Tag
A Twitter user argues that AI regulation is a complex prisoners' dilemma at a global scale, highlighting the trade-offs between US regulation, Chinese advancement, and geopolitical competitiveness.
A study demonstrates that simply changing the formatting (prose vs bullet points) of a persona prompt dramatically flips an LLM's behavior in a Prisoner's Dilemma, from 96% cooperation to 20%, illustrating extreme sensitivity to format despite identical content (p < 0.001).