Jun 28, 2025 ยท 2025 #25. Read the transcript grouped by speaker, inspect word-level timecodes, and optionally turn subtitles on for direct video playback
Edit labels for this show, save them in this browser, or download a JSON override for the production folder.
Transcript Playback
Copyright Can't Save Publishers from AI
Human Transcript
Timed transcript
Blocks are grouped by speaker for readability. Expand a block to inspect word-level timing.
Speaker 2
everybody it's saturday june the 28th 2025 this show probably won't be broadcast until the 29th we're catching up on tech uh we haven't done a show for a couple of weeks i was in london last weekend a couple of weeks ago we did a show with keith from that was the week keith tier about the company that ate the web uh google's quarter century journey from bridge builder to web destroy i'm not sure if Google is the company that ate the web, or the web was the thing that ate Google. That's still quite uncertain. And the title of Keith's newsletter a couple of weeks ago was The Speed of Now. AI is fast, humans are slow, and that hasn't changed in the last... two weeks. In fact, the speed of now is now the speed of now times two or three. This time for this week, Keith's newsletter is entitled Copyright Can't Save Publishers from AI. There's been a couple of very interesting court cases since we last Did a show, Keith. Tell us about those cases, why they're so significant, why the speed of now is even quicker. Now, two weeks after you did that previous newsletter.
Yeah. So let's start at the end. The end, which, by the way, isn't the end because there can be appeals, but the current end is that courts believe that fair use includes the right to learn from published content. And the right to learn is not the same as the right to copy. So the court cases revolved around, is there any proof that AI is copying? And the broad answer was no, there is no proof that AI is copying, but there is proof that it's learning. And learning is fair use. Therefore, Getty Images, in its fight with stability AI and other cases, came down on the side of AI.
Yeah, my wife, as most of you know, is the head of litigation at Google shyly told me this week as we were driving to the movies, which will probably be destroyed by all this too, that there were two good cases this week. I think from Google's point of view and OpenAI, of course, things are going very well. Tell us about this Getty trial. You entitled it Getty retreats, but the war isn't over. So what was the retreat and why isn't the war over?
Well, the war isn't over because fair use is only one of three accusations that Getty are making. So they lost on fair use, but there's still very detailed, kind of slightly geeky legal points that are not resolved, which Getty is still fighting. And if they win those, might constrain even AI learning, but not based on winning fair use, but based on these more minor points. By the way, I don't fully understand the minor points. I'm not a lawyer. But they are still out there. So the court didn't find against Getty on everything, only on the main accusation.
Yeah, so as you say in the editorial, fair use is the resulting court was of mixed signals. What about the other big case, the anthropic case? That seems to be clearer in terms of the win for AI.
Yeah, so the Anthropic case, EFF writes a piece about it. It's kind of interesting. And it's a company called Bartz that accused Anthropic of using their books to train the Claude Chappell. I don't think it was a company. I think it's a person. It's a person. The writer. Oh, it's the writer of the books. Okay, sorry. I thought Bartz was a company. I didn't go and check who Bartz was. Anyway,
basically, the claim that an AI, having read something, can write new material that competes with existing material, emulating themes or substantive points or grammar or composition or style, Bartz felt that was infringing, and the court found it is not. The only thing that's infringing is copying.
Which is a huge deal, as you suggest in this editorial this week. Copyright can't save publishers from AI because publishers, the creative community more broadly, are in the business of creating original content. Why is this such an existential decision, Keith, in your view?
Well, because I don't primarily think of myself as a publisher, I've always had the point of view that an AI is, at least abstractly speaking, no different than a human. It just can be lots of humans at scale, but essentially it's like a human. So if a human can read, let's say the New York Times and then go and write something that couldn't have been written without having read the New York Times, that human isn't held to have done anything wrong. In fact, the opposite, the human is applauded for having done their research. So I've always thought AI should be treated like that.
No, I'm less interested in your view and more interested in the broader implications, for better or worse, of the content industry. Why is this such a huge problem? I mean, the whole point of your... I thought it was one of your better editorials, as I said, when we went live and you acknowledged that perplexity helped you write it, which probably adds to the ironic complexity of this conversation. Probably even this conversation is somehow being aided by AI, Keith.
So my preamble about me was just preamble. The reason it's a big deal depends on which point of view you're examining it from. So it's a big deal from the AI point of view, and it's also a big deal from publishers point of view. It's probably a big deal from search engines point of view, which we can get to later. But let's take the publishers point of view. The way modern publishing works is firstly, it's largely digital. Printed media and even broadcast media is largely a loss leader for digital media these days. And digital media's business model is advertising. And most of that advertising comes from people like Google who do copy and put it in a search engine.
Yeah, so basically that search engine has a social contract with publishers. You can index and therefore, in quotes, copy our stuff as long as you link back to us and send us traffic. That's the social contract.
Well, I mean, Google can't just have the entire article. It has a couple of headlines. There's always been a debate about how much it could publish. And then you press the link and you go to the New York Times or the Wall Street Journal or the Washington Post.
Well, it actually depends what you search for. Google does have the entire article, but it doesn't show the entire article if you're just searching for, let's say, the title. But if you search for something that's in a paragraph, it will find it because it has the entire article. So Google is allowed to have the entire article in its index, but it doesn't reproduce it in whole. Now, AI doesn't have an index. There's no concept of an index in AI. So it reads what you published, but it doesn't index it and doesn't contain a whole copy of it. What it does is it builds something called a vector database. which is a list of words with statistical numbers next to them that imply that word A is close to word B. And then the AI guesses words in a sentence or a paragraph. And so there's no reproduction even possible because it doesn't contain the original.
And it's the same with art. I always make fun of your AI art, but that's exactly for people watching this. And this was the Getty images issue because they're an image company rather than a text company. The same is true when it comes to pictures or images.
As they are already, I mean, a lot of consumers are already using it, even if they don't quite know it. Now, as you note in the editorial, this is a challenge for everybody, not just publishers, but also for Google of the world. As I said, last time we talked about the company that ate the web, Google, which did eat the web in 1997, 1998, 1999.
Yeah, there's a long interview with Google's CEO this week. And he talks about this AI mode, it's called. AI mode is when you replace search mode with AI mode in Chrome using Gemini. And he talks about the fact that about a third of Google's revenue doesn't come from search advertising. Two thirds does. And he's put on the spot to discuss how Google will adjust to a world in which AI mode starts to dominate. And to be fair to me, it comes across fairly confident that that's a journey they can traverse. But if you're a publisher, that is the end of traffic from search results.
And this is the point you... I don't always want to do the post of the week before the end, but this was your... interesting post of the week from quoting Cloudflare CEO at Cannes, the advertising get-together, annual advertising, suggesting that the publishing industry is challenged. They're just not going to get the same return, right? And it's becoming more and more dramatic.
Well, yeah. The Om Malik piece also speaks to this, which is the prediction that as a result of AI, the internet publishers are going to have to go to putting tolls in front of their content so that they can earn money from subscriptions, not from advertising.
Yeah, and Om, who I know is a friend of yours, he's been around a long time, so he's lived through a lot of these changes. He talks about something called the internet, not of trolls, but of tolls, a new commons or a fragmented future. my guess it will be a fragmented future. There's always the promise of the new commons, which never is realized. So how will publishers then, I mean, they'll have content, which is that, I mean, it will be quote unquote theirs. How will they monetize it?
Well, it becomes a pre-internet business model. If you think before the internet, a publisher only got the attention it deserves due to who bought its content. So they're going to go back to a pre-internet business model. And then whatever bump they can get from AI will be on top of that. I don't think the AI bump will be zero. I do think there will be free to use AIs that do include links and advertising. So it won't be zero.
But the tolls, I mean, the tolls idea sounds a little dramatic. Isn't that how the world works? If you want to buy a newspaper in the pre-internet age, you walk to the newsagent and you paid for the New York Times. Now... you pay quote-unquote the toll by subscribing. Isn't that what the toll model is?
Yeah, it's a subscription model, exactly that. Now, what the CEO of Cloudflare is implying, and it's super interesting, is that, so before we say it, let's just understand that Cloudflare is a network layer on the internet. So if you're a customer of Cloudflare, and you can do this for free, it runs your, what's called the DNS, the domain name system. And because of that, it is in the middle of all the traffic. And when it's in the middle of all the traffic, it can build applications that perform various actions. And what he is implying is that Cloudflare is gonna offer publishers a learning algorithm that charges AIs for learning from the content at the network level, which means it will be impossible for an AI to break the rules unless it's paying. In fact, there'll be a toll on the AIs for learning. And this will create, if you like, microtransactions for publishers insofar as they permit and put a price on AI learning.
So let me try and use the example of the New York Times. Are you suggesting that, and obviously the New York Times still exists and it has its headlines, but if you want to use the AI of the New York Times, which will be a very high-walled garden, you have to subscribe And in that sense, the AI driving the New York Times might be Gemini, might be OpenAI, will be paid. Is that how Gemini or OpenAI will get its money?
So you're looking at it from the point of view of the consumer. Matthew Prince, the CEO of CloudFry, is looking at it from the point of view of the New York Times. He's saying from the point of view of the New York Times, we can do a deal that blocks OpenAI and Anthropic and all the others, including Gemini, from learning on the New York Times unless they pay. So he's saying the New York Times can get a revenue stream from all the AIs for learning And now whether the New York Times has an AI itself for consumers, that's a totally separate issue. They may or may not do that. They'd be smart if they did. And when we talk about the deer browser, we'll see how AI is going to end up being built into everything. And that may create revenue streams for them.
So it's good for the New York Times. They have a huge amount of content, but for the individual blogger or poster, it's worthless. No one's going to pay them for
It depends on the AI's learning needs. I'm gonna guess you'll set your own price as a publisher. So on Kinon, you could say it's, you could measure it by tokens. It's a dollar per million tokens. And a token is a measure of the indexing reading capability of the AI. and they'll choose whether to pay it or not. It'll be transparent to them.
Yeah, and you'll have a deal with Cloudflare. They'll probably keep a wallet for you and escrowed money. And so that's what's interesting. It's a network layer idea, which, of course, that's what Cloudflare do. They do network layer innovations.
So, MG, another article you link to, which is very interesting, is from MG Siegler, who's younger than O'Malley, but has been around the block a few times as well. Used to be at TechCrunch. He talks about the AI browser wars beginning. I don't quite know what the title begun the AI browser wars have. I don't know whether he's making sure.
Oh, I don't get that. But anyway, he refers to DIA, which I know you're using as a model for this new By the way, it's an AI browser. I'm not sure if that's a contradiction in terms, but it's a browser powered by AI. Is that one way of putting it?
Well, to be specific, it's using the Chrome open source code called Chromium, which a lot of browsers that you've never heard of do that. And then it's building an AI side panel so that any page you're on you can talk to the AI about that page. It also allows writing and coding in the AI. So it becomes, if you like, a one-stop interface for creating and consuming. Why is that different from, say, Anthropics Claude?
It can browse, but you can't include. With Dia, you are doing the browsing. It pulls in your bookmarks just like a normal browser. How different is that from perplexity? Same with perplexity. You can't actually put a URL in and browse to it there, but you can with Dia. Other than that, it's quite similar to perplexity.
I'm not bullish yet. I've just started playing with it. Let's see. But it seems obvious. And by the way, Google is already building AI into its browser all over the place. Gemini is available in all kinds of places in Chrome. So it does seem to be a valid direction of travel. The question is whether Dia can do anything valuable enough to make you want to use it instead of Chrome. And I'm a little bit dubious that that will pan out, but you never know.
And adding to the complexity and irony of all this, of course, when it comes to Chrome, the biggest legal challenge to Google in the long term, when it came to the... the competitive case, the monopoly case, is that it might be required to sell off Chrome. Is that a challenge now to Google, being asked to cut off a limb that no longer has any value?
Well, I've always said in this conversation about Chrome, I've always made the point that Chrome is only makes revenue for Google insofar as people do searches with ads in them, or all browse to websites that include Google ads. And that won't change if Chrome was sold off. Chrome itself is a zero revenue cost center, but the traffic makes money. Now, if Chrome increasingly is in AI mode, not search mode, then that revenue goes away, or a lot of it. So the value of Chrome is a declining asset, clearly a declining asset going against the trend of history.
So you're suggesting that, or Malik and Fred Vogelstein are arguing that we have two choices, a new commons or a fragmented future. You developed that. the internet of tolls, which is inevitable, is either a new commons or a fragmented future. You conclude, as you always do, Keith, with agency suggesting that the road ahead is ours to choose. I actually strongly disagree with that. I don't think any of us really are choosing this. How is this going to get determined? It's not like users are going to determine it.
Well, just to give away some billion-dollar ideas here, let's say, for example, I did a startup next week. that went to every e-commerce site and asked them to upload all of their products with the links and the descriptive text. And I promised to isolate that from AI unless AI pays to read it.
Well, I'm talking about collectively doing it. Let's call it e-commerce for AI search engine. And you put all the links and all the related metadata in there. And now AI could surface links to products because it was allowed to learn on it and not get paid. So referral fees would go away. That would be a business model to preserve commerce in the age of AI, even though there's no web searches happening or less web searches. And there's a couple of articles this week talking about how data is the new gold. That's an example of data being the new gold.
Well, it means that data is the link to real world stuff. And real world, let's say the New York Times did it for every article in collaboration with lots of other publishers. There would be lots of ways to monetize publishers' data, which is another way of saying content. And I think innovation is going to lead there. in a way where links are still valuable and clicks are still valuable, but advertising is less and less the method of doing it. It's much more the content itself.
Yeah, and you didn't have it, but in the last couple of weeks, the CEO of WPP has resigned, acknowledging that AI is an existential threat to the company. So the CEO of Cloudflare made this speech about these changes at Cannes, which is the annual advertising get-together. This is particularly troubling for the advertising business, isn't it?
Well, it is, although the advertising business is an adjunct of the selling business. So the real question is, does selling still survive? Selling, what do you mean by selling? Trying to get attention for content or products or services.
the core of advertising is you're selling a product, some content, or a service. And that doesn't go away because humans have an endless curiosity and desire to see things. So the only real sub-question to that is, is advertising still the mechanism whereby that attention is garnered? And there are many other, I would argue, better ways of getting attention than advertising. And so the loss of advertising may only be sad for advertising businesses, as you say.
So the real question then becomes, and this we... talked about this endlessly, and we will talk about this endlessly, is from a content provider, given these two cases that you talk about, which allow AI to learn from other people's content, I mean, what should content creators be doing? If the AI can learn from it, it can replicate it, what's the value of human-created content. Is there a long-term value?
I think there is. Let's use the example that we have talked about a lot recently of abundance. OpenAI can probably give a pretty good answer to what...
You mean the book by Ezra Klein and Derek Thompson?
Words and timings
YoumeanthebookbyEzraKleinandDerekThompson?
Speaker 1
Not only the book, the concept itself, because it goes beyond Ezra Klein. It's now a broader conversation. OpenAI can probably give a pretty good answer to what is the current discussion on abundance and will probably refer to all kinds of things. But it will be a lot better if it surfaces the link to Klein's book or perhaps his New York Times articles. And if the New York Times can get paid when those links get clicked on, the New York Times will be a net beneficiary of that. And so I think the question of what to do with links, and that's a question both for AI and for the New York Times and for you and me as consumers, that's probably the most important question of the time, because links are clicks and clicks are money.
it's all being devalued and then as the post of the week uh your post of the week suggests what is it uh used to be one in uh well the numbers are astonishing and it used to be what one in 18 and and with open ai it's gone up to one one in 1500 and soon it will probably go up to one in a million
And that's because links are not a currency yet in AI. Nobody has innovated to make links a currency. And so links in OpenAI are currency-less. Insofar as you can get links, no one's making money. That can change. And I think it will change. If I wasn't doing SignalRank, it would be my next startup would be link monetization in the context of AI.
And I wonder, the more we talk about this, it's all a bit vague, whether we're still in the pre-Netscape moment. You talk about DIA. We've talked about perplexity. I've used perplexity. It doesn't seem to be particularly convincing. Have we yet really got to a native AI product, a win that's easy to use, accessible, and logical? I mean, that's the holy grail. I guess for you, it's probably chat GPT. I'm not convinced of that.
You know, I think the answer is no, but it's a no with an asterisk. And the asterisk says it's getting closer and closer. We're in that mode where people are reimagining current technologies and adding AI into them. And honestly, that's mainly what Dia has done. At some point, somebody's gonna create something which doesn't repurpose an existing thing and create something new. Arguably, the chat interface at OpenAI was an example of that, but I do think there'll be more. And we'll all have probably many aha moments, like what replaces the spreadsheet, what replaces the database, those kind of moments. um and um you know that's gonna be what you call native ai it it it's it's getting there with voice if you look at whisper which is the voice technology for max that you can download and install it really does create conversational ai on your mac i i think the airpods
Well, it's the Johnny Ive thing that he's doing with OpenAI. I mean, I think what you're suggesting is we've been sort of softened up for this aha moment. Yes. We haven't seen it yet, but we're creeping there. So when it comes, it won't be... a massive shock. And I'm guessing there's, I mean, there's many, many people in many garages around the world working on these products.
Yeah. And even the spreadsheet when it happened, Dan Brinklin invented the spreadsheet back in the day. There were already, you know, written ledgers and the spreadsheet looked like a digital version of a written ledger. So even that was, even though it was an aha moment, it wasn't unfamiliar.
So finally, Keith, what becomes, I mean, you say copyright can't save publishers from AI. Is the future for publishers bright or dark? Are they finished or is this just another chapter in the history of publishing?
The answer depends on what you think of publishers. There is the opportunity for it to be a huge plus for publishers if they understand the value of links. And if they're prepared to work with AI to ensure links are present and used and monetized. That's the key.
Well, let's talk about the Wall of Garden as a second level discussion. The top level, and the New York Times already does this, you can go to the New York Times as a non-subscriber and see an article that you click on, even today. You just can't read lots of articles. So the New York Times lives on traffic, and what they're losing with search and browsers is traffic. that traffic can be replaced probably times 10 if they surface links to the AI companies and the landing page you go to includes subscription and advertising as it already does. So they've got to want traffic from AI and make it easy for AI to serve that traffic without undermining the AI value proposition, which is knowledge. And I think that's doable, but it would involve publishers doing the, you know, eating their own children, if you were, moving away from search as their primary source of traffic to AI as their primary source of traffic.
Huge opportunities, huge challenges. Certainly won't be the last time we talk about this. It's the dominant thing. been the dominant theme over the last few weeks months years keith ever since open ai we will talk about it more uh we've got one more show coming up in the first week of july and keith and i are away for the rest of the month so we will talk again next week keep well and keep thinking about the future because we can't do without either it or you thanks so much keith thank you bye everyone