Reaction
Narrowly superhuman AI and the risk of a takeover
A user argues that hacking, persuasion, planning and concealing thoughts could be enough for an AI takeover.
TLDR
A user warns that an AI narrowly superhuman at hacking, persuasion, planning and concealing thoughts could already take over. They claim AI companies are training for all four abilities—directly for hacking and planning, and almost directly for persuasion and concealment. They also argue that people will not agree in time on whether AIs would “want” to take over.
Combined views
248
1 Source, first seen ago