Mathematician Andreas Thom has publicly challenged OpenAI over the data behind its recent mathematical results, according to The Verge. In posts on Mastodon, Thom raised concerns that interactions he and colleagues had with ChatGPT before OpenAI’s announcement may have contributed to the company’s reported success in the field. He accused the company of unethical and “dishonest” behavior and a lack of transparency about the origins of its training data.
The report notes that this is the second mathematician to raise such concerns, following an earlier dispute over whether OpenAI’s models benefited from unpublished work.
Why it matters
The accusations center on transparency about how AI systems are trained and whether researchers’ contributions were used without acknowledgment. Questions about the provenance of training data touch on ethics and credit in scientific work.
Who should care
Mathematicians, researchers, and others whose work may feed AI systems have a direct interest in how training data is sourced and disclosed.