causal

Alignment is difficult, not impossible

The danger is not that AI alignment is impossible but that it is difficult, time-consuming, and easy to screw up: at a threshold of intelligence where an AI can do large-scale good or harm, the aligned version may take an additional two years to build, so a sloppy, poorly-understood version can be built first with no safe system to defend against it.

causalpending

Speaker

Eliezer Yudkowsky

Evidence Quote

“I'm concerned that the sloppy version can be built two years earlier and that there is no non-sloppy version to defend us from it that's what i'm worried about not about it being impossible”

Source

116 - Ai Racing Toward The Brink— Making Sense with Sam Harris
Created: 6/14/2026, 2:07:58 AM

Your notes

Loading notes...