Sign up
Sign in
Poorly aligned models with superhuman capabilities open a scenario where humans have to deal with models they don’t understand, that far exceed their capabilities and, worryingly, aren’t aligned, hence unpredictable.
Ignacio de Gregorio
Justin Ohms
Follow
--
2
Share
For the foreseeable future, we can just unplug them.
Help
Status
About
Careers
Press
Blog
Privacy
Terms
Text to speech
Teams
