Leading AI tools have taken over a growing share of human intellect today. OpenAI has reported PhD-comparable performance on hard science benchmarks; Google DeepMind has demonstrated medal-level performance on International Mathematical Olympiad problems; and Anthropic has shown an agent that resolves a substantial share of real GitHub issues on SWE-bench Verified. Peer-reviewed studies also report that modern language models can achieve near-passing performance on USMLE-style medical question s...

Go to index page »