A curated newsletter covering recent AI research and developments, including an Anthropic study showing AI coding assistance reduces developer skill retention by 17 percentage points, research on LLM dishonesty detection using censored Chinese models, LLM reliability in materials science, Geoffrey Hinton's views on AI consciousness and the 'Volkswagen Effect', prompt injection attacks on AI agents, LLM safety vulnerabilities via colluding LoRA adapters, and benchmarks for LLM code translation efficiency.
Table of contents
Miro WengnerSort: