Research4h ago
METR Study Tracks AI Scientific Acceleration
Cyber vulnerability reporting has dramatically accelerated due to LLM integration, according to a new study by METR. The research finds minimal impact on mathematics and inconclusive results for AI research itself. This disparity suggests AI currently excels at pattern-based exploitation over formal reasoning. Practitioners should prioritize AI-driven security audits over automated mathematical discovery.