WorkThe story, in brief
Import AI 444: LLM societies; Huawei makes kernels with AI; ChipBench
Google researchers found LLMs aren't just answering questions—they're simulating entire societies inside their weights.
The KeyNews take
Why it matters
A Google research finding on how LLMs internally model multi-agent systems has implications for AI safety, interpretability, and how we should think about model behavior as complexity increases.
The key facts
5 to knowGoogle paper: LLMs simulate multiple personalities/agents to answer questions
Pattern emerges as model capability increases
LLMs build and manipulate rich multi-agent world models internally
Huawei making kernels with AI optimization
ChipBench mentioned (likely chip benchmarking framework)
Go to the source
Import AI (Blog)jack-clark.net
Publisher excerpt: Welcome to Import AI, a newsletter about AI research. Import AI runs on arXiv and feedback from readers. If you’d like to support this, please subscribe. Subscribe now Google paper suggests that LLMs simulate multiple personalities to answer questions:…The smarter we make language models, the more…