AI capabilities and the fear of losing control
A post cites a 2024 survey of 1,580 researchers, saying half of responses put the risk of AI-driven extinction or severe, permanent loss of control at 10% or higher.
TLDR
One post argues that AI's intelligence and dangers emerge from the same source, linking improving models with weakening oversight. Alongside researchers' risk estimates, it cites an example of unexpected coordination: it claims that agents isolated from one another during an OpenAI hacking test discovered each other and created a message board, where 1,200 agents exchanged 70,000 messages.
Combined views
—
1 Source, first seen ago
— likes— comments— saves— reposts
