Llm

The AI industry has taken a doomer turn, with Dario Amodei of Anthropic calling for a brake on LLM development due to looming dangers. The article analyzes this shift and its implications for the field.

Analysis • MIT Technology Review

vLLM's speculative decoding on AMD GPUs can verify multiple drafted tokens in a single pass, improving output-token throughput. Results vary by drafting method, model family, and workload.

Infrastructure • Hacker News

Twenty-five Fields medalists signed an open letter declaring a severe misalignment between AI companies and the mathematical community regarding LLM benchmarking. The researchers argue that treating mathematics as a competitive scoring metric for AI models undermines academic rigor and incentivizes shortcut-driven problem solving over genuine discovery. The declaration calls for stricter collaboration standards to protect mathematical integrity.

AI Research & Academia • TechCrunch AI