arXiv:2610.35919v1 Announce Type: new Abstract: Advances in artificial intelligence (AI) raise the prospect of substantial changes in how research is conducted and who can direct discovery. This perspective argues that universities should prepare for these changes by developing AI as well as using it, rather than relying primarily on capabilities designed elsewhere. The relationship is reciprocal: AI can advance disciplinary research, while scientific questions, experimental evidence, and observed failures can guide AI development. The proposal distinguishes three complementary commitments: shared operation of existing AI models, disciplinary AI development, and foundational AI research. Their resource requirements differ, and no single university needs to reproduce a frontier AI company. Shared computational services should be connected to sustained technical teams, laboratory experimentation, and knowledge records that make findings traceable and reusable. Independent evaluation, alignment research, transparent resource allocation, and inclusive academic governance are integral to this infrastructure. Consortia and national facilities can support capabilities beyond individual campuses while preserving room for investigator-led inquiry. Initial programs in experimental and mathematical research would test scientific value, the effort required to validate results, and improvements to AI before expansion. The argument concerns research rather than educational reform and calls for building institutional capacity before the extent and timing of disruption are settled.
Continue to arXiv Mathematics →You are being forwarded in 6s — stay on this page