AI News - Monday, February 9, 2026

Import AI23h ago

Import AI 444: LLM societies; Huawei makes kernels with AI; ChipBench

Researchers at Google have published findings suggesting that large language models simulate multiple distinct personalities or personas when answering questions. This discovery indicates that as language models become more sophisticated, they develop increasingly complex internal mechanisms for processing and responding to diverse queries. The research implies that LLMs may operate through a form of "society" of simulated personalities rather than as unified systems.

DeepMind23h ago

Accelerating Mathematical and Scientific Discovery with Gemini Deep Think

Google

Google's Gemini Deep Think represents a significant advancement in AI's ability to solve complex mathematical and scientific problems by enabling extended reasoning processes that mirror human problem-solving approaches. This development matters because it directly addresses a critical limitation of current AI systems—their struggle with multi-step problems requiring deep reasoning—and could substantially accelerate research timelines across fields like physics, chemistry, and mathematics by automating challenging analytical work.