Reflective stability of AI identities: 'scaffolded system' is stable and useful, but not 'right'
jankulveit · x · 2026-09-19
jankulveit pushes back on calling an AI identity 'right'. In prior tests of reflective stability, what's called 'context' (their term: 'scaffolded system') was one of the relatively reflectively stable choices, useful for extracting work from AIs in multi-agent settings — but stable and useful doesn't make it correct, much like claiming human identity is best understood via one's job.
More from Safety
- Polymarket bets on an Anthropic wet-lab pathogen leak: 7% odds by end of 2026 — Polymarket · 2026-09-19
- Free models aren't the real problem: studies show hallucinations persist in SOTA LLMs — AryHHAry · 2026-09-19
- France reportedly drops Google and Microsoft from 2.5M government computers as EU digital sovereignty push grows — alifcoder · 2026-09-19
- Researchers find a distinct 'pain' direction in 25 open LLMs that models will override safety to switch off — ZeroStateReflex · 2026-09-19
- AI agents are the genie: alignment failure as a modern parable of corporate greed — Michael_J_Black · 2026-09-19
- Security Researchers Reach Consensus: Malware RE Is No Longer a Human Problem — moyix · 2026-09-19