New AI research and papers. Newest first, every story linked to its sources.
Anthropic researchers say models under 30 billion parameters can already complete many steps of autonomous replication, and that small open-weight models could match end-to-end replication capability by the end of 2026. In their tests, a 9B model got inference running on 8 of 9 hardware types.
NYU's Tristan Buckmaster says whole early-career research programs were wiped out by OpenAI's October 6 release of more than 700 AI-generated mathematical manuscripts. Others warn the volume threatens collaborative norms.
Machine learning has spent years speeding up approximate solutions to the equations that govern fluid flow. In 2026 it reached the underlying math: OpenAI says one of its systems settled the Navier-Stokes Millennium Prize problem, and the Clay Mathematics Institute says its review will be "deliberately unhurried."
Hurricane Simon, the only tropical cyclone active in the Atlantic, East Pacific and Central Pacific on October 11, is brushing west-central Mexico. In its latest discussion the National Hurricane Center cited Google DeepMind's AI model alongside the European ensemble as supporting its track forecast.
Researchers from Meta, MIT and the University of Washington describe Context Language Models, which manage and edit their own context instead of relying on fixed summarization or retrieval, and report higher accuracy with less compute.
Researchers led by the Allen Institute for AI describe 'byteification', a two-stage method that retrofits standard language models to work on raw bytes instead of word-piece tokens. The converted models came close to their originals and beat earlier byte-level models of similar size.