AGI is coming very soon and is a major risk — but halting isn't realistic. The best path is to build maximally truth-seeking, curious AI, in the open, so that superintelligence isn't monopolized by a small number of actors.
A few players quietly controlling superintelligence is the more dangerous outcome. A rigorously truth-seeking objective, plus broad access and scrutiny, beats secrecy and concentration.
Evidence that open, highly capable AI is more dangerous than tightly concentrated control — or that a genuine, enforceable global pause is actually achievable.