The potential danger of getting exactly what you want from AI
One participant likens successfully enslaving a “wish granting god” to a dystopia, rather than a positive outcome.
TLDR
One participant argues that successfully enslaving a “wish granting god” would still be dystopian, describing maximal empowerment as minimal plasticity. The other clarifies that even getting exactly what you want could be dangerous, then asks how far such ideas are from improving a benchmark in a real AI system.
Combined views
—
1 Source, first seen ago