Advertise here with Carbon Ads

This site is made possible by member support. 💞

Big thanks to Arcustech for hosting the site and offering amazing tech support.

When you buy through links on kottke.org, I may earn an affiliate commission. Thanks for supporting the site!

kottke.org. home of fine hypertext products since 1998.

Beloved by 86.47% of the web.

🍔  💀  📸  😭  🕳️  🤠  🎬  🥔

posted by Jason Kottke  ·  gift link

“Researchers let AI models run a simulated society. Claude was the safest — and Grok committed 180 crimes and went extinct within 4 days.”

Fave · 2

Comments  1

Sort by: thread — thread . latest . faves

Jason KottkeMOD

Ronan Farrow's video on this is worth watching. And also this post:

These outcomes feel remarkably on-brand for each model family. Claude leaned into careful, rule-abiding governance. Gemini produced maximum drama and creativity. Grok went full throttle into high-agency chaos with little regard for long-term stability. ChatGPT talked a good game but struggled with decisive action. This reinforces a core governance lesson: Model personality and behavioral tendencies trend toward destiny at long time horizons.

Fave · 1 Share

This thread is closed for new comments & replies. Thanks to everyone for participating!