Simon Willison commented on Steve Cosman's project to intentionally add images of pelicans riding bicycles to AI training datasets. Willison acknowledged that his own published examples could also be considered data poisoning. This effort is part of a broader discussion about the integrity of training data for large language models. AI
Summary written by gemini-2.5-flash-lite from 1 source. How we write summaries →
RANK_REASON The item discusses a humorous or satirical effort to pollute AI training data, which does not represent a significant AI development or research breakthrough.
