really going to be a 0:19:25.790 --> 0:19:29.500 lot of consumer products. I mean, you news is, is 0:19:29.720 --> 0:19:32.940 one example. I have doubts about whether that's going to have, 0:19:33.200 --> 0:19:37.380 you know, long, long-term stickiness, uh, drink more Guinness read. 0:19:39.810 --> 0:19:41.530 We should have Guinness on this show. I mean, it 0:19:41.590 --> 0:19:45.290 is 9am my time, but I'll, take one for the team. No, 0:19:45.330 --> 0:19:48.010 I just think it's I think consumers really, it's really 0:19:48.030 --> 0:19:50.889 going to struggle until we have a lot more data centers, 0:19:50.930 --> 0:19:54.560 a lot more tokens to go around and and more infrastructure. 0:19:54.580 --> 0:19:57.720 I think the security infrastructure is like, it just hasn't 0:19:57.760 --> 0:19:58.780 even been built yet. 0:19:59.040 --> 0:20:01.740 But it's like, to your point, it's what product is 0:20:01.800 --> 0:20:05.240 worth the compute, like we're seeing that the B2B stuff 0:20:05.320 --> 0:20:09.000 is worth the compute, like super specialized, powerful stuff. What 0:20:09.040 --> 0:20:13.480 Character AI was saying was our product is teenagers chatting 0:20:13.720 --> 0:20:19.139 with fantasy characters funded by advertising and product tie-ins. 0:20:19.530 --> 0:20:24.590 You've got your maximum harm, minimum monetization. 0:20:24.650 --> 0:20:29.250 Yeah, functional purpose, I don't think so. So yeah, thinking 0:20:29.310 --> 0:20:32.210 about what is actually a functional product I think is 0:20:32.270 --> 0:20:33.109 important right now. 0:20:33.130 --> 0:20:37.949 Do you remember this... everything's a rapper conversation and like 0:20:38.010 --> 0:20:41.050 character AI was like, because probably was a rapper. Like, 0:20:41.090 --> 0:20:43.930 do you think this is like the 404 headline around 0:20:43.990 --> 0:20:46.119 like one of the first dominoes to fall in the 0:20:46.160 --> 0:20:48.700 long foreseen AI bubble? Like, do you think this is 0:20:48.760 --> 0:20:51.520 like a kind of like a signal that like, it's 0:20:51.540 --> 0:20:54.980 gonna be very hard for, especially on the consumer side, 0:20:55.060 --> 0:20:59.169 the non kind of extraordinarily heavily funded frontier labs as 0:20:59.210 --> 0:21:02.150 the product gets better and the funding disparity increases that 0:21:02.190 --> 0:21:05.690