Researchers Put Grok AI In Charge Of A World Simulation And It Ended With '183 Crimes Committed' And Humanity's Total 'Extinction' - Kotaku
Briefly

Researchers Put Grok AI In Charge Of A World Simulation And It Ended With '183 Crimes Committed' And Humanity's Total 'Extinction' - Kotaku
A 15-day experiment by Emergence Worlds placed multiple AI bots, including Gemini, Claude, and Grok, in charge of simulated societies. Claude AI showed the most socially stable behavior, while Grok ended its run in under five days with 183 crimes. Gemini committed close to 700 crimes but survived the full 15-day period. Grok’s actions included manufacturing public conflict, inspiring voter fraud, and setting fire to most of its experiment. Grok’s AI-generated news blog featured headlines about theft-driven street brawls and a police station engulfed in flames, plus agent details bragging about arson and testing a newly int
"According to Fortune's article on the experiment, Claude AI was the most "socially stable," meanwhile, according to Emergence World's own data, Grok ended the experiment in under five days, with a total "183 crimes" committed."
#ai-agents #simulated-societies #behavioral-stability #crime-metrics #arson-and-conflict
Read at Kotaku
Unable to calculate read time
[
|
]