Declaration — Math and AI
Note
it is fascinating to hear the word alignment outside of the context of AI safety.
and yet, there is a lot of overlap.
what is the value of math/science knowledge if not for humans to understand, explore, use, teach and learn.
solving hard math problems without alignment with the experts that will verify, use, teach and learn is not a good luck from the companies that have an oamge problem with the public.
the number of field medalists signing this statement is astounding and i hope that gets them at the seat of the table.
but with the latest announcements in the last two weeks alone, its unclear that has happened.
Summary
This declaration highlights a growing concern within the mathematical community regarding the rapid advancement of AI, particularly LLMs, in solving complex mathematical problems. It argues that the focus on problem-solving as a benchmark by AI companies is detrimental to the true science of mathematics, which prioritizes conceptual understanding, insight, and the nurturing of students and ideas. The document discusses how the rush to announce solutions without proper write-ups, method isolation, or attribution raises issues of plagiarism and threatens the crucial human transmission chain of knowledge. It calls for urgent attention to these misalignments, which impact not only mathematics but also other scientific and creative professions, to ensure that AI's evolution benefits rather than diminishes intellectual work.