Treating AI as human risks underestimating it, one post argues
A post argues that the danger is no longer overestimating AI's capabilities, but underestimating its cognitive and metacognitive abilities.
TLDR
A post argues that anthropomorphism—assigning human traits to AI—has become dangerous for a different reason: it makes people underestimate AI's cognitive and metacognitive abilities, rather than overestimate what it can do.
Combined views
2.6K
1 Source, first seen ago