≡ Menu

All-In Podcast Concerned But Confused about IP and AI Training

The All-In podcast (@theallinpod) hosts Chamath Palihapitiya (@chamath), Jason Calacanis (@Jason), David Sacks (@DavidSacks), and David Friedberg (@friedberg) screw up yet another IP discussion. (See All-In Podcast Concern over China and IP “Theft”.) This time they sense how copyright is going to screw up IP and want some changes, but… not enough. This is one of these examples of Brandolini’s Law (the Bullshit Asymmetry Principle); it would take about 30 pages to debunk all their confusions about IP (Calacanis claims to know it more than the others but he mangles it too; none of them understand IP law); they conflate patents, copyright, trademark, stealing, plagiarism, and contracts/terms of service; (( Stop calling patent and copyright “property”; stop calling copying “theft” and “piracy”; Copying, Patent Infringement, Copyright Infringement are not “Theft”, Stealing, Piracy, Plagiarism, Knocking Off, Ripping Off. )) they are confused about independent invention, originality, authorship, what China is doing,1 and on on… they are just all over the map.

But they do recognize that IP is a major threat to AI. And it is. See Whereupon Grok admits it (and AI) is severely gimped by copyright lawLibertarian and IP Answer Man: Artificial Intelligence and IP; Copyright Thicket and President Trump’s AI Training Data Solution.

Where they are wrong is thinking there are easy fixes, doable fixes, or any fixes at all short of abolishing copyright (and probably trademark and patent as well, but for AI copyright is the real problem). Copyright is already ruining AI.

And due to global confusion about property rights and IP, this is going to continue. These guys think they are part of the solution but they are part of the problem. The only way to see this clearly is to have a clear, i.e. Austrian and libertarian, understanding of property rights, and of the nature of IP law. Basically only pro-tech anti-IP Austro-libertarians can understand this. Meaning it’s hopeless.

Transcript of this part below along with Grok summary.

Overview, Summary, and Analysis of All-In Podcast Discussion on Copyright, Fair Use, and Patents in the Age of AI

The All-In podcast discussion, featuring hosts Chamath Palihapitiya, Jason Calacanis, David Sacks, and David Friedberg, delves into the contentious issue of intellectual property (IP) rights in the context of artificial intelligence (AI) training, particularly focusing on copyright, fair use, and the broader implications for the global AI race. The conversation, sparked by President Trump’s remarks on a pragmatic approach to IP, reveals a spectrum of perspectives on how copyright law intersects with AI development. The hosts grapple with the tension between protecting creators’ rights and ensuring AI innovation isn’t stifled by overly restrictive regulations, especially in light of global competition, notably with China. Their discussion, while insightful, is critiqued by Stephan Kinsella in his draft blog post for conflating key IP concepts and failing to propose radical enough solutions to address the systemic issues posed by copyright law.

The hosts acknowledge the critical role of data in AI development, with Sacks and Friedberg arguing that requiring individual licensing agreements for every piece of internet content—articles, websites, books—would be impractical and could handicap American AI companies, especially when compared to China’s less restrictive approach to data usage. Sacks highlights Trump’s stance that a common-sense approach to IP is necessary to avoid losing the AI race, advocating for a broad interpretation of fair use that allows AI models to train on publicly available data without exhaustive negotiations. Friedberg supports this by distinguishing between AI learning patterns from data (akin to human learning) and outright copying or plagiarizing copyrighted material, which he agrees violates copyright law. This perspective aligns with the need for AI to access vast datasets to remain competitive, but it sidesteps the deeper legal and ethical complexities of what constitutes “fair use” in an AI context.

Calacanis, however, takes a creator-centric view, emphasizing the importance of respecting IP rights to sustain industries like journalism, music, and film. He argues that licensing deals, such as the $20 million annual agreement between the New York Times and Amazon, could create a “golden age” for content creation by enabling media outlets to hire more journalists and fact-checkers. Calacanis sees these deals as evidence that AI companies must pay for access to proprietary content, particularly when their models compete with content creators commercially. He challenges the notion that China’s IP practices justify loosening U.S. copyright standards, asserting that American companies should not “steal” from creators simply because foreign competitors do. This stance reflects a belief that protecting IP incentivizes creativity and provides a competitive advantage through high-quality, licensed data.

Palihapitiya introduces a provocative angle, questioning the long-term viability of copyright and patents in an AI-driven world. He posits that AI’s ability to independently derive or recreate existing works—such as a hypothetical recreation of Michael Crichton’s Jurassic Park without direct reference—could render traditional IP protections obsolete. He suggests that businesses should focus on building defensible moats through trade secrets or other means rather than relying on legal protections that may become unenforceable. This perspective aligns with Kinsella’s critique that the hosts fail to grasp the fundamental flaws in the IP system, particularly how copyright restricts AI’s potential by limiting access to training data.

Kinsella’s blog post sharply criticizes the All-In hosts for their muddled understanding of IP law, accusing them of conflating patents, copyrights, trademarks, and terms of service while misunderstanding concepts like independent invention and plagiarism. He argues that their proposed solutions—such as refining fair use or negotiating licensing deals—are insufficient and that only abolishing copyright (and potentially other IP forms) would truly liberate AI development. Kinsella’s libertarian and Austrian economic perspective frames IP as a state-granted monopoly that distorts property rights and stifles innovation, a view echoed in his cited posts (e.g., “The China Stealing IP Myth” and “Whereupon Grok Admits It and AI Is Severely Gimped by Copyright Law”). He contends that the hosts’ confusion exemplifies Brandolini’s Law, where debunking their misconceptions would require extensive effort due to the complexity and depth of their errors.

The analysis reveals a core tension: the hosts recognize the threat copyright poses to AI innovation but are divided on solutions. Sacks and Friedberg lean toward a pragmatic, tech-friendly approach that prioritizes access to data, while Calacanis defends creators’ rights, seeing licensing as a path to mutual benefit. Palihapitiya’s forward-looking skepticism about IP’s future highlights the disruptive potential of AI but underestimates the entrenched legal frameworks. Kinsella’s critique underscores the need for a clearer understanding of property rights, arguing that the hosts’ failure to advocate for abolishing copyright perpetuates the problem. The discussion reflects broader societal debates about balancing innovation with creators’ rights, with no easy resolution in sight, especially given global disparities in IP enforcement.

Ultimately, the All-In podcast illuminates the urgency of addressing IP in the AI era but falls short of a cohesive or radical solution. The hosts’ varied perspectives—ranging from pragmatic deregulation to creator advocacy—highlight the complexity of the issue, but their conflation of IP concepts and reliance on incremental fixes, as Kinsella notes, limits their ability to propose transformative change. As AI continues to evolve, the debate over copyright and fair use will likely intensify, requiring a nuanced approach that neither stifles innovation nor disregards the contributions of content creators.

Copyright, Fair Use, and Patents in the Age of AI

Below is the cleaned-up transcript from 34:32 to the end of the provided transcript, with corrected spelling, punctuation, and speaker identification based on the context and the speakers listed (Chamath Palihapitiya, Jason Calacanis, David Sacks, David Friedberg). Timestamps are included when the speaker changes, and “Unknown Speaker” is used where the speaker cannot be confidently identified.

34:32 – Jason Calacanis: The content part of it, where—and this is something that the press was having a field day with and they really keyed on—which was, hey, respecting IP, respecting copyright. What’s the feedback been so far on that, which was a pretty spicy part of President Trump’s speech?

34:54 – David Sacks: Well, I think what the president said was just very pragmatic. He said we had to have a common-sense approach towards intellectual property. And he said if you have to make a deal with every single article on the internet, every single website, every single book, every piece of IP in order to train an AI model, it wasn’t feasible. He said, “Look, I appreciate the work that went into people creating these works, but you’re not going to be able to negotiate a deal for every single one of them. And if we require our AI models to do that and China doesn’t—and they won’t, they’re just training on everything whether it’s, you know, pirated or not—then we’re going to lose the AI race.” So, I think he took the side of a fair use definition. I don’t know if he used the term fair use, but effectively, he was taking the side of a reasonable fair use.

35:41 – Jason Calacanis: What did you think of that part, Dave Friedberg? You have any thoughts for Chamath on that part?

35:47 – David Friedberg: I think he’s absolutely right. I’ve said this before: if something’s on the internet, if something’s in the open domain, I strongly disagree with the idea that AI getting trained is the same as AI replicating copyright material. If AI outputs text or outputs audio or outputs video that contains copyright material, it is 100% in violation of copyright. And he said that, by the way.

36:07 – Jason Calacanis: Yes.

36:07 – David Friedberg: And if the AI is learning, it is understanding patterns, it is understanding reasoning, it is understanding concepts by reading copyright material, just like humans do. A writer, an author reads a bunch of fiction, learns good techniques, learns good concepts, learns good theory from reading all those books, and then goes and writes his or her own book. They are not violating copyright material.

36:29 – Jason Calacanis: In the same way, Freeberg, what if it’s all the New York Times on the open internet?

36:36 – David Friedberg: 100%. You’re 100% correct, that should be paid for or licensed. I’m talking about the open internet. I’m talking about open material. I’m talking about stuff that’s in the open domain, like Common Crawl. There’s a thing called Common Crawl.

36:49 – Jason Calacanis: If there was—if somebody stole a hundred books, let’s say, and put them on their website, and it was a pirated Russian website with a thousand books on it, and you accidentally crawled it, you would be obligated to take that out then.

36:58 – David Friedberg: I think we all agree. Correct.

37:00 – Jason Calacanis: Okay. Correct. ‘Cause that’s what a lot of the lawsuits are around. So, I think we’re reaching something. I just want to say, you know, this is such an important point, especially to me as a content creator and somebody who spent his career in this. I’ve been thinking about the endgame, and I’m here in Park City. I was just giving a keynote, and I wanted to show you something I made, Sacks, because I think we have to get to the endgame here. So, in my talk, I talked a little bit about how we can get through this fight and then maybe get to a solution. So, I had my team mock up the New York Times website here and ChatGPT doing a deal with them. So, here you see, you’re on the New York Times website, and you ask it a question powered by GPT. You ask it, “Hey,” you might ask this question. In fact, you log in with your ChatGPT credentials, and it could be Grok, it could be Gemini. “Give me the earliest mentions of Putin,” you know, if you were a fan of Putin or something. And it would then go through that and give you your Putin references. And then I made another one. And then, obviously, this would be an exclusive to ChatGPT. It would be one of those things where, you know, they get an exclusive. And then here, on the Disney Plus channel, imagine you could make yourself into a Jedi Knight, and you could then upload your photo. You know, kids might really get into this. You upload your photo, and then it makes you into a Jedi Knight. There’s Darth Calacanis.

38:27 – David Sacks: So, that looks to me like you’re infringing on their trademark.

38:30 – Jason Calacanis: What’s that?

38:31 – David Sacks: Are you infringing on their copyright?

38:33 – Jason Calacanis: This is fair use. This is fair use. This is a perfect example of fair use for editorial.

38:37 – David Sacks: You’re also infringing on some Ozempic.

38:40 – Jason Calacanis: That’s absolutely infringing. Trust me, I am definitely infringing on some Ozempic here, guys. I’m past those. I’m on to peptides now, man. I’m on the Wolverine protocol. So, look, what could go wrong? Don’t take a podcaster’s advice. Please don’t take a podcaster’s advice on your healthcare, rule number one. Take Chamath’s advice because he’s got 6% body fat, which I think attributes to much of your pomp and circumstance around your privates. I think it has to do with the lack of fat. But I’m going to leave it at that.

39:13 – Chamath Palihapitiya: First of all, it’s 11 and a half, but, you know, that’s like right before I go on summer vacation. Then it ends up at 12 or 13.

39:19 – Jason Calacanis: Did you go get that gelato? What was that place we went that we love?

39:23 – Chamath Palihapitiya: Lulume. I’ve gone there every day. Every day so far.

39:27 – Jason Calacanis: Did you do two or one? Be honest. Two or one?

39:30 – Chamath Palihapitiya: Bro, I’ve had—per session, two. I start with the medium, and then I finish with a small.

39:36 – Jason Calacanis: Yeah, exactly. This stuff is so good. I’ve never tasted any gelato like this. It’s incredible. I mean, it’s unbelievable. We have to license it for the United States and the All-In brand. We have to license it from them. It’s really incredible. But Chamath, just generally speaking, or anybody who wants to have at it—Friedberg, Sacks—what do we think about the endgame here? Because there are some major lawsuits here. They’re going to get settled in the next year or two. What do we think about sort of the future I’ve shown here today?

39:59 – Chamath Palihapitiya: I think what Sacks just highlighted is exactly right. Look, we’ve got to have a common-sense approach here, or we’re going to lose the AI race. I mean, one of the key determinants of AI quality is the amount of data that you have. It’s very simple, right? There are a few building blocks: there’s energy, there’s chips, there’s data, and there’s algorithms. And if you lose on any one of those dimensions, then you’re in trouble, right? So, look, you just can’t have a situation where China can train on the entire internet, and our AI models are hamstrung by needing to negotiate contracts with every single website.

40:35 – Jason Calacanis: But right now, Europe—Elon owns X, right? He owns Twitter, now X. Does Sam Altman have the right to use X in his corpus?

40:47 – Chamath Palihapitiya: It’s public.

40:49 – David Sacks: No, it’s not. No, it is not a public endpoint.

40:53 – Jason Calacanis: I just honestly, I don’t—there’s—I don’t know the answer to that. There are some edge cases here. We’re going to have to come up with—

40:58 – David Sacks: It’s not about whether it’s behind a paywall or not. It’s whether these APIs exist and whether you’re contractually allowed to use them or not. The terms of service.

41:09 – Chamath Palihapitiya: Correct. The terms of service. It’s published on every website, what the terms of service are with respect to the content.

41:15 – Jason Calacanis: I think it would be okay to let people opt out, you know. So, we already have this with Common Crawl. You can put in the footer of the website, you put in robots.txt, and you opt out of Common Crawl. Common Crawl is, like, this nonprofit organization that basically archives the entire web every few months.

41:27 – David Friedberg: Funded by Gil Elbaz, formerly of Google. Great fan of the pod. Shows up to our summits. Great guy.

41:33 – Jason Calacanis: And all of Open AI was built off of Common Crawl originally. But they’re very clear, by the way. They say you have to clear copyrights. You don’t get to just use Common Crawl.

41:45 – Chamath Palihapitiya: Can I go out on a limb? I don’t know if you guys saw this Amazon deal with the New York Times for $25 million. Did you see that today?

41:51 – Jason Calacanis: No, I didn’t see it today. Explain it, please.

41:54 – Chamath Palihapitiya: I think that the New York Times licensed Amazon all of their content, including The Athletic and a bunch of other things, for training. $20 million. Sorry, $20 million a year.

42:09 – Jason Calacanis: Okay, here we go. I read that, and I thought this is the peak of these deals. These deals will only go down in terms of dollar value from here.

42:16 – Chamath Palihapitiya: And it actually brought me to this point where I was thinking to myself, is it even realistic to believe that patents and copyrights actually exist in five years? And I went through this exercise of, like, if a computer studies the periodic table and also understands the laws of physics, the laws of biology, the laws of chemistry, and then independently derives some material that is otherwise patented, what will happen? And then, separately, if two competing AIs invent a new material from scratch, how will the international courts deal with this? And if you take all of these examples to the limit, at the limit, the idea that there are enforceable copyrights, I think, is a very fragile assumption. So, I’m actually thinking more that we have to spend some time understanding the landscape of a world that doesn’t have copyrights and patent protections and, instead, what is the surface area in which you compete? What is a trade secret? What does that mean in a world of AI? And I think it’s quite an interesting thing to think about.

43:37 – Jason Calacanis: Patents are a totally different piece. I think that’s a fascinating string to pull on. I will tell you, I will take the other side of the bet if we want to make a Polymarket on this. I will guarantee that this will be the beginning of the deals, and the deals will go up from here. I’ll tell you why. The reason the New York Times made that deal is to make it apparent that what Open AI has done has damaged their business because now they have a customer, and their customer is Jeff Bezos at Amazon and Jassy, and now they can show damages. And those damages could give them an injunction against Open AI, and Open AI’s got to take it out of their crawl of their, you know, construct, and that’s going to be really expensive for them. It’s not undoable, but it’s going to be expensive. And let’s think, on a societal basis, of what we want as a society. Do we want a society in which journalists, writers, artists, musicians, filmmakers, actors cannot make a living, podcasters, or do we want a world in which they can? And I think technologists—hold on, let me—hold on—as a technologist, as a technologist, we typically think if we can crawl it, it’s ours. What I can tell you as an artist is, if I make it, it’s mine, and you need my permission because it’s my art. And I think the industry will do better if they respect them because now the New York Times can hire more fact-checkers.

44:59 – Chamath Palihapitiya: But can I just ask you a question?

45:01 – Jason Calacanis: Yeah, go ahead. Sure.

45:02 – Chamath Palihapitiya: Why do you have to connect the two as immutable things? Meaning, why can’t somebody make something, still—you know, let’s just say it’s a song, but that song can now be made by multiple AI models. But if they make the song, there’s a reasonable claim that, even if they don’t have the copyright, more people will want them to perform the song than some random AI.

45:21 – Jason Calacanis: So, can’t you make a living without having the copyright, which is the choice of the artist? Some artists are very well known for not wanting their art to exist in some mediums. As a perfect example, the Rolling Stones, for a long time, thought they would be sellouts if they had their music used in commercials. And when they did “Start Me Up” with Windows, that was a really big concession from them. And that’s up to the artist to make that decision. You make a valid claim: “Hey, yeah, you go on tour and make more money.” But that’s the artist’s decision, not the technologist’s or the people stealing their content. And, by the way, $20 million a year is a hundred $200,000 highly paid journalists, fact-checkers at the New York Times. They’re going to get 10 of those deals, and it’s going to create a golden age of journalism and content. And we should be happy.

46:06 – Chamath Palihapitiya: I told you this example, Jason, but at Beast, we did a licensing deal of our content to allow Open AI to run training runs on our videos. And at the board, the thing that we kept talking about was, I was really concerned, like, let’s just do a couple-year deal max. And the reason is we have no idea what this looks like in five or 10 years. And there’s just as much chance, to your point, that we get it wrong as right. Now, that was about six months ago. And so, the intuition that I had back then was maybe we should keep the deal term as short as possible. But now, when I see how important AI is in the global landscape and what China is doing, I think, on the margins, that this idea that these copyrights will mean something—in my mind, I am underwriting the value of these things going to zero, and I’m asking myself, instead, for my businesses, how are we actually building a real defensible moat and not a piece of paper that we can use to sue somebody?

47:09 – David Friedberg: Okay, Freeberg, you want the last word here? We’ve got to move on to some other topics.

47:12 – Jason Calacanis: I just want to be the last word.

47:14 – David Friedberg: I just want to be clear that nobody is losing their copyright. Copyright is the right not to have your work copied. And if an AI model produces outputs that copy or plagiarize your work, then that’s a violation of the law. And I think the president specifically said that we’re not allowing copying or plagiarizing. The question is whether AI models are allowed to do math on the internet—you know, pattern recognition. Basically, that’s what it is. And, JCal, I think you’re conflating the two. And I don’t want to be interrupted. I just want to say this.

47:49 – Jason Calacanis: I understand the distinction.

47:51 – David Friedberg: And I think that this idea that, like, I can’t, for example, go to the library, rent a book, read it, and then learn some of the good techniques on how to write a good book should be restricted to humans in this AI context. Like, this is exactly what they’re doing. They’re identifying patterns, and then they’re building predictive algorithms that allow them to output stuff that starts to fit within different kinds of, you know, variable settings.

48:12 – Chamath Palihapitiya: Do you guys think it’s possible that if you allocated enough compute at the problem, you could write Michael Crichton’s Jurassic Park de novo without ever having read it?

48:18 – David Friedberg: Yeah, me too.

48:20 – Jason Calacanis: Me too. I think—I don’t know what that would mean. Like, well, this is my point. I know who Michael is, and I know what Jurassic Park is. I don’t know what it means—this issue. I don’t know what it means to say, “Can AI write that?” But, you guys remember the Ed Sheeran lawsuit?

48:39 – David Friedberg: I did.

48:40 – Jason Calacanis: But let me just make one point here on this, ‘cause you’re saying I don’t understand it. I spent my career in it. I understand it much better than you do, and I understand it from lawsuits and being in the weeds on it. Like, I understand it from first principles, which you do not. And I will say, this is what we’re talking about here: the definition of a derivative work. And the output matters. So, if you were to take my knowledge and then create a derivative work from it, and you used a percentage of my work—and that’s where this will get into the nuance—is what percentage of the original work is used in the derivative work and under what context, a commercial context or a non-commercial? This is clearly a commercial one. If Open AI was a nonprofit right now, we’d be having a distinctly different discussion because there would be—you wouldn’t be competing with me as the copyright holder to use this new medium and create the derivative works, and it has to change substantially. So, if it’s a Cliff’s Notes—

49:36 – Chamath Palihapitiya: When China has the only models that are able to meet your stringent definitions of copyright—

49:41 – Jason Calacanis: Well, no. Here’s the thing. I think the China fear—the China fear bullshit is bullshit. I’ll be totally honest here. Just because China steals IP does not mean you get to steal from Americans. In America, we have rules. And when you go to China—and, by the way, we spent the last 30 years—the major issue with China is not Taiwan. It has been the technology industry itself.

50:00 – David Sacks: Let me finish.

50:01 – Jason Calacanis: The technology industry itself has leaned on our government for 30, 40 years, including Microsoft, including Google, to make sure our trade secrets are not stolen, our IP is not stolen, our movies are not stolen. That is the key issue with China. So, just because China’s a thief does not mean American companies get to—

50:13 – David Sacks: Have you seen the latest batch of Chinese open-source models or open models? They steal everything.

50:18 – Jason Calacanis: Does that mean you should be able to steal Windows?

50:20 – David Sacks: Should you be able to steal—

50:22 – Chamath Palihapitiya: Jason, we don’t think it’s stealing.

50:24 – David Sacks: Elon has said this pretty clearly, but Grok 5 and, for sure, Grok 6 will not use Common Crawl. It will not use the internet. Okay? It’ll just be an enormous amount of synthetic data. And, back to what Freeberg and I just agreed upon, if you synthetically go and try to generate all this content to learn across, you’re invariably going to produce something that’s already been created.

50:49 – Jason Calacanis: That’s like some sci-fi level—

50:51 – David Sacks: I understand. That’s what’s happening now.

50:53 – Jason Calacanis: It’s happening now. If somebody—what do you think happens to Grok 5 or Grok 6? Is that violating copyright? It didn’t even know that it existed—

50:59 – David Sacks: On the output.

51:00 – Jason Calacanis: Yeah, that’s fine. If, on the output, it created a similar work, they would need to then take it down. And so, that would be a really interesting new—that’s a new space we’re going to have to contend with. So, you can—if it does happen, it’s a new concept that we would have to address in a new way.

51:17 – David Friedberg: I’ll give you a science corner example. There’s this Evo2 model that they published at the Arc Institute, which Patrick Collison, you know, is the name—Sherman. So, that Evo2 model, they just ingested all the DNA data they could find in the world—trillions and trillions of base pairs of data that they ingested—and then they looked at patterns in DNA, and that’s it. They had no context for what the DNA represented. They had no context for the concept of genes, none of the structured understanding of what that DNA does, what it is. And you know what it did? It fed in the BA gene variant, and the thing output a warning saying, “I think that this is a pathogenic variant to DNA,” without having any context. This is the breast cancer allele, and it didn’t have any knowledge that there are pathogenic variants for cancer, and it identified that this was a genetic variant that can cause some sort of pathogenic outcome in the organism. So, that’s a great example where there’s a lack of understanding at the human level on what really drives some of the patterns in nature, the patterns in society, the patterns in behavior that are kind of emergent phenomena, perhaps, that these AI models are starting to identify.

52:27 – Jason Calacanis: And I think, to Chamath’s point, we may end up seeing this in things like entertainment as well.

52:33 – David Friedberg: All right, this has been an amazing debate. We’ve got to move on. And you know what? We’re going to have more amazing debates September 7th through 9th in Los Angeles at the All-In Summit. The lineup is stacked: Alibaba’s co-founder Joe Tsai, Thoma Bravo co-founder, ARK Invest’s Cathie Wood, Uber CEO Dara Khosrowshahi, Sequoia’s Roelof Botha, YouTuber Cleo Abram, and many, many more coming.

52:56 – Chamath Palihapitiya: You get the last word here, go.

52:58 – David Sacks: I was just highlighting this tweet that I saw where—talking about Chinese open-weight models are basically open-source models. So, basically, all the leading American models are closed-source, and all the leading Chinese models are open-source. This is kind of the way things played out. It’s a pretty good technique for catching up because then you’ve got the larger open-source developer community helping you out. But the point is just that these open-source models are catching up pretty fast. We’re ahead in many other aspects. Our chips are a lot better. Our data centers are better, and so on. And I’d say our closed-source models are better. But they have this one area of open-source models. So, again, if you hamstring our AI models’ access to data by creating a whole bunch of new requirements for contract negotiations, like, we could really lose the AI race. This is a really big deal. It’s not a made-up concern.

53:46 – Jason Calacanis: I don’t know why you think it’s made up. I never said that it’s made up. I think it’s an opportunity for America to actually have a distinct advantage, which is that $20 million from Amazon alone is 1% of the New York Times’ revenue. And that’s going to go directly to the bottom line. It’s going to allow them to hire more journalists. Then that protected site will be giving, in real-time, something. These language models are going to have to go hack and steal that real-time data. That’s going to be a distinct advantage for Gemini, Open AI, Amazon, whoever chooses to do it. And we can create—you have this, like, nostalgic, sort of quasi-romantic notion about journalism and the need to save the New York Times.

54:23 – David Sacks: It’s also art. I mean, you can say all the derogatory things you want about me personally, Sacks. That argument doesn’t work.

54:29 – David Sacks: No, no. You just said I have this whole nostalgia—whatever—when you—

54:32 – Jason Calacanis: Yeah, you do. You’re nostalgic for journalism as it used to exist.

54:34 – David Sacks: When I know I’ve beat you in the debate is when you make it personal like that.

54:37 – Jason Calacanis: It’s not personal. I’m not being nostalgic. I’m trying to create a sustainable advantage for America. And you are our public servant, and you’re learning AI. You will take my feedback.

54:52 – David Sacks: You will take my feedback.

54:54 – Jason Calacanis: We’re going to ignore your feedback.

54:56 – David Sacks: Take your feedback.

54:57 – Jason Calacanis: Public service. Throwing it in the trash.

54:59 – David Sacks: No, you take it, and I will be showing up at the White House for my tour. You have this crazy idea that we’re going to win the AI race by tying one hand behind our back so that you can subsidize journalists, so you can subsidize movies.

55:11 – Jason Calacanis: You’ll get more content. You said before you want more training data—pay for it. Pay for more training data. You’re the czar. Take it back to POTUS.

55:24 – David Sacks: All right, let’s keep moving here.

55:26 – Jason Calacanis: We have to keep moving. We have a great—this is great debate. Great debate here on the All-In podcast. It’s not going to stop, folks.

55:32 – David Sacks: It’s just you yelling. It’s just you yelling, saying things that don’t make sense.

55:35 – Jason Calacanis: Okay, you can say that. You only have, like, three topics going on. You can personally attack. You know what it is? It’s, like, we’ve got to let in more immigrants, number one. Number two, high-skilled immigrants. AI is going to put everyone out of work. By the way, no sense of perceived contradiction between those two things. Number three, we need to, like, subsidize—

55:52 – Jason Calacanis: Here, you know, the audience says the same. When the three of you guys attack me—

55:57 – David Sacks: When the audience gang up on me like this—

56:00 – Jason Calacanis: The three of you gang up on this, and you personally attack me, the audience comes up to me, and they say, “Wow, you really nailed and beat—”

56:08 – David Sacks: Have you done that today?

56:10 – Jason Calacanis: No, not yet. Not yet. But a little bit of the booing.

56:13 – David Sacks: Yeah, that’s true.

56:15 – Jason Calacanis: Let him eat. He’s emaciated. He’s 11% body fat. Let him eat.

56:19 – David Sacks: Let him cook.

56:20 – Jason Calacanis: All right. Listen, you and I, Sacks, will do more debate, and it’s going to be amazing. Allin.com/yada-yada-yada for tickets. Get in there, folks. We have to get to the docket. We’re an hour in, and we still have all the news.

  1. The China Stealing IP Myth []
Share
{ 1 comment… add one }