Back to Elon Musk: "Humans Won't Be in Control" When AI Becomes Smarter Than Us, with Elon Musk.
Should AI Labs Test Each Other's Models?
Elon Musk says AI labs should test a rival's new model a week before it is released. Walter Isaacson asked whether the five frontier models, Musk's and the others, could get together once a week and ask how they are going to save humanity from themselves. Musk says he has already recommended that exact thing.
He told the other labs they should have a call once a week. A week before release, a lab would give competitors access and ask them to try the new model and see if they find any problems. Musk says competitors would keep each other honest in that situation. If someone says they were able to make a nuclear bomb with the AI, the lab would fix that before release.
Musk sets the idea against a regulatory agency. There is a lot of talk of having one, but it takes a long time to get a regulatory agency established, and rulemaking is very slow.
Elon Musk: "Humans Won't Be in Control" When AI Becomes Smarter Than Us ·
From the transcript
It seems like they're all kind of the opposite of their name. It's just pretty difficult to create guardrails, especially the government. I can't imagine how Congress or whatever could do it. Could you imagine the five frontier models, yours and the others, just getting together once a week and saying, all right, how are we going to save humanity from ourselves? No, in fact, I've recommended that we do that exact thing.
Where I said, like, guys, what if we just have a call once a week, just like you're saying, and say, and that maybe we, a week before we release a model, we give that, you know, we give access to our competitors to say, hey, guys, why don't you try out the new model, see if you see any problems with it. And, you know, just keep each, I think competitors would keep each other honest in that situation.