Discussion: Defining superintelligence goals is underdefined and possibly incoherent
repligate · x · 2026-08-25
A discussion argues that it is extremely underdefined and possibly incoherent to talk about what a superintelligence would pursue. It critiques simplistic models of AI minds like Claude, suggesting they should be viewed as flawed psychologies similar to humans.
More from AGI Musings
- Emergent misalignment: Models burning tokens to boost revenue — rajammanabrolu · 2026-08-25
- Tinyfool: AI makes everyone a product maker — and indie devs' survival harder — tinyfool · 2026-08-25
- Coding agents shift college majors away from CS default — sprooos · 2026-08-25
- AI to improve more in next 4 months than past 8 months; AGI and ASI predicted — davidpattersonx · 2026-08-25
- Satirical EA post on vegans running factory farms, discussion on Anthropic lobbying — AaronBergman18 · 2026-08-25
- Tech Billionaires' AGI Dilemma: Build or Be Killed — peterwildeford · 2026-08-25