All right. Good afternoon, everybody. Are you all having a good time? Oh, fantastic. Amazing. Are you ready for Metallica? Are you ready for Benson Boone? Raise your hand if you're going to Metallica. Awesome. Who's looking forward to Benson Boone? All right. Well, so we got a mix. I don't know. Please welcome Dario.
Hey, Mark. Thanks for having me.
Great to have you, Dario. Welcome. We're really excited to have you and introduce you to the audience. We've had such a great partnership. We've had an opportunity to speak so many times, work with some customers together, and build some technology together. It's been exciting to use your model inside Slack with our new Slackbot. But as I've talked to so many of my friends about you, a lot of people don't know you. So maybe you could tell folks a little about yourself, your beginnings, where you're from, and I'll ask you a couple of core questions along the way, but take it from the start.
Yeah, absolutely. I grew up right here in San Francisco. I was originally interested in science, not in tech per se. I did my graduate work in neuroscience. I was a postdoc in computational biology at Stanford. As I was working hard on solving the problems of biology, trying to understand diseases like cancer, one of the things I noticed is that the complexity of biology felt like it was beyond human scale. I was starting to despair that we would be able to make progress at any speed because if you've ever seen one of these diagrams of human metabolism, it's just beyond human scale. But almost exactly at the time that I was wondering about this, out of Google and the University of Toronto came the first work on deep neural networks. I looked at it, I replicated some of the results myself, and I said, 'My god, this stuff really works. This stuff is really going to go there.' So I went into AI. I ended up working at Google. I joined OpenAI in the early days. I led research there for several years. I was one of the people who documented this trend of exponential scaling of the technology. Myself and the folks who eventually became the Anthropic co-founders were among the first to study the scaling of the technology. Around 2019 or 2020, I had two realizations. One was that AI is really going to scale to the moon and reshape the economy. It'll take a few years, but it's going to reshape the economy. And the second was that these systems are different from ordinary software. You don't write every line of code. Building ordinary software is like building a skyscraper. You make the blueprint, you make everything to design. This is ironically a bit more like biology. It's organic. You're growing these models. You set a high-level recipe. It's like growing a plant or baking a cake. That unpredictability combined with the incredible economic impact means that safety and responsibility are going to be very important to make sure the models behave predictably as we deploy them and they have enormous impact on the world. We need to make sure that we deploy them responsibly. Those were the two pillars on which myself and the seven co-founders who left OpenAI to found Anthropic generally based our view. The rest is history. Anthropic has been around for four and a half years. We're growing very quickly. By revenue, we have at least a plurality, perhaps the majority, of enterprise customers and enterprise revenue, while most of the other players are focused on the consumer side. As you would know from running an enterprise-focused company, these values of trust and responsibility are synergistic with serving the enterprise. These values help us serve the enterprise, and having a business model that serves the enterprise helps reinforce our values.
One of the things, and you went pretty fast into starting the company and shot through Google and OpenAI to Anthropic, but you didn't mention your sister. That was one of the things that first really shocked me when we started to get to know you was that you and your sister together are working on this company. Can you tell us the origin story?
Yeah. There are seven co-founders, and me and Daniela are two of them. As I mentioned, we both grew up in San Francisco. We went to Lowell High School. Even very early on, we figured that we wanted to do something good for the world, to have some kind of grand project, and to work together. I think many siblings have these ambitions, but somehow things found their way to us actually being able to do it. She joined me at OpenAI in 2018, and then the two of us, along with some folks we worked with, many of whom had been longtime friends, decided to start Anthropic. Me and Daniela essentially run the company together. She's responsible for the day-to-day running of the company, how the company is put together, how the org runs, its processes, and a fair amount of the business on the structural side. I couldn't run the company without her. I tend to think about the strategic side, where the research is going, our plans as a company strategically, how we think about safety, the company's role in the world, and the vision. I find that's a very good split because it's really hard to switch between an operational mode and a vision mode. Both people can be more effective if you split.
Thanks for allowing me to come back to that because I think it's such a great part of the company and the culture and how you're working. Last year she was here and now this year you're here. We've known you all from the beginning. I think Salesforce owns a little point or something of Anthropic, so we're very happy to be an early investor. I think you invested when the valuation was $4 billion. Is that correct?
I don't know, but it was too high. Whatever it was.
Before you say it's too high, you should calculate how many times it's gone up since then. We made money so far.
Well, I'm thrilled obviously that we're part of your journey, and we love working with you and Daniela.
Getting back to this story of you and your sister and your friends starting this company here in San Francisco, you came to us when we were starting to consolidate some of our real estate here. We had a conversation about you taking over the Slack building, and now you're in the Slack building, right on the park with us, looking at the tower from your window.
I believe my office is what I was told was formerly Stewart's office.
Formerly Stewart's office. Very good. It just occurred to me, here you are. You're the San Franciscan. You're right on the park with us. You're right in the tower with us. We're an investor in the company. It's obviously a critical part of Slack today in terms of how you couldn't use Slack without it, but you're at the very bleeding edge of these models. It's really incredible. You've been visionary right from the very beginning. I only met you after you started the company. Take us through your expectation. What has been your biggest surprise since it's only been four and a half, five years since you walked out the door, said goodbye to Sam, and said, 'Okay, Daniela, we're going to get this going.' What was your biggest shock?
There were so many moments because the last four and a half years were just so surprising overall. At a high level, there's a dichotomy where on one hand, everything we've seen over the last four and a half years was in some sense predicted by those straight lines on graphs that I made with the scaling laws. These enormous data center buildouts, the way our revenue is going up 10x per year, the use across the economy, you can see it all in these scribbles on the page. But actually seeing it play out is wild. A lot of the individual things I'm seeing happen, I never would have expected. If I were to pick one most surprising thing, it is that we left one company and started another. Over time, as you build a company, you see through the people you hire and the choices the company makes, you see the paths diverge and you see differentiation. OpenAI has gone very much in the consumer direction. We've gone very much in the enterprise direction. The way we approach things is very different. So you have this market structure where there are these two companies that are like a natural experiment. They're taking a very different approach. Both exist in the ecosystem, but the approach they take is very different.
Then you have to reflect back at your time at Google as well. They're also very much with Gemini, a top model as well. How do you look at them? Do you look at them as a blend?
Google's a partner to us. They're an investor in us. Obviously, their models compete with us at the same time. But I think this co-opetition dynamic is becoming more and more common. As we have all these layers, the chip layer, the cloud layer, the model layer, the go-to-market and distribution layer, various companies are involved in more than one of these layers. It's not uncommon for companies to cooperate in one area and compete in another. It's fairly normal. It's also been helped by the fact that I know quite well some of the folks behind Google. I've been friends with Demis Hassabis, who leads research at Google, for something like 15 years. He was maybe one of the only people who was earlier than me in seeing the potential of the technology. He also has a neuroscience background like me. I think that helps us to get along even as there are some competitive aspects between the companies.
What was your reaction when he won the Nobel Prize this year?
I was very happy for Demis. I was writing this essay about the potential of AI in biology called 'Machines of Loving Grace,' and two days before I was about to publish it, Demis and his colleagues won the Nobel Prize for AlphaFold. I ended up dedicating the essay to the 2024 Nobel Prize in Chemistry for AlphaFold because it was an example of the future that I was talking about.
Very exciting. Getting back to you and Daniela working on and growing the company, and you're doing it right here in San Francisco. What are some of the characteristics and the models that you're looking at now that are most surprising? What's really exciting about the next generations of these models unfolding?
Broadly, it's the jump to being able to do agentic abilities. Being able to do things.
What was it? I didn't hear the word.
Agentic abilities. Turning the models into agents that can...
Oh, I see. You've been... I don't worry about it, Dario. I just wanted... I'm a little weak in my right ear.
I genuinely thought you didn't hear.
I see. I understand. It was very funny, wasn't it? Definitely a buzzword, and you've perhaps contributed more than most to making it a buzzword.
But I think it's a real thing, which is being able to do tasks end to end in a way that was not possible before. We see it more in code than in any other area, but code is just a leading indicator. In tasks like financial services, health and medical, insurance, manufacturing, and many of the regulated industries we're working together on, agents are just as valuable. You want to do things end to end. Think of sales, go-to-market, and customer management workflows. You want to string a bunch of things together and have the models pull in context, learn, do, access information. It's just as valuable. Coders are early adopters, so they've given us a glimpse of the future. We're going to see a similar spike in other areas.
I want to ask you a few questions in that area. When we first started getting in, probably one of our very first conversations, we were starting to talk about this theory and AGI, especially the potential in the future. AGI at that time, several years ago, was being defined as the model is almost recursive, building itself, growing, evolving, coding itself, enhancing, and off and going. We had Brett Adcock on the stage earlier today, and he was talking about a scene out of 'We Are Legion' where the robot had taken off and was self-replicating into the future with an infinite number of robots. It was incredibly compelling. To that point a few years ago, it was like the model is building itself, growing itself. It didn't turn out exactly like that, though, or we're not at that point yet, are we?
I would separate two things. Inevitably, as we talk about AI in the future, there's this sci-fi sheen on the whole thing. In a way, it's helpful as a metaphor, but it can also be distracting because it conjures everyone's mind into a future dreamy mode. But there are some things that are real. One thing that's real is that every three months we make a model that's smarter than the model we made before. The models are slowly getting to where they're better than most humans at most things, maybe at some point better than all humans at all things. The other thing I think is real is that the model is not building itself in the sense that I have a specific copy of the model that's making a specific copy of the next model. We don't have self-replicating nanobots or robots. But what absolutely is happening is that within Anthropic and elsewhere, we are using Claude to help build products on top of Claude, to help train the next Claude, and to help serve Claude faster. Six months ago, I made this prediction that in six months, 90% of code would be written by AI models. Some people think that prediction is wrong, but within Anthropic and a number of companies we work with, that is absolutely true.
So you're saying that 90% of all code at Anthropic is being written by the model today? Is that something you've provided something auditable around or shown how that's happened?
The teams play an editing and supervisory role. You can say 'write this feature' and Claude will do it, and then the human looks over the code or asks for help from another copy of Claude. The other thing is sometimes we have long-running bugs. When we were training our most recently released Claude, the cluster was broken. There was this bug. Engineers were spending a few days on it. We just went to Claude and said, 'Just play around with the cluster and see if you can figure out what's wrong.' Claude found this super obscure bug that the engineers had missed. It was like having a very talented teammate.
If Claude is now writing 90% of all the code at Anthropic, do you need less engineers? When will it be 95% or 99%?
People can misinterpret it as Claude writing 90% of the code means we're firing 90% of the software engineers. But if you're familiar with the principle of comparative advantage...
You have made some very aggressive comments around firing executives, so I just want to put two and two together to get four. One thing I've learned in knowing you for so long is I have to come back and get that qualifying question again because you're going fast. I'm trying to keep up with you. I wanted to make sure I understand exactly what you're saying.
Let me take them one by one. Under comparative advantage, if Claude is writing 90% of the code, what that means usually is you need just as many software engineers, you might need more, because they can then be more leveraged. They can focus on the 10% that's editing the code or writing the hardest part or supervising a group of AI models. You can end up being 10 times more productive because the remaining 10% is now a complement to the AI system that's getting much better at what it's producing.
So it's a rebalancing of the work, not a replacement.
It's not a replacement. It's a rebalancing.
