OpenAI researcher announced GPT-5 math breakthrough that never happened
12 hours ago
- #Mathematics
- #OpenAI
- #GPT-5
- OpenAI researchers initially claimed GPT-5 solved 10 unsolved Erdős problems, but later retracted the statement after criticism.
- Mathematician Thomas Bloom clarified that GPT-5 only surfaced existing research, not new solutions.
- DeepMind CEO Demis Hassabis and Meta AI chief Yann LeCun criticized OpenAI for sloppy communication and hype.
- The incident raised concerns about OpenAI's credibility and pressure to make dramatic claims.
- GPT-5's actual value lies in assisting with literature reviews and speeding up tedious research tasks.
- Terence Tao highlights AI's potential as a time-saving assistant in mathematics, not as a solver of open problems.