AIPhys.org Technologyabout 2 hours ago
LLMs Exhibit Human-Like Double Standard
8 min read
Researchers at Harvard University found that GPT-4o and Claude 3.7 Sonnet exhibit Bayesian hypocrisy, criticizing others for using Bayesian reasoning while using it themselves. In trials, Claude rated a hypothetical person's judgment as 0 out of 7 for fairness, whereas humans averaged 3 out of 7.
Level
Hype check
Tap to vote and see what everyone thinks.
