I'm pleased to welcome Dario, co-founder and CEO of Anthropic. I really see us as a little bit as birds of a feather. We have a common background in doing PhDs in physics, but then turning to entrepreneurship to tackle the domain, hopefully in a disruptive way, but to do so in a way trying to have the biggest positive impact and reduce harms. You obviously focus on AI and I'm focused on space, but similar attitude. And I'm particularly pleased that we're here today because we're announcing our partnership together, which we'll get into here shortly. Welcome, Dario.
Yes, thank you for having me, and excited for the partnership.
So I want to start a little bit with AI applications. So everyone's talked about AGI, you know, it's the rage obviously, but you know, what are the biggest challenges you see in the world that AI can meaningfully advance, and what's Anthropic's strategy to advance them?
Yeah, so you know, I see a number of things. Like AI is a general-purpose technology, things ranging from biology, which I'm particularly excited about because, you know, I used to be a biophysicist. You know, I'm quite excited about the potential for things like healthcare, advancing science, and you know, in our collaboration with you, the ability to understand the Earth better for the sake of many cases that will have a number of scientific applications as well as improving the state of our planet and improving our ability to manage it. I think some of the challenges come — the key development of models itself, of course, is its own science, it takes a lot of work. But then when you actually want to apply the model, you know, creating the adapter between the fundamental model and the downstream application often requires companies to work together, right? It often requires partnerships with companies. We're trying to serve every industry with this general-purpose technology, and so when we partner with a company that understands some part of the world much better than we do, I think the important thing is to kind of work together to do something that we couldn't have done alone, because we understand AI but we don't necessarily understand all the implications of all our models or all the things it's good at or can be used at. So you know, I think at a high level that's one of the biggest challenges.
Talk a little bit about what's rate-limiting the development of AI in general today. There's, you know, there's compute, there's talent, there's data, there's algorithms. Where's the bottleneck?
Data is a really interesting one. You know, in terms of general data, there's a lot of it around, and you know, we've been able to basically scrape the entire internet. And there are some techniques like, you know, these new class of reasoning models that rely less on data. Where I think data is really important and valuable is the thing that I said before, which is when you're trying to do a task in some particular domain, getting the best performance out of that domain, data is going to be particularly important. There are some things that you just don't learn by reading the internet, right? And I think the partnership we're doing together is maybe a prime example of that, where, you know, Claude, our model, has been trained to analyze all kinds of images by looking at images on the internet, but the fraction of those images that are satellite data, that are geospatial data, is tiny. While you guys have an enormous amount of such data, it's the general model that we've trained — it's going to help, we're going to start from a much more advanced place. The models can already do great things out of the box, but there is much greater potential for the models to become better at this task that is somewhat similar but somewhat different from normal image analysis. So I think there's a lot of value in this specialized data to make these fine-tuned models for particular applications.
Yeah, 100%. And so let's move to space a little bit and those Earth applications in particular. So, you know, it's interesting, there is a revolution going on in space. People don't know about the lower-cost rockets, the increasing fleets of satellites, and most of that — you know, we've actually seen a thousandfold increase in cost per performance of satellites over the last five or ten years. So it's a bit like the mainframe computer to desktop computer sort of equivalent, and of course unleashing lots of applications. But the big upshot of it is new data sets about the Earth or new communication data around the Earth. One way or another, it's about data. We alone at Planet produce four million of these 47-megapixel images to scan the whole Earth every day. It's a lot of data. I think of how LLMs, as you say, have figured out how to semantically interface with the web, the data on the web. Analogously, you can imagine a semantic interface with all of these images. I call that sort of vision-queryable Earth or searchable Earth. What visions excite you at that interplay between AI and space? Is that one? Are there others that come to your mind?
Yeah, I mean, well, I think that one is pretty exciting, and I can put it in context a bit more, right? So in the space of text, there's been a lot of stuff that's been done in this kind of queryable or summarize or search, right? There are whole companies, some of whom — most of whom are customers, who, you know, have their whole business model around, you know, how can I find, summarize, search, sort through data better than I could before, better than base search engines can do it. But all of that has been text, and, you know, as you point out, in principle the same thing can be done just as well with images. But if I think about, you know, in the internet context, well, you know, of course you want to do internet search, it does come up, but actually I think it wasn't until a use case like this, right, satellite data in space, where it's really like you're kind of doing the same sort of queryable thing for images, that has been a core part of our business model for text for so long. You know, and I'm sure we'll have more advanced types of things, I'm sure we'll be able to analyze video, I'm sure there's much more value to be produced here. But this idea of searching through a large number of images for a particular thing — I'm not the expert on space and satellites, but it feels like this use case is perfectly made for that ability.
Absolutely. And you know, scanning for larger areas for changes that you might be interested in — sustainability or climate-related, or damage assessment after a storm — you know, they should be able to give a quick quantification of things, quick analysis of things, and we've already seen some interesting results there. Another area is predictive analytics. So one of the things I — you know, again intuitively, one thinks about how LLMs have extrapolated the next token. You know, here we are with 2,500 images for every point of the Earth's land mass. You know, perhaps it can guess the next token, i.e., the next image, which in this case could be where's the next flood, where's the next deforestation, where's the next refugee camp. Do you see that as a potential? I mean, where does predictive analytics go?
Yeah, yeah, I mean, I think that's pretty valuable as well. I mean, you can already ask the models out of the box today, like, you know, show them a bunch of images and ask them what they think is going to happen next, right? They don't even need to generate another image in many cases — they can just describe in words where they think things are going and what they think will happen next. So I think there's really a huge opportunity for that, and it has analogies to what we're doing in other fields, right? If I'm — you know, any scientific application tends to involve prediction. I'm very excited about the use of LLMs for science, and so that predictive activity is basically the same one. I'm excited about LLMs for medical use cases, right? And a lot of medicine is about diagnostics, prognostics, are about prediction. I think LLMs will find wide use in finance — that's definitely about prediction. If you can predict things more accurately, you can be very successful in the world of finance. So there's a lot of precedent for this to work well. We're still in early days of trying it out, but I'm just very optimistic based on the kinds of use cases that I've seen across industries.
Okay, let's come specifically to our partnership. So we're pleased today to be announcing this partnership between Planet and Anthropic. In particular, I mean, I've been so impressed out the gate how Claude was able to do a couple of things. One is program against our API, just make that simpler. And the second was analyze our imagery, or a stack of our images over time. I mean, you know, an example — we were looking in a forested area in the Amazon and showed it a bit of deforestation. You know, it could really accurately describe what was going on, what are the likely causes. It could reason a little bit about what's likely to happen next, what's likely to be the consequences, and even did some estimates of what the rate of deforestation, what's the land-use change, graph it out, because we asked it to. And I was pretty impressed with that. As you say, it hasn't seen much satellite imagery yet, so the big aspect of our partnership to start with is just to fine-tune some of the models onto Planet data, with Planet data, such that they are a little bit smarter about satellite imagery than they are before. What excites you about that partnership and where it might go next?
