An AI model named GPT-6 Astra reportedly achieved a 98% score on a physics test, surpassing human experts. The AI's performance revealed errors in the test's answer key, indicating a potential shift in scientific discovery and validation processes. AI
IMPACT This event suggests AI's growing capability to not only solve complex problems but also to identify flaws in established knowledge, potentially accelerating scientific research.
RANK_REASON Frontier-lab model release with system card. [lever_c_demoted from frontier_release: ic=1 ai=1.0]
Read on Mastodon — mastodon.social →
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →