Welcome back. AI and quantum tech firm SandboxAQ announcing this morning a partnership with Anthropic to integrate its large quantitative models with Claude. The companies are looking to boost access to drug discovery and material science research, among other major issues. Joining us now in a CNBC exclusive is SandboxAQ CEO, Jack Hidary. Welcome back, Jack.
Sarah, great to be here and break this news.
Yes, right on CNBC. Tell us about why this is such a big deal for you.
This is a big deal because, as we talked about before, LLMs are fantastic. Large language models like Claude and Gemini and others give you great access to AI for the digital world. You can make documents and images and videos, but how about AI for the real world? AI to make new medicine for cancer, AI for new medicine for Alzheimer's, for material science. That's really the domain of large quantitative models. Today is the fusion of the two. Today we're announcing that via Claude, one of the largest LLMs out there, you can access Sandbox LQMs: large quantitative models. So you're on your Claude, you're prompting, and now when you ask a technical question like biology, chemistry, physics, engineering...
Exactly. You can access our model. Instead of just...
So you now have access to all their enterprise clients.
Well, now basically, if you're on Claude, you can access all these quantitative models, and then we're hoping that large enterprises also start adopting it. But now, right from the Claude prompt, you can start accessing these models. It's very exciting.
And it's not exclusive. You could do this with others.
That's right. Hopefully you'll see some others also join in. LLMs converging now with LQMs, this is really the best of two worlds coming together.
Yeah, explain. I mean, you laid it out pretty well. The digital world and the real world. But what is, first of all, what is different about the quantitative model?
Okay, so the LLM is trained on what data? The data of the web, right? All the words of the web and the images of the web and YouTube videos. That's what large language models are trained on. They're great tools and they're having wonderful adoption time, as we see from the revenues of Anthropic and others. But there's a whole other world where we train not on the words of the web. We don't use anything from the web to train on. The data sets that we use, we build from scratch. SandboxAQ owns those data sets. We have to build them. It takes years sometimes to build these data sets in biology, in chemistry, in material science. And then once the data set is built and validated, we can then train a specialized model to make predictions about what would be a good drug. That's not a transformer model. That's not the model that drives the language models. That's a different architecture. So the two are complementary: language models and quantitative models.
Yeah. Your argument is that the language models get commoditized, right? They all sort of wind up in the same boat?
Well, language models are great. They're the interface to the AI world. They understand natural language. You can type into them very easily. Look at the adoption. It's been fantastic. But then the question is, what's the horsepower on the backend? What can it do for you? And what it can do for you right now is great digital work, right? It's taking over a lot of white collar tasks in a huge way. But how about when you ask it a question about the actual world of engineering and medicine and things like that? Right now, those answers are not the best in the world, right? We know about hallucinations, we know they're getting a lot of answers wrong. Now we have those real answers for the LLM community, and that's the fusion.
Is this all an enterprise story? Or is there a consumer element to this?
I think there's a prosumer element, right? I think some consumers may want to tap into this if they're curious about medicine or things like that. But I think it's both a prosumer, you know, so this B to P, business to professional, and then I think it's also B2B, and Claude obviously has had a huge run-up in its adoption in B2B.
Well, I'm trying to understand from a Sandbox perspective, is this a change in your business model? You've been selling your data, haven't you? Or the access to this, to companies themselves to help them. So does this change your approach?
