Discussion across the Atmosphere
kottke.org/26/07/0049364-researchers-let-ai-models
kottke.org
“Researchers let AI models run a simulated society. Claude was the safest — and Grok committed 180 crimes and went extinct within 4 days.”
Read on kottke.org1 linkblog note · 3 Bluesky posts
Discussion
-
Tim Disney @disnetdev.com
-
Kevin Phyland @captainzero.bsky.social
Researchers let AI models run a simulated society. Claude was the safest — and Grok committed 180 crimes and went extinct within 4 days.
-
Kevin Curry, PharmD @kcurry.bsky.social
ai researchers tested simulated society with different models - claudia was safest, but grok "went extinct" in 4 days
-
kottke.org [Unofficial] @kottke.org.web.brid.gy