The fundamental control architecture for safe AI must keep humans in the loop: the AI can suggest architectural improvements, but humans must approve and run new training runs, rather than allowing the AI to modify its own weights autonomously.

normativepending

Speaker

Dave Blondin

Evidence Quote

the the AI Alec Radford that suggests the next improvement in its own architecture and then runs the test that's already underway and that's fine you know and that does create a new training run that generates new weights. That's different from then saying, "Oh, go ahead and change your weights by yourself." Uh, so to me, that's what keeps the human in the loop

Source

AI Experts Debate: AI Job Loss, The End of Privacy & Beginning of AI Warfare w/ Mo, Salim & Dave 176Peter H. Diamandis
Created: 8/11/2026, 7:21:56 AM

My Notes

Loading notes...