BREAKING
OpenAI Tests 'ChatGPT for Science'
A Vertical Plan for Research
0
M
sci/math msgs per week
GPQA Benchmark Scores
GPT-5
92
Human
70
GPT-4
39
Promise Versus Verification Risk
Wins
●
Solved cosmic string problem
●
Proved 1970s open problem
Concerns
●
Verification errors
●
Risky for non-experts
Announcement Expected in Weeks
AI NEWS BLITZ
A code leak reveals OpenAI is testing a new science plan for researchers.