Are AI Labs Pelicanmaxxing?
ai
Researcher Dylan Castillo put a humorous thought experiment to the test: whether artificial intelligence labs had secretly optimized their models to draw pelicans riding bicycles. Hacker News is discussing the investigation, which examined forty-eight prompts to see whether models showed unexpected proficiency at this absurdly specific task. The premise traces back to a running benchmark from writer Simon Willison, who has long asked AI models to draw a pelican riding a bicycle as an offbeat gauge of their abilities.
Source: https://dylancastillo.co/posts/pelicanmaxxing.html
Listen to this story
Hear this and more stories in a personalized audio briefing.
Open The Chonkerton