Don't slow AI progress, accelerate AI safety investment instead

tristanbob · x · 2026-09-13

Responding to Rauchh, the author agrees AI progress shouldn't slow down but argues investment in AI safety must accelerate, suggesting directions: making good behavior a natural outcome of training, ensuring models explain their thinking for interpretability, keeping manual failsafes, and more. The author isn't a safety expert but stresses the field needs many more experts.

Original post →

More from AGI Musings

AGI Musings channel →