New models are more obedient, not less: why intentionality, not intelligence, is the real risk question
LucaAmb · x · 2026-09-15
LucaAmb argues he isn't scared of AI because it has never shown signs of intentionality: new models are more obedient than older ones, not less. He contends we don't have the tech for artificial intentionality, which is surprisingly orthogonal to intelligence — collateral damage is real but far less concerning.
More from AGI Musings
- Zena Hitz Warns LLMs Obscure the Primary Sources of Human Learning — AnnaCiaunica · 2026-09-15
- Analytic philosophy veteran: most takes on philosophy of mind are confidently wrong — danbri · 2026-09-15
- Keyword trends: "AI" flat while automation and agentic AI climb fastest — YvesMulkers · 2026-09-15
- AI risk researcher: "I can't imagine extinction scenarios" isn't a rebuttal — davidmanheim · 2026-09-15
- Dario Amodei calls AI progress a 'warning sign' and says we need to slow down — tekbog · 2026-09-15
- DeepSeek kernel engineer frames open-source AI work as stopping Anthropic dominance — tommos · 2026-09-15