The future of content with AI technology with Nicholas Thompson

Hello, everyone, and welcome to This Week in NET. It's the November 22nd, 2024 edition, and this is a special episode with Nicholas Thompson, the CEO of The Atlantic and former Editor-in-Chief of Wired. I'm your host, João Tomé, speaking to you from Lisbon, Portugal, via the Internet. Last week I attended Web Summit, a massive international event with over 70,000 attendees right here in Lisbon, and I had the opportunity to speak with Nick Thompson. Here's my conversation with him. Hello, I'm Nicholas Thompson. I'm the CEO of The Atlantic. I'm based in Brooklyn, New York, and I'm from Boston, Massachusetts. Let's start in terms of conversation. You're at Web Summit. You typically look at technology in its larger scale, in a sense. What have you learned in these days at Web Summit? What has put some thought into your mind in terms of technology, in terms of the Internet? Well, I've mostly been talking about artificial intelligence, and the first conversation I had on stage is one that I followed up on with a number of people, and it was about the dynamics between open -source AI and safety. And there are a lot of people who have made the argument that open-source AI is less safe. You give everybody access to these incredibly powerful tools, so many things will go wrong. And the consensus here is in the other direction, that actually open-source AI, where you have more people looking at problems, you have more people trying to solve the riskiest use cases, maybe open -source AI is safer. And one of the more convincing arguments was, of all the bad things you've seen in AI, how many have come from open-source AI? And the answer is, not many. So the consensus is a little different here from maybe back in New York and Washington. Specifically around the Internet, there has been a lot of changes in general in the industry, new players coming around, especially in the AI area. But in the last two years, what are the main changes you've seen and surprised you the most around the Internet, around AI? Well, I think what's going to happen with the Internet is that the big large language model, the big foundation models, whether it's open-source, whether it's based on Lama, whether it's GPT-4 or Anthropic, are going to become kind of a new base layer for the Internet. And all kinds of stuff will go through them, and people will build things on top of them. But I think you will see kind of like five more companies or eight more companies that are sort of intermediaries between a lot of what you see online and yourself. Specifically around your industry, the media, one of the talks you did was around the media landscape with AI being the focus. You spoke about this specifically on the fact that before Facebook, Google were bringing traffic to news organizations, with AI that's not really the case. You're getting all the information right away. What are the main challenges in that relationship? I mean, the main challenge is almost existential, right? You can imagine right now the way the world works, the plurality of people who come to the Atlantic, they come from Google, right? They type in a search request and they come to the Atlantic. And you can imagine that completely breaking, right? You can imagine that Google no longer sends you out to links. It just answers your question. And so instead of getting Atlantic article, you just get it answered by Google, which has already read the Atlantic article. So then we don't get any new readers. You can also imagine the same thing happening on other apps. You could see our Facebook traffic diminishing. You could see our Twitter traffic diminishing. Then simultaneously, you can also see the ease at which people can create new websites using AI, making it so that the Internet is, I call it the inshittification process. There's so much stuff on the Internet that you can't find the serious stuff. So you can imagine both fewer people are able to navigate to serious journalism, and there's so much other stuff distracting you. So if that happens, it's not great for our business model. And so we have recognized for a couple of years that this is a process ongoing, and we've tried to harden our business model against it, build direct relationships with our readers, figure out how to get them into our apps, into certain ecosystems, to read our print magazine, to get newsletters, to have direct social relations. So figuring out how we can survive in a world without normal Google, Facebook, Twitter, et cetera, traffic to theatlantic.com is an important challenge. Specifically on the media landscape, we've seen crisis at even CNN, big networks in terms of potential layoffs, funding, it's not there. Democracy depends a lot on good information. Where do you see this good information going for a good democracy going? With the social media aspect, the changes that occur there, but also the media, the traditional media, not being so much trustworthy. There are studies in the US regarding that. It's a crisis six ways from Sunday. The trust in traditional media has declined for all kinds of reasons, partly from errors made by the media, and partly because people who don't want us reporting on them have run a huge, massively funded campaign to discredit us. And so we have to build that back in every way we can. We have to win that battle. We have to continue to be completely dedicated to being as accurate, as fair, and as trustworthy as possible. That's extremely important. When it comes to AI, you can imagine AI being a tool to increase accuracy and veracity and truth. And there's some studies that have come out recently about the ability of AI to get people out of conspiracy theories and bubbles. And so you can imagine a future one where people come out of filter bubbles and use AI to create their own facts that support their magic reality, which is a little bit the way some of us live now in America. Or you can imagine a world where AI counters that and where you have trusted news organizations, trusted individuals who can say what they believe is true to the best of their abilities and the best of their research. And you can have AI back things up and you can have people, if they believe something that is factually inaccurate, go to AI to have their opinions sorted out. And you can imagine people who are in a left-wing filter bubble learning new things. And you can imagine people who are in a right-wing filter bubble learning new things. And maybe, maybe, maybe, maybe we can have more productive discourse in America in the years to come. There's some of the discussion on news organizations are still important. Even the information they do goes online, goes on social media. So they are a part of the, even today, the information ecosystem, a relevant one, even if people consume them on social media. Where do you see the traditional media keeping, keep it going in terms of having funding, maybe millionaires helping or others? Yeah, there are lots of business models for traditional media. So some of whom are supported by your billionaire benefactors, like affluent people have always bought publications and have been willing to lose money on them in order to keep them alive or to get their opinions out. I don't like that model. I like capitalism and I like media companies that make money. And so I've worked really hard to do what I can to make the Atlantic profitable. We obviously are owned by a very wealthy woman, but our goal is to, we are now profitable to sustain that as long as possible. So we are not beholden to anybody. I want to run it like a real business. So then the question is, how will you make money? We make our money, a lot of it through subscriptions, right? Figuring out how to have direct relationships. People pay us some of it from advertising. How do you create a place where brands can get real benefit by appearing next to our content, whether in print or in digital. And then some through other sources of revenue, streams, licensing, events, affiliate. So every journalism company will have to figure out a business model and it will be different because the way we used to make money has been totally disrupted and it will be further disrupted. Regarding the Internet, you work at Wired, you work with a lot of different tech companies in terms of reporting on them, seeing the evolution of the past few years. For those who look at the Internet as this tool, what is the main thing about the Internet most people don't realize, but they should? The main thing about the Internet they don't realize? I don't think people realize quite how complex, maybe it's because I'm talking to someone from Cloudflare, quite how complex and interesting the process is by which information gets to you and the number of different players along the stack from your CDN to your service provider, whether it's whatever company is providing your ISP. And I don't think they realize the influence that those different companies have and the decisions they can make. And I think you're going to see some interesting ways and I have no idea how that plays out in the years to come. Are you confident in terms of the Internet, in terms of AI, what's to come? Are you optimist on that regard specifically? So net-net, I am definitely optimistic that AI will make the world better, at least in the short term. I think in the next few years, there will be churn, there will be change, but I think that it will make us more productive, it will solve problems, it will make us smarter, it will make us more interesting, it will help us learn things. We'll learn new languages, we'll learn all kinds of incredible things. AI will be great at that. The thing I worry about is that the pace of change, not now, it's not really disrupting too many things right now, but in a few years, the pace of change will be more than society and its social safety nets can handle. So I worry about that, but I think I am largely optimistic. Specifically about Cloudflare, you've seen the company grow for a few years. Do you have an opinion in terms of evolution of what you saw, a story related to Cloudflare you want to share? About Cloudflare? I mean, look, I think that the work of a CDN and the work of a company that protects you from bad actors, particularly as the Internet gets more complex, and that makes sure that accurate data gets to the right place. Cloudflare can't make sure that what you write is true and accurate, but it can make sure that the flow of information is true and just, and that's absolutely essential. I'm curious to hear specifically on the landscape of what's to come. Social media, there are studies on this. In terms of kids not having access really young to social media, I want to hear your thoughts on that. But also, do you think AI could be more positive from the get-go, especially chatbots, compared to social media? That's a hard question because AI could be worse than social media, could be better. There's no question that the worst thing about social media is what it did to our children, right? And it made people insecure. It pulled them away from their friends. It pulled them from being outside. It pulled them from many of the learning experiences that they have, right? If the average kid is spending four hours a day on TikTok, what are they not doing? Are they not climbing trees? Are they not talking to people? It's been terrible, right? Social media is terrible for our children. Now, AI could make it worse, right? It could just further addict children. You can imagine AI being embedded in stuffed animals, and the stuffed animals become manipulative. Or you can imagine it being helpful and a way for people to learn and explore. But I really feel that children should be off devices as much as possible. And so, my hope for my children, and I'm certainly an imperfect parent of this, my hope would be they spend as much time without their devices as possible, while also being excited and learning everything they can about AI, because it's going to be important to the future. I'm curious also regarding your use of AI. Do you use chatbots typically, even to write, to prepare for conferences, things like that? What's your use? I use it all the time. I probably do 50 AI queries a day across all the different products, testing them out. I use them. I have interviews today on stage, and I actually had Anthropic write what it expected the replies would be of the various people I was going to be interviewing, which was helpful for coming up with questions. I certainly test out my ideas on them. I use them for search in all kinds of ways. Basically, just ways of getting information more quickly and being smarter about it. Do you feel it got better in the last year or so? Oh, it gets better every month. There's no question. And I'm writing a book about running. I would never use a chatbot to write because I think it's unethical and unfair, but I use it for research. And what it can do to help me go through my notes is better last week than a month ago. You don't use it to write, but do you use to gram checking, to things like that? I actually don't think it's very good at grammar checking. I don't think it's good at fact checking. What it is good at is, hey, here's something I wrote. Here are some notes I took. Please analyze the notes and tell me whether I've actually represented it. But it's a crappy copy editor. It's a horrific fact checker, but it's a very smart thought partner. I like to think of AI as, is it the best available human you have for whatever task you have? And so for some tasks, for copy editing, no. Fact checking, absolutely not. But one of the questions I was recently asking, help me understand the optimal ketone intake in an ultramarathon. I don't have anybody I can ask who could tell me that. Nobody. Maybe some ultrarunner friends I could call up, but they're hard to get on the phone. It is much better to say, what is your answer? And it'll be like 90% accurate, which is good enough. I'm also curious on this area that you do, you communicate a lot about technology. Yeah. You use different tools. You do the most interesting thing in tech. What have you learned doing those types of things? What has changed in the past few years? And why do you feel it's relevant? Well, for my most interesting thing in tech series, I use AI all the time where I will say, hey, I'm trying to come up with a topic. Here are some things I'm thinking about. Which of these do you think would be most compelling in a video? Or I'll upload a paper and I'll say, hey, here's a paper I really like. I don't understand this thing about it. Like help me understand that. I mean, if you went through my anthropic search history, you would find all kinds of most interesting thing in tech. You just use Anthropic? No, I use them all, but I happen to use Anthropic for analyzing papers. I think it's, whatever reason, I think it's the best. I use Anthropic. I use GPT-4. I use Perplexity. I use Llama. I use Gemini. I use all of them. And I just happen to use Anthropic for that particular use case. I think that the capacity, one of the things that AI is really good at is synthesizing and understanding information. And I use Notebook LM actually. If I'm going to like going in transit, I might upload a paper, have it create a podcast. And then I listened to the podcast about the paper. So the ability for me to learn things quickly, like what are the things about my job? I have to learn a lot about a lot of different subjects very quickly, right? Because I spend most of my time doing my job, right? Like managing people and figuring out what strategy to be on this. But then I have this little side thing, like trying to analyze technology. And without AI, it would be a lot harder. Regarding those types of segments you do externally, what has changed in the way you do it? Not AI related, but even the way you communicate things, the way things, communication resonates, technology communication resonates with people. It has been NetQuazit for you, I would say. What would be the evolution of that area of doing that? Well, I mean, the interesting thing is like, so take my most interesting thing in tech series. I do it the same as I've done it for however long I've done it, six or seven years. How long until somebody can create a synthetic Nick who can do it all using AI and be better than me, right? They can't do it right now, but like next year, right? Will I have a competitor who is actually better because they can read everything infinitely quickly. And they can be more handsome than me because you have the AI avatar and they can be more compelling and they can have better backgrounds. Like it might happen. So now, I don't know. I don't know. Will they have the same relationship with their viewers? Maybe not, but like I might be wiped out by an AI next year. That'd be a big change, wouldn't it? That's kind of interesting because I remember I was a tech journalist and one of the things I saw was music being played by robots. And it was amazing. Music was really wonderful, but no one cared because it was robots. It was not like an orchestra. It was this element of humans doing something that was engaging. The sound was perfect, amazing, but it wasn't humans. So do you feel that we're not as replaceable as... Well, I think that if people know something is AI generated, they will... The first one who's AI generated will have an advantage. People will be like, oh, that's cool. And after that, everybody will be like, yuck, right? But the interesting question will be when one comes along and you can't tell, right? And they follow it and they're like, that's better than Nick. About the Internet, is there a wishlist about things that are related to the Internet? Maybe it'd be infrastructure. Yeah, there is a wishlist. Or even... Here's my wishlist. This is my wishlist. I would like content creators, whether they are media companies, individuals, authors, to be able to stop the large language models from scraping their sites. And I would like a little button I can press. Right now, I can put a little thing in robots.txt and it says, hey, please don't scrape my site. Please don't take my content. Please don't use it in your models. If you are, please pay me. And they run right through that if they want it. It's like a little sign in the yard that says, hey, please don't come by. And I would like there to be a button I could press that would just let everybody I want see it and everybody I don't not be able to scrape it. Do you think that will help even the future media business? Totally. Because it will shift the power balance. Right now, the power balance is entirely with the AI companies who can scrape all the content they want. And the media companies and the content creators who are like, wait, wait, wait, we don't want this. We don't want this. We don't want you doing this. We don't want you building competitive products off of our data. And if you are, you better pay us. We're happy to make a deal with you, but you should not take this. And to some degree, the crime has already been committed, right? The AI companies have already scraped the whole Internet. They think they have a legal right to it. Maybe they do. Maybe they don't. But I would like the ability to block them because it is my stuff. It is our stuff and it's the content creator's stuff. Golfer launched like last month. I know. That's why I mentioned this. The AI audits that does that. And I would like everybody to be able to use whatever your CDN is or whatever layer in the stack is. I would like it to be easy for every content creator. I would like anybody who has their own content to be able to say, I want to license it or I don't want to license it. Makes sense. Last but not least, you run a lot. You actually do several of your segments when you're running. How is running important for you, for the way you communicate things, even for your brain? And I'll give you like a small sip. There's this talk about Peter Levels that is a known entrepreneur, solo entrepreneur. He talks about lifting weights and if you lift weights, your brain has advantages. If you run, your brain has advantages. In what way running has been helping and can give us the way you approach running? When do you run? Things like that. So running is like meditation for me. It's a break. I get up, I make breakfast, I hang out with my children, I run to work. I try to turn my mind off, you know, and I like have a chance to be outside and disconnect. I work, I run home. So it creates like space in the day for meditation and thought. I often have my best thoughts while I'm running. Sometimes I listen to podcasts. A lot of times I'm just engaging with the world around me. I think that the discipline you get from intense running training helps with the discipline and work. I think there's a lot you learn from a deliberate practice of focused running that translates into the rest of your life. So I believe that running is an essential part of the life I lead. It also distracts me, takes time away, and maybe I'd do better at everything if I didn't. As for my communication skills, part of the reason I film those videos when I run is that that's just often where an idea comes to me, right? And my, I like to, my videos, I like to just do when the idea, the idea. I don't like to wait. I don't plan. I don't script. And so if I'm running and the idea comes, I'll just put in my AirPods, put my phone up on a lamppost and record. I'm curious on this element. You have, you talk about your kids often with examples regarding tech. Thinking of your children in terms of technology, what would be the things that puts you energized and optimized in terms of my kids will have it rate with a good technology perspective? And also what worries you? Well, I think that they, you know, they haven't used it as much as they could, but I think they're using it to learn. Like I've seen my children, all of them have used chat GPT to understand things better, right? Like what was Germany's contribution to the start of world war one, right? Like in chat GPT can answer better than I can. You know, they've, we've used it. I've studied Spanish, like had chat GPT come up with Spanish programs for me and my middle son. So I think it's a great educational tutor and tool. And what worries me is just the, the, the real world is so much richer than the digital world. And the digital world can be so much more addictive and manipulative. And so just trying to keep them in the real world as much as I can. But like in five, 10 years, where technology is going, are you more optimist for them or? I'm optimistic for them. Yeah. I think, I think the benefits will be greater than the cost. And I think they have the right self-control and self -knowledge to be able to not get wrecked by it. You speak a lot about technology, things that people are not aware of. Yeah. In terms of the next few years, what is regarding AI or the Internet, things that excites you, but you think are not that much talked about? Um, I think we haven't really focused on AI and robotics and what it will mean to have machines that you can communicate with. I think that's coming quickly. Since Nick Thompson mentioned protecting content creators from AI crawlers, here's a short clip from a conversation we had on this program with our CEO, Matthew Prince, on the same topic. Our customers are content creators. Um, they are, uh, publishing information, publishing applications, making kind of their work available, uh, online. And what, if you think about the, the previous sort of era of the web was I think really defined as the search era, right? Where, and some of the original inspiration for Cloudflare was Google's crawler would create so much traffic. How could you sustain that if you, if you, if you were just getting crawled all the time? And yet, you know, Google's quid pro quo, what they gave you in exchange for, you know, crawling your web was they sent you traffic and you could do things with that traffic. You could monetize it through advertising. Sure. You could monetize it by, uh, you know, selling products. You could generate value that had nothing to do with money around just the ego of, you know, I built something people care about and want to see. And, and that was the, the, the promise of the search era of the web. As we move to, and it's, you know, it has been kind of a, it's not been as dramatic a transition as people, as people think, but as we move to a web, which is maybe a more AI driven web. And, you know, this started with Google itself, where you type a question and they give you the answer right on the page without sending you to the page. That changes the quid pro quo. And if you go to something like, you know, chat GPT or anthropic or perplexity, and you ask it a question, you get back an answer. And the nature of these LLMs is oftentimes they can't even like from a very technical perspective, they can't even credit back where they necessarily found that answer. In the same way that, you know, if one morning you wake up and you're really craving Thai food, you, you might not know exactly why you're craving Thai food, but somewhere in your brain, you know, the neurons fired in such a way. LLMs work in a really similar way. And that's created a real challenge for the web. Because if original content creators can't generate value from the content they're getting, if they don't get that traffic, so they can't sell ads, they can't sell products, they can't just even get recognition that they're the reason that the answer is correct, then why will they create content at all? And as we have talked with content creators, they've all said, you know, they've had a real fear around this. At the same time, as we've talked to the best leading AI companies, they've all said, you're absolutely right. That's a problem we need to solve. And you can see it where they're striking deals with the Reddits and Quoras and, and large companies in the world. What's been missing though, is how do you give people the visibility, the control, and then the marketplace, if you're a small content creator or a niche content creator, to say, I want to figure out how I'm going to give permission for these various AI systems to be able to use my content. And some people are just going to say, you can do whatever you want with it. Other people are going to say, you can do what you want with it. Just send me a t -shirt. Other people are going to say, here's the price list if you want to do something with it. And what we want to give people is the power to do that. Before we go, here are some great Cloudflare blog posts you should check out. Bigger and Better explores how DDoS attacks sizes have grown exponentially over the last decade. And you can see different metrics there. Do It Again goes into how we use durable objects to enable web socket support and authentication for AI gateway. So a bit something for developers. There's also the introduction of account-owned tokens for better access control. And also Xeraz Automation Actions. That's for streamlined event tracking. There's also Preventing DNS Conflicts that explains how we use formal verification to avoid logical contradictions in our authoritative DNS configuration. And a bit of PostQuantum. So that's PostQuantum signature candidates that reviews the latest developments in PostQuantum signature standardization efforts. That's it for this week. And that's a wrap. Thanks for watching.