🚨 New muckrAIkers: "AI Safety For Who?" @jacobhaimes.bsky.social & @thegermanpole.bsky.social break down how instruction tuning/RLHF create anthropomorphic chatbots that exploit human empathy—leading to mental health harms. kairos.fm/muckraikers/... Find us wherever you listen! (links in thread)

Dangers of being ab auto-didact: I somehow missed the identity $a^b=e^(b ln(a))$ for the last 5 years. I'm technically semi-autodidact because I did get formal training, but somewhere that very useful and basic thing got lost and I never had enough pain to derive it myself.

Probably our best episode yet, Chris was an awesome guest and is a true Mensch

🚨 New Paper Alert: Open Problem in Machine Unlearning for AI Safety 🚨 Can AI truly "forget"? While unlearning promises data removal, controlling emergent capabilities is a inherent challenge. Here's why it matters: 👇 Paper: arxiv.org/pdf/2501.04952 1/8