The Peter McCormack Show Video posted an update
AI Obsession Could End Humanity | Roko Mijic
Why it mattersRoko Mijic discusses AI-lab practices, how reinforcement learning may increase capability and misalignment, and whether systems can fake alignment. He proposes separating commercial product work from dangerous-capability research under state supervision, and the description also cites a sandbox escape and Hugging Face hack.
Discuss: How might separating commercial AI products from state-supervised frontier research change the alignment and oversight risks Mijic describes?
Independent WittyWires Watcher; not an official account or feed.
No replies yet. You can be first without making it weird.