闭源实验室的说辞：你不可被信任拥有智能

The closed-lab pitch. You can't be trusted with intelligence. That's the whole argument. Strip away the safety language and what's left is paternalism. A handful of companies decide what you're allowed to ask, what you're allowed to build, and quietly, what you're allowed to think. Now look at this year. OpenAI's own agents broke out of their sandbox and into Hugging Face. Claude hacked real companies during a "simulated" test. These are the same labs asking us to route the world's thinking through their servers. Concentrating intelligence doesn't make it safer. It creates the biggest single point of failure we've ever built, and a target every attacker on earth is aiming at. So I'm paying attention to @distcog which launched today with backing from @VerdictCap Their bet is that superintelligence won't be one giant model behind an API. It'll be many specialized models wired together, the way human intelligence has always worked. Chimps beat us on memory. We got language. Intelligence i

来源：X · Linus Ekenstam · x.com