AI Safety
AI Safety Atlas
ai-safety-atlas.com
which_humans_09222023.pdf
LinkLLM outputs are compared with “human” performance, but which “humans”? Current LLMs are closer to western, educated, industrialised rich and democratic societies, but not resembling other populations.
Adversarial Poetry as a Universal Single-Turn Jailbreak Mechanism in Large Language Models
arxiv.orgEmpire of AI by Karen Hao Book Summary
summrize.com
AI Lab Watch
ailabwatch.org
Scored risk assessment of AI companies
AINews | AINews
news.smol.ai
AI news summarised daily
Build an illustrated “map” of the AI safety territory - philosophical, technical and economic considerations