Alex Karp on Open Secure AI Alliance
작성: Tim Official: Alex Karp
37 tech companies just formed the Open Secure AI Alliance to push back against closed models from OpenAI, Anthropic, and Google, none of whom joined. Palantir CEO Alex Karp says his customers are angry about being 'token maxed' and watching their business value drift to the labs they pay. The alliance includes Nvidia, Microsoft, Dell, Cisco, CrowdStrike, and others. Nvidia warns policymakers against banning open frontier systems. Karp compared AI to uranium and admitted regulation is inevitable, even as the alliance argues against it.
스크립트 (en)
One of the big problems we have in AI is it is actually dependent on the models being orchestrated and managed, both for safety reasons and for value reasons. That's an application layer. We see on the front line in our classified deployments where we're using open-weight models, that we can get those models to be as good or better than Frontier models. Even the Chinese ones? We can get our models to be as good or better than any model. Yours. So just a minute. What you want in a stack is an American stack that's American based on every single part of that stack is the best part. You're talking about chips? Chips. Servers. Servers. Racks. No, actually, it's like so you have GPUs and then you have the models. And in between, those models are powered by an application layer. Part of what's happened, honestly, is that, you know, and Dario is an incredibly world-historic figure. Dario Mode. of anthropics. You're selling a model that actually only works at that level safely if you modulate it and use it. They're number one in enterprise. People want that safety, Alex. So the application layer allows for the safety, but also for creating these things, using them so they're actually valuable. A way of not addressing that issue is to say, oh, okay, we can't have any competition because the competition might be dangerous. Now, of course, I THINK IT A LOT OF PEOPLE THAT ARE GOING TO BE DANGEROUS OF COURSE THE COMPETITION COULD BE DANGEROUS BUT YOU ALWAYS HAVE TO BE VERY CAREFUL WE CANNOT GO DOWN THE PATH OF I could be dangerous But you always have to be very careful We cannot go down the path of I mean I very pro I watched Europe regulate itself out of business You end up with businesses that no one believes are businesses because they only exist behind the firewall of regulation. That doesn't mean we shouldn't watch all of these models very carefully, both open and closed. But most importantly, what our clients want, primarily U.S. clients, is they They want to be able to control a model, either a closed model, or they want to be able to control the weight so they keep the alpha in an environment that's safe. And they have this kind of vulgar, you know, no application there, no love saying that they're all saying. And they want to make sure that it's valuable on the front line. Was this a way to shut out or is it an attack on Anthropic? Because they didn't sign on. And you have complained that they very much control a lot of what would be the tokens that that's a. Well, I can tell you, I can look, I can't. Look, this revolution has taken off, and you can't put it back in the bag. And I think we helped launch it, maybe even the leaders and a lot of people are now following on. I'm not anti-anthropic or anti-any of the closed models. I am pro my customers, and my customers are angry. And while my customers are angry because they feel like they got token maxed while transferring their alpha to a frontier lab that, by the way, believes their business is only valuable because they're going to put it out of business. It's money. It costs a lot. No, it's not just it costs a lot. THAT'S ONE PROBLEM. THE BIGGER PROBLEM IS YOU ARE MIGRATING THE VALUE OF YOUR BUSINESS TO THE PERSON YOU'RE PAYING A LOT OF MONEY TO.
