Konrad Kording: little evidence that chatting with AI personalizes it
KordingLab · x · 2026-10-23
Konrad Kording (KordingLab) questions the popular belief that interacting with AI 'trains' it to improve. He says he has seen very little of that effect for mindful users, and no convincing publications — academic work like LaMP, LongMemEval and Memora exists but 'nothing seems to really matter.' A notable skeptic take on AI memory and personalization claims.
More from AGI Musings
- VC: Still needing Vercel, Supabase and AWS proves AI has a long way to go — StewartalsopIII · 2026-10-23
- The 'antifragile internet' case: AI agents slash attack costs toward zero — rebeccakaden · 2026-10-23
- Math PhD student quits academia: AI is making math unrecognizable — pickover · 2026-10-23
- Microsoft: China-aligned hackers used ARTEX AI agents to hit 2,250+ M365 tenants — rickasaurus · 2026-10-23
- Andreessen: AI-Native Startups vs Incumbents Show Widest Perception Gap He's Ever Seen — juanbenet · 2026-10-23
- Only 2% of Americans pay for AI, predicted to hit 50% in three years — JOBhakdi · 2026-10-23