Listen to this episode

Speaker A: Jason, can you tell me about the most striking video you have ever seen that was created by OpenAI’s Sora?

Speaker B: Oh my God, that’s such a good question.

Speaker A: Jason Kebler is a co founder of 404 Media.

Speaker B: The one that comes to mind immediately was Spongebob standing at a, like a bully pulpit, wearing a N*** uniform and giving a Hitler speech in German to like an audience of bikini bottom fish.

Speaker A: This video, Jason saw it on the very first day Sora launched. And it was not just spongebob doing this Hitler cosplay.

Speaker B: Because of the way that Sora worked, it’s like you could remix those and so I could like swipe right and left and the next video was like Cartman. The same thing. There was like Snoopy.

Speaker A: There was a video of Sam Altman being arrested for shoplifting a dog, plucking a steak out of a grocery case, and Pikachu DJing. None of this was real. Nothing on Sora was real. But with a few simple prompts, this AI tool could make stunning videos that brought your most manic dreams to life. Jason always sort of wondered, who was this tool for? That didn’t seem to matter though. Even a few months back, Sora felt unstoppable.

Speaker B: It just felt like there was a moment where like, everyone was kind of losing their minds all at the same time over Sora.

Speaker A: Now it seems like someone has come to their senses because last week OpenAI pulled the plug on Sora.

Speaker B: I’m shocked that it died so fast. I mean, because all indications from OpenAI and sort of from, I don’t know, AI boosters was like, Sora is doing great. Like, it’s the top of the App store. It’s, you know, it’s one of the fastest downloaded apps. It had like an invite system where, you know, there was so much demand that you had to be invited by another person. And so the fact that it died within a few months, and especially that it died after all this hype was surprising. At the same time, I report on this stuff and so I go on Sora every now and then and I could see like a lot of the videos I was being served didn’t have a lot of likes, they didn’t have a lot of comments. Like, it seemed a bit like a ghost town.

Speaker A: One of the weird things about OpenAI pulling the plug on this app is that just a few months back, the company announced they’d made this billion dollar deal with Disney. And that deal involved boosting Sora by allowing users to make videos with Disney characters. And those videos were even supposed to show up on Disney plus.

Speaker B: I found it to be very odd because there was this huge press release. You know, Disney CEO Bob Iger, who is now not the CEO or is outgoing, like, had this big quote in the press release saying, like, this is going to revolutionize how we do business at Disney. And it really represented this maybe, like, acceptance from Hollywood and from, you know, the most famous entertainment company in the world that, like, AI is to some extent the future of how we’re going to make things. And then so quickly it all fell apart.

Speaker A: I mean, at the time, an OpenAI executive said that this deal represented a cinematic sea change on par with the end of the silent film era. Is the fact that this is all falling apart a sign that this cinematic sea change is off?

Speaker B: I mean, I think that there’s still a lot of hope among Hollywood business people and executives that they are going to be able to use AI to drive the cost of making content down. But I think it’s going to be a lot more gradual and less disruptive than has been promised.

Speaker A: Today on the show, Sora is dead. Does that mean Hollywood is breaking up with AI? I’m Mary Harris. You’re listening to what next. Stick around. Let’s just back up a bit and explain Sora’s importance in more detail. So for listeners who’ve never picked up this app, we should explain it’s an app that’s run by OpenAI, same company behind ChatGPT. How did it work?

Speaker B: So Sora is a smartphone app that allows you to generate AI video using text prompts.

Speaker A: Like, if you had it on your phone, do you still have it on your phone? Or is it like, boop, it’s gone.

Speaker B: It was working as of yesterday when I tried it last. So Sora does still exist. It’s. It’s not clear when they’re going to shut it down.

Speaker A: Have you done it?

Speaker B: I have done it to test it. I don’t. I don’t, like, use it in any way. I use, like, when it. When it was released, I used it a few times, and I’ve also used it a few times to test what the guardrails are.

Speaker A: What have you been able to get it to do?

Speaker B: So I wrote an article about how fake videos of ICE agents arresting people at Walmart were going really viral on Facebook, and these videos were generated using Sora and then they were taken off of Sora and they were posted to Facebook and Instagram where they went viral. And so I tested whether I could make videos like this by typing in, like, Ice Agent, Raid, Walmart. And the answer was yes. I could make videos that were very similar to kind of what. What I was seeing on Facebook.

Speaker A: How long did it take? Like, were you just like, typed in some words and it was like, boop.

Speaker B: It’s like five seconds, you know, like, it’s just. It’s like one. You type one sentence, you click generate, you wait 30 seconds for it to. To generate, and then it’s done.

Speaker A: 404 also wrote about how there were these videos that were cropping up from Sora that were of women being choked. Did you ever document a use of Sora that was positive?

Speaker B: I mean, we never wrote an article that was like, look at these. The next generation making awesome Sora videos. I. I mean, I would say a lot of what I saw on Sora is like, innocuous. You know, people making memes. I mean, it’s not like everything on Sora was bad. I think it’s just that there was quite a lot of abuse of the tool. And then also, I mean, just like endless copyrighted characters. Like, there’s so much content on there that is like Peppa the Pig, South Park, American Dad. Weirdly, it’s like a lot of those three. There was a lot of Pokemon stuff at the beginning. Like, there was a lot of videos of Pikachu, who was like, pikachu is doing, like, shoplifting at a target.

Speaker A: It’s so funny because usually in the American system, which is capitalistic, you know, part of the way we control how content gets out there is people get mad when you use their trademark characters in these ways. And so that becomes a lawsuit, that becomes expensive, that ends up sort of moderating how tools like this develop and the cautiousness with which one launches a tool like this. Doesn’t seem like that happened here.

Speaker B: Yeah, that was one of the really interesting things to me because I don’t know, Disney and Nintendo especially are notoriously litigious and notoriously protective of their characters. I wrote an article 10 years ago about a fan who threw like, an unofficial party, Pokemon party at this video game conference. And Nintendo sent that person, like, a cease and desist letter and sued them over this. Like, I think it was like $5 to enter this. This party at a bar. And it used like, a clip art of Pikachu. And so, like, that’s how litigious Nintendo is. And then you have here just like, thousands of people making really messed up content of Pikachu Charizard, like, these iconic Nintendo characters. And we didn’t hear anything about Nintendo, like, doing anything about this. I think what’s happening here is one, like the big Hollywood companies and video game companies, they are wondering what their play in AI is going to be because they see AI as inevitable in some way.

Speaker A: So they’re letting things cook.

Speaker B: I think they’re letting things cook, or I think they’re thinking, like, oh, well, OpenAI is the biggest player in this. And so do we really want to pick a fight with them? Like, if we sue OpenAI? OpenAI has lawyers and will fight that case versus, like, oh, if we send a scary letter to a teenager, like, they’re just going to stop.

Speaker A: Well, I guess that brings us to December, when Disney made this deal or announced this deal with OpenAI that would have, you know, involved them supporting Sora, essentially giving their trademark characters to users to use. Which, given everything you’ve told me about how people were using Sora, I could see why it would be concerning that Disney would be like, sure, use all of our characters. I mean, I remember back when this deal was made, like, most of the conversation was, oh, we’re gonna see p*** made with a Little Mermaid. And I don’t think that that’s. I don’t think that’s a crazy expectation. If this deal had gone forward as expected.

Speaker B: Yeah, I suspect that’s one of the reasons why this never actually launched is Disney was going to give over the likenesses to 200 different characters.

Speaker A: Wow.

Speaker B: I found that to be very wild. Like, a very wild step for Disney to take. But what we have learned is that these big AI companies try to put guardrails on what people can do with the different characters. But one, they can’t imagine how creative the people who are making this content can be and how dedicated they are to, like, circumventing those guidelines. And so, yeah, you might try to say, like, oh, you can’t make p*** of Ariel from the Little Mermaid, but there’s going to be someone who figures out how to do that anyway. And then maybe that content will be downloaded, shared elsewhere and go viral in some way. And I think that that’s a nightmare probably for Disney to sort of try to litigate and figure out, like, what the guidelines are, how. How should they be written, all of that sort of thing. Like, I can. I can imagine it being very complicated. And then, I mean, the other thing is, I follow all AI generated slop and content, like, congratulations.

Speaker A: I’m sure that’s really good for your mental health.

Speaker B: It’s very good for my mental health. And Disney P*** is very popular. Like, I will just say that. And there’s, like, a lot of it. And so already there is, like, quite a lot of horrible AI Disney slop out there. And I think the idea that Disney was going to like, invest a billion dollars to subsidize the creation of even more just, like, weird AI slop, I think it. I think it really, like, ran the risk of watering down their product and watering down their characters.

Speaker A: Do we know how this deal fell apart and why? Like, who got KG first?

Speaker B: I have no idea. I mean, I suspect that OpenAI was burning money on this product that not that many people were using because the actual compute and, like, computer power needed to generate video, audio, sync it together, let people remix it. Like, all of that, it was costing them. They’d been reporting, like, $15 million a day.

Speaker A: Yikes. Yeah. I mean, that’s the other thing. Like, we talked about how the content, like, a lot of it was benign on the app, but it was wasteful. Like, that’s something that galls me, which makes me feel like, elderly. But it was just. It seemed, like, so wasteful to me to, like, create this app where you could create free videos that were maybe silly. And that’s funny, but, like, involves so much money and energy just being burned up.

Speaker B: Yeah. And just for no one to see or look at it, you know?

Speaker A: Right.

Speaker B: Like, and so I think that. I think that’s likely. What happened is that OpenAI saw that after the initial surge of interest in Sora, there was not, like, retention, really. People weren’t sticking around. People weren’t, like, using it in the way to revolutionize social media or whatever they imagined when they launched it. And meanwhile, they’re burning all this money. They’re using, you know, a lot of their compute that is somewhat limited, like, in their data centers and things like that. And I think that they were like, well, maybe we can just reallocate all of this computing power to other things.

Speaker A: Yeah, I mean, the decision to pull the plug on Sora, I feel like, has implications both for the artificial intelligence business and for Hollywood. And I wonder if we can focus on the AI business first. Like, I mean, OpenAI has been preparing for an initial public offering. How much does that factor into this decision about Sora and setting it to the side?

Speaker B: Yeah, I think there’s a few things happening on the AI side of this equation. I think. One, there has been a new obsession over the last few months on AI agents and AI for coding, and a lot of that hype and a lot of that use has come for Anthropic and Claude, which is OpenAI’s biggest competitor. Like, there’s a lot of people who have switched from OpenAI products to anthropic products.

Speaker A: So they’re feeling the competitive pressure.

Speaker B: I think they’re feeling the competitive pressure. And I think that OpenAI has been like, the leader in the consumer space for a really long time. Like, when you think of, like, oh, AI, you think chatgpt.

Speaker A: And increasingly I think people are like, claude, is there a wider implication for AI more generally here? And I ask that because Nitish Pahwah, who writes for us over here at Slate, the first line in his article about what was going on with Sora was the AI bubble might finally be on the verge of popping. And that’s a bigger implication than what we’ve been talking about. Not just like, this one company repositioning itself within the world of AI, but like AI more generally feeling a squeeze. Do you buy that?

Speaker B: Yeah, I think it’s hard to say. I would say that this is the first, like, massive failure of a really hyped AI product from OpenAI, which is, of course, like, the darling of this industry. And so I think that a lot of people have been waiting for the other shoe to drop. We know that these companies are spending so much money. We know that they are burning cash, they’re taking so much investment, they’re not profitable. And there’s a thought like, how long can this go on before some sort of shoe drops?

Speaker A: We’ll be back after a quick break. Let’s talk about the implications of what happened with Sora for Hollywood. Does the dissolution of this deal mean that artificial intelligence is losing its shine for creative industries, for movie making, for, you know, people like Disney, essentially?

Speaker B: Yeah. I think that there is still quite a lot of hype around AI for Hollywood and, like, what it will do to creative industries. And there’s a lot of fear among writers, among actors, among editors, composers, people who work on movies and TV shows. But there’s been such a level of hype about how this is going to hollow out Hollywood, how Hollywood is not going to be able to compete with cheap AI generated content. And I think that has not come to pass. And it’s not clear if it ever will come to pass. I think that people are maybe happy to consume AI slop on free platforms while they’re, like, scrolling Instagram, but it’s not clear that they want to, like, pay to go to an AI generated movie or pay to watch an AI generated TV show on A streaming service. I think that Hollywood needs to and to some extent will lean in on the fact that they have very talented human beings creating TV shows and movies that can’t be replicated by a machine. The other thing I’ll say is that I’ve been watching AI generated movies for a long time and talking to people in the AI movie industry. Like, there’s an entire class of people who are working on these tools who are like, we are making AI generated movies. They’re getting better. They’re getting, you know, it looks bad today, but check back in a year.

Speaker A: Are you buying it?

Speaker B: Well, they say check back in a year. They’ve been saying check back in a year for like a decade. I mean, I think that there’s still a lot of hope among Hollywood business people and executives that they are going to be able to use AI to drive the cost of making content down. That has been a hope. Like, that was a big part of the writer strike negotiations and things like this, like using AI in the writers room. And obviously it’s been a big fear of Hollywood creatives. Like, are they going to try to replace us with AI? And I think that Hollywood really. Well, the business people want to do this, but the attempts to do so have been really lacking so far. Like, I live in LA and I went to an AI film festival about a year ago for tcl, which is this Chinese television manufacturing giant. They’re like one of the biggest TV companies in the world and they have this free streaming service. And they were, they created a movie studio that was going to be all AI generated films. And they were showing these films and they were saying, like, we’re going to put these on TCL plus, which is our streaming service, and this is going to revolutionize how movies are made. Well, they made five shorts, the five shorts that they were showing at this film festival. They made one 10 minute video that was laughed at widely, like widely mocked on the Internet. And then they released one other one that has 1,000 views on YouTube. Like barely anyone watched it on YouTube. And now it seems like that entire project is dead.

Speaker A: It’s funny to me that you say this though, because you’re saying that the great AI sloppification, the threat of it is maybe not as great as we thought it was. But you’re saying it at the same time as we’re seeing, for instance, this AI generated actress Tilly Norwood release a music video which is basically an attempt to do PR for AI. I don’t know if my listeners have, have watched it but it’s very funny. Like the chorus is, AI is not the enemy. It’s the key. And I’m like, well, this is clearly still coming down the pike. You know, like, even though you’re saying it’s we’re learning, it’s maybe less of a threat than we thought, it’s still knocking on the door.

Speaker B: I mean, it’s knocking on the door. There’s a lot of people trying to make it. There’s a lot of people like, who have a lot of money invested in AI becoming popular in some way in sort of like, yeah, AI generated influencers, AI generated actresses, AI generated musicians. And it’s not going away, but I think it’s, it’s like you have people who are doing this kind of like on their own to make money on social media or to get buzz or whatever. And then you have the like kind of legacy Hollywood studios that have creative people working on them who do know how to make movies and all that. And it’s like, maybe AI will lead to some efficiencies in the way that they work, but is it going to like completely revolutionize what they’re doing? I don’t know.

Speaker A: I guess what I’m seeing, and I wonder if you’d agree with me or not, is a two tier system emerging where human content is valuable and it’s also a premium. And so it’s like the price of admission for some free platform, Facebook, Instagram, whatever is AI slop. And those spaces are filling up with this stuff. That doesn’t mean that your movie theater is going to fill up with it. But that means that human interaction, human nuance, human context is not going to be as available or as readily available to everyone. And I’m not sure people have really processed that.

Speaker B: I think that this technology is coming for like the YouTubers of the world and the influencers of the world first. And we’re already seeing that. And also just like news websites, journalists, musicians, things like that. Because discoverability is such a critical thing for like a small creator where let’s say you wanted to start a YouTube channel tomorrow and become an influencer. Like finding that initial audience is going to be so hard because there are so many, not just people doing it, but there’s so many AI channels that you have to compete against. And so building an audience is becoming really, really difficult, I think. But if you are already kind of an established player, like if you are already a Hollywood studio, I think that human touch might help you stand out more because you already have like the distribution mechanisms. You don’t necessarily need to like, go viral on social media to get people to watch your movie. And so I do think it’s going to be a two tier system where it’s like, there’s going to be really popular AI channels, they’re going to be mass produced, there’s going to be a lot of them, there’s going to be really big companies that make prestige things, and then there’s going to be human influencers and human creators who already have a big audience that are able to maintain that audience because people already know them. But I think breaking into that world is going to become very difficult.

Speaker A: So Sora is dead, but long live AI Slop.

Speaker B: Oh, AI Slop’s not going anywhere.

Speaker A: Jason, I’m really grateful for your time. Thanks for spending a bit of it with me.

Speaker B: Yeah, thank you so much.

Speaker A: Jason Kebler is the co founder of 404 Media and that’s our show. What Next is produced by Elena Schwartz, Rob Gunther, Anna Phillips and Madeline Ducharme. Paige Osborne is the senior supervising producer of what Next and what Next tbd. Mia Lobel is the executive producer of podcasts here at Slate. Ben Richmond is our senior director of podcast operations. And I’m Mary Harris. Go track me down on Blue Sky. Say hey, I’m Ary Harris. Thanks for listening.

Speaker B: Catch you back here next time, Sam.