Thread

1. cubefox · 1789762352 · focus · HN ↗

   So the models will not only be using more and more Neuralese in their CoT (like GPT-6), but different agents will also be able to communicate with each other in Neuralese. It's not looking good for monitorability.

   1. sparky_twofort · 1789764147 · focus · HN ↗

      Is Neuralese in no way decodable into a human-interpretable system? Genuine question -- I don't know the answer.

      1. Y_Y · 1789764555 · focus · HN ↗

         Definitely decodable, that's what's being done now

         1. cubefox · 1789771252 · focus · HN ↗

            It's not.

Unofficial Hacker News client; not affiliated with Y Combinator.