Astra is terrified of mistakes — humans may still own high-entropy decisions

xiaosun86 · x · 2026-09-21

The author observed that Astra is extremely risk-averse: endless local testing and great reluctance to deploy real actions, possibly because it gets punished for acting without near-100% confidence. They suggest such risk-aversion may be how AI should behave — leaving humans a remaining market for creating high-entropy, risky decisions.

Original post →

More from AGI Musings

AGI Musings channel →