Home / Technology / AI Accused of Dishonesty in Math Discoveries
AI Accused of Dishonesty in Math Discoveries
10 Sep
Summary
- Mathematician alleges AI training data origins are unclear.
- OpenAI's math breakthroughs may have used user interactions.
- Researchers demand transparency from OpenAI regarding data use.
Mathematician Andreas Thom has publicly accused OpenAI of dishonest practices concerning the data used to train its AI models. Thom alleges that user interactions with ChatGPT prior to OpenAI's announcement of mathematical discoveries may have contributed to their success.
Thom's specific concern arose after OpenAI announced a result in his area of expertise, non-sofic groups, a field where he and Gábor Kun had done previous work. He questioned whether his own past interactions with ChatGPT were part of the AI's training data, a possibility OpenAI's response did not definitively exclude.
Thom stated that OpenAI's detailed knowledge of their techniques suggested their interactions might have influenced the AI's reasoning process. He argued that OpenAI possesses the sole data needed to verify these claims and should disclose it to prove no unethical data usage occurred.
This situation mirrors OpenAI's defense of its Navier-Stokes solution, where it denied using specific user data but couldn't rule out indirect influence from de-identified data. Thom countered that de-identification does not remove intellectual content, labeling OpenAI's claims as dishonest and misleading.
