Baby Please Don't Go

Speaker

I've got sunshine on a cloudy day When it's cold outside, I've got the month of

Speaker

December the 15th, 2023. If it's Friday, it must be that was the week and it's the last. That was the week of the year. Keith has decided to take the rest of the week or the year off to recharge his batteries. One thing can be sure from my experience of that was the week. Art, Dali generated or all these other AI generated art, they're not going to put artists out of work. The quality of the art is dreadful and it gets worse. Every new piece of art that Keith creates for that was the week seems to be worse than the one before. Keith, am I being unfair? Speaking as not an artist myself, I'm not sure my judgment counts. But I would say compared to what I could have come up with without Dali, it's a piece of excellent artwork. What it's showing by the way, what it's showing is the media on the left. Yeah, we need to show because some people just be listening. We have two images, one of a, I guess, a traditional Google world and another of a Google AI world and a group of people are shouting at one another and the catchphrase is baby, please don't go, big media needs Google, web search. So what's the point, Keith? Well, so obviously web search started back in 1997 and initially was responsible for sending lots of traffic to places like the New York Times still is. And at some point in history around the middle of the 2010-ish kind of era, media started to get the idea that Google was stealing its lunch and wanting Google to pay it for the content that Google indexes. And of course that never panned out because Google is a search engine and an index and the media benefits, one would argue, by traffic, but they never saw it that way themselves. So there's been this constant narrative. Although to be, I mean, it's not that simple because of course in Europe in particular, Google convinced, not Google, Google's critics convinced some regulators to force Google to pay for links back. I think they almost never have. There are some exceptions, it's true, but they almost never have. I think they were going to pull out Google. I think they did it in Canada. They threatened to pull out if they had to pay aggregators or newspapers, online newspapers for Google links. And the reason I bring up, by the way, that thing about putting people out of work, Dali putting artists out of work, is of course this is part of an ongoing discussion about whether or not AI is going to put creatives, journalists, writers, artists out of work. Yeah, anyway, so with the images, the left-hand one is historically saying it's the media shouting at Google, you're stealing our content. On the right-hand side is this week's news, which is baby please don't go, with the media screaming at Google to not adopt AI in replacement for search, but to keep search going because the media feeds off the traffic it sends them. So it's kind of turned into its opposite. And all this comes from, I thought, a particularly interesting piece in the journal, Wall Street Journal, which tends to have very good quality, high quality tech reporting, on a piece, I think it was from yesterday, news publishers see Google's AI search tool as a traffic destroying nightmare. So what's the point of that piece, Keen? Well, obviously, when you use AI, Google's version is called BARD, and underneath the surface of BARD is Gemini Plus, which is the the large language model. Well, they just came out with a new version last week, too. Exactly. So basically, they cooked. So if Andrew Keen, instead of going to Chrome and typing in a search and getting search results, starts using BARD, there will be no links, but BARD will give you the answers to the queries, it will have been trained on all kinds of data, including the ability to do web searches. But those web searches will be hidden in the background, and BARD will just give you the results, often with no link. So the fear is, AI will get rid of traffic that emanates from links delivered by Google. And that is a genuine fear. I mean, you and I were playing around this week with ChatGPT, and you played some audio directly linked with ChatGPT over your phone that was actually quite compelling. So we're not that far away from this supposed dystopia, at least for search engines. Yeah, I think the big point is that web search quite probably represents a stage in the evolution of technology that will become rapidly outdated with AI. And so the very concept of web search, even the concept of traffic, may go away. Because traffic is only, the only reason traffic exists is because users want content. And if content is delivered to them by AI, there will be no traffic, there'll just be flows of information. And this is, I mean, this word gets used all the time in Silicon Valley, disruptive. But this is, this is, and to borrow another word radical, this is radically disruptive for the web2 economy, isn't it, Keith? Very disruptive. It take, it take, web2 was really about the move away from portals to distributed content. It coincided with things like YouTube embedded videos, or RSS feeds, or newsreaders. And this is the transition from content to be navigated to, or even pulled in. Now you're moving to content being understood and explained to you, and you don't actually see the content. And all this is in the context, there was the, you have a link to the Wall Street Journal piece on whether or not this Google AI search is a nightmare, or perhaps the solution for the creative community. And alongside this, there was a very interesting announcement of a partnership between Axel Springer, the German publisher, and OpenAI. So tell us about that and how this connects to the question of the future of search. Yeah. So I'm going to assume that OpenAI drove this deal, but it could be that Axel Springer did, or it was roughly equal. But OpenAI has a vulnerability, which is that it trains its models on content, publicly available content for sure, but it does train on content. And there is an underlying narrative, which I believe to be wrong, that says they're stealing copyright. We hope it's publicly available, and we hope it's legal. We don't know for sure, though, do we? Well, they assure us that that's true, and I've got no reason to disbelieve them. Right, you have no reason to believe them either. They're not going to admit that they're illegally taking content. No, but it'd be easy to prove. If they're doing logins and passwords, it would be super easy. Well, it isn't that easy. They'd have to open their black box. They'd have to show exactly how their chat GPT LLM is learning. Well, that's where your technical knowledge is failing you. On the receiving end, Axel Springer would see a crawler login using a password and gain access to content that requires a login, and they would be able to trace that. No, no, you've got my... Leaving aside the Axel Springer piece, which I want to come to, my point is that we're still not entirely clear, to put it mildly, of how chat GPT is learning and whether all the content that they're learning off these guardrails, these learning systems, that they're getting their hands on this stuff legally. There's all sorts of... No, let me repeat what I said then in a different way. You can see, if you host content that is not publicly available, you can see when someone accesses it in your log files. So you would know, and you'd out... They would have to see their log files. No, but what I'm saying is the offended party, let's say it's the New York Times, they would say, our content's being stolen, here's the proof. And no one said that.

Speaker

You may... I'm not technically proficient enough to call you on that. That sounds all too easy. But anyway, go on. So we've got this open... The Axel Springer, yeah. The Axel Springer thing. It's fantastic. It's basically Axel Springer, which is a major publisher, saying, you can use our stuff to train on and we'll do a commercial deal. They're obviously paying them something. And that was... Who's paying who something? OpenAI will be paying Axel Springer. For access to all their information. But why do they need that? They're opening up their... Well, because Axel... Do you know Axel Springer or their content? Because Axel Springer sees the future and realizes that if you don't play, you're going to be a net loser. Because these systems don't need... They're not like a web So by definition, you only need enough content for the training to be successful. You'll see there's another article, by the way, implying from Michael Parekh saying that the LLMs are getting smaller and smaller whilst getting cleverer and cleverer. Yeah, that was an interesting piece. So presumably they can create... Sorry, that they can create small LLMs or SLMs, specifically what, for Axel Springer content? Is that how it would work? Well, anyone can create small LLMs that are really good at a discrete set of things. And that probably represents a big part of the future. So Axel Springer is basically saying we want to be in this party. Plus, we can make some money by being in this party. Whereas, for example, the publishers that are complaining about Google, at this stage, at least, are resisting fully jumping into the pool with both feet. And if the future truly is no websites and no traffic, that's a foolish decision.

Speaker

But the future, if the future is no websites and no traffic, there's still a future for journalism and for content creators. In some ways, there's more of a future. We go back to the pre-web 2.0 period, don't we? Well, how will it be consumed if the average individual, and I'm thinking about my children here, consumes content no longer on social media, no longer on websites, but through an AI interface that knows everything? You know, that's a very different world. Well, let's use an example. We've got the Spurs-Forest game coming up. That's why we have to end at noon Pacific. I want to know the Spurs team today. Normally, I would just do a web search and find the team. But I would be able to ask OpenAI or Google, Gemini, all these and say, tell me the Spurs team for Forest today. How's that going to work in terms of, does advertising go away? So I'm just going to get a free information. Who's going to make money out of that? And the economics change dramatically, leaving aside the technology. Well, we've always said that the value of content goes to zero. It's like in music. The bands make most of their money from live performance now. So clever people who want to make money probably are not going to make it from writing content. But you're skirting the issue. Who's going to make money out of that? Normally, I would go to Google, I would find out, and I'd have to see some advertising to do it. That's the price, so to speak, of Google. What's the price on OpenAI? Well, let's test something.

Speaker

Hey, can you tell me the Tottenham Hotspur team that's playing Nottingham Forest today? You'll have to do a web crawl.

Speaker

Here it goes.

Speaker

The Tottenham Hotspur team playing against Nottingham Forest today has been confirmed. Anj Pastakoglou, the Spurs manager, has named an unchanged starting... I'll stop it because it's going to give the answer. So it gave us the answer, but no one's making any money out of that. I mean, I guess you do pay $20 for... I pay $20 to chat GPT a month, and we'll continue to do that. And I'll probably pay more if they have enhanced services. So the information about the Tottenham team is worthless.

Speaker

Well, it was worth... It did have value in the Google search engine days, in the sense that I would be willing to see some ads in exchange for knowing the team. Well, that goes away. If there's no traffic, there's no ads. But there is, in a sense, traffic, Keith. I mean, there's... Well, not really. I am traffic, and I'm asking the search engine. Yeah, but no advertiser is going to show an ad to chat GPT. So basically, the way it's going to work... And by the way, this is why Google and the media companies, which that article gives the impression they're adversaries in some way. Actually, if you read the article, the Google one, they're not adversaries. Google's working very hard with the media companies to try to figure out how traffic and web search survives. Well, but the thing with Google is that they're hedging their bets because half the company is dedicated to search and half to AI. So I don't see how they lose on this. Well, they lose if the web search people can't retain user usage of the interface. Well, not if they configure out an AI business because... The AI business, between you and me, you think about it. Chat GPT is so far ahead and doesn't have ads. Would you go to a worse AI? Well, not everyone, and Keith agrees on that. By the way, there was another interesting piece you said in the news. It was a particularly good newsletter this week, last one of the year, by Eric Hull. Excuse me, but the industry's AI is disrupting and not lucrative. Would that include the media business? Is that what he was saying? I think he's saying that, yeah. There's a good riposte by Rohit Krishnan, who disagrees with him strongly. I put both in because actually they're both long reads. These are Christmas reads. They're both very long essays and very thought through. But he's saying that the truth is that really valuable content isn't accessible to Chat GPT and you're still going to need to pay for it. The things it's disrupting, I think he's completely wrong, by the way. It's going to take some time. It's not going to happen overnight. Meanwhile, the issue of creatives and AI is enormously hot. Hamish, I was going to call him Hamish Substack. Hamish McKenzie, the founder and CEO of Substack, suggests that the AI revolution is an opportunity for writers, the humankind. He always writes these things. Any new technology is good for creatives as long as they're on Substack. Is it right, though, do you think? Could AI be good for Substack writers or online writers?

Speaker

Well, it could be, but he's not right in the way he's thinking about it. What he's saying is, okay, the content is going to become worthless, but the human cultural aspects of writing and reading are going to become more important. It's a little bit like saying bands will do concerts. He's basically saying creators will meet each other and their readers in the real world, and that that is where the value will be. And I think there's some truth in that, although the forms of that are yet to be figured out. And for Hamish McKenzie, all roads, for better or worse, lead to Substack. So that's given. What about photography, Keith? You have another interesting piece. Everything seems to be changing. Why Vision Pro, the new Apple virtual reality thing, will change photography. How's that going to change everything? So this is written by Om Malik, who is... You're a big Om fan. Well, he's a friend, and I like his thinking process. And he's basically a camera guy. His best favorite cameras are made by Leica, and with super expensive cameras and lenses. And he takes fantastic photographs. And this week, he looked at iPhone photos, iPhone videos, and his Leica photos on an Apple Vision prototype. And he's reporting back saying, this completely changes photography because you get completely subsumed in the photograph down to a very detailed level, especially good photographs with good cameras. And he makes the point that the outcome is the opposite of what he expected. He thought the outcome would be that the iPhone is good enough to replace his cameras. But his Leica photos were so good looked at on the Vision Pro that the opposite he thinks is true. He's going to have to keep buying expensive cameras because the impact, once you look at it in 3D on one of these things, is so fabulous that you don't want to give up on... It's interesting, just as the Google AI might be cannibalizing their search. So what you're saying or what Om is saying is that the new Vision Pro might be cannibalizing the quality of the Apple photography because of the Leica or other high-end stuff. Om is still around. Another person who has reappeared is Larry Lessig, my old sparring friend, enemy, whatever you want to call him. He has come out saying how he learned to stop worrying and love AI. What is it about AI that Larry Lessig has learned to love? Well, so Larry is a lawyer at Stanford. Actually at Harvard. And he got thrown out of Stanford. Did he? Okay. Where is he now? Oh, he's at Harvard. Okay. Well, that's not too bad, is it? Getting thrown out of Stanford and then... I don't think he was thrown out. But anyway, he's not at Stanford anymore. So Larry has always been an advocate for democratization of various things. And in this article, he explains how his subjective initial negative feelings about AI were completely transformed when he argued a case with ChatGPT. And ChatGPT understood his logic and reaffirmed that in fact, Larry is right. On RFK assassination, he's really taking on the new loony RFK Junior. Yeah. So he basically came to the point of view that ChatGPT is a better lawyer than most lawyers and a better judge than most judges. And he completely changed his view. I'm not sure whether that's something to be celebrated. What happens to judges and lawyers and journalists? I'm not sure if Larry thought of that. I'm not sure he really cares because his experience of judges and lawyers in particular is not healthily positive. Law professors like Larry Lessig out of work too. That would be entertaining. Meanwhile, we've talked a lot about Google. You have a piece from the information about Wall Street yawning at Google's monopoly threat. The Google stock seems to have held up reasonably well. Where are Google in this economy? You think OpenAI is miles ahead. Not everyone agrees, Keith. Well, I think you'd find it hard to find someone who disagrees, actually. Actually, there have been lots of pieces suggesting that Gemini or the new version of Gemini competes quite well with OpenAI. Anyway, that's another issue. Yeah, I've used it and I beg to differ. I don't think it's as good, but there you go. So where is Google? Google has been on a constant transition from web search to whatever the future holds, really since 2007 when the iPhone came out. You could argue that Android is part of that. And the finding this week was that the Android app store is a monopoly because Google has used it to the detriment of Epic. Yeah, they lost that big case. They lost a big jury trial. I'm guessing they don't have very good lawyers at Google. Yeah. Maybe they should replace some of their lawyers, Keith. We won't go there. But losing a jury trial is kind of significant because it means you've lost the audience. But Apple also kind of implicitly lost that Epic trial in the sense that it was, in a way, a judgment against both the Apple and the Google store business model. That's not my reading. My reading is that Apple, because the iPhone is not, unlike Android, which is an open platform with lots of developers and phone manufacturers like Samsung, the app store that Google owns is more of a problem. Because Apple is the sole owner of the iPhone and also the sole owner of the app store, that's less of a problem because it's the app store for the phone that they own. So I think Apple's on solid ground at not being a monopoly. It is a closed environment, but that's not the same as being a monopoly. We are speaking with Keith Teer, CEO of SignalRank and the author of the excellent weekly newsletter, That Was The Week. You should all subscribe. This week, he's showing off his amateurish AIR once again, talking about Google. Are you a buy or sell on Google? It seems like you could make either case or you're simultaneously baking both cases, Keith. Should Google be worried about this future or should they embrace it? If you look back at my writings on TechCrunch from even 10 years ago, I've been a sell on Google ever since mobile came along. Some people might say, well, every clock's right twice a day. Yeah, but if you'd have put your money into Apple in 2013... Yeah, but that's true of anything. Yeah, but in 2013, no one knew that, but I did.