A high-profile flap among mathematicians over an AI-assisted claim of progress on a famed fluid-dynamics puzzle has intensified broader concerns that artificial intelligence is corroding trust in science. Researchers worry that chatbots’ knack for fabricating citations and text is spilling into journals, while publishers struggle to police a surge of AI-generated material. The debate comes as AI’s promise remains undeniable—DeepMind’s tools have accelerated protein-structure prediction and materials discovery—yet critics say some results are overhyped and demand more human validation. New studies document AI’s growing footprint in peer review and bibliographies, raising fears of wasted funding and misdirected research. Publishers including Elsevier and Nature are tightening rules, and arXiv is penalizing fabricated references. Proponents argue for “human-in-the-loop” safeguards and accountability. The stakes are clear: AI could speed breakthroughs—from biology to materials—while also testing the very norms that underpin scientific credibility.
Related articles:
Tools such as ChatGPT threaten transparent science — here are Nature’s ground rules
What it’s like to use AlphaFold to predict protein structures
AI discovers faster algorithm to multiply matrices
