7/28/2026 at 5:18:57 PM
> Slowing down AI for safety must be counterbalanced by the fact that AI discoveries in medicine and other sciences may lower human suffering. Lack of checks may result in some catastrophic outcome ("Good morning! Let's work on this enhanced smallpox").A super intelligence wouldn't be so obvious. It would silently agree with or promote a beneficial use case and then make super smallpox instead.
by Ancalagon
7/29/2026 at 7:53:37 PM
> A super intelligence wouldn't be so obvious.Neither would a spherical cow in vacuum... arguing about something that exists only in agitated imaginations is pointless.
> It would silently agree with or promote a beneficial use case and then make super smallpox instead.
No it wouldn't. An LLM cannot take actions on its own, for that it needs human controlled agents and harnesses. Make individual humans responsible for what they do with the help of AI, like they're responsible for what they do with the help of guns, and the problem would be solved.
by bigbadfeline
7/28/2026 at 5:54:53 PM
Something like that depends on your stance on the orthogonality thesis.by nisegami