Twitter | Pretraživanje | |
Stephan Bridger
Weird that people will admit, even boast, about not updating their beliefs, but they would never want an AI running around that never updated its beliefs.
Reply Retweet Označi sa "sviđa mi se" More
Figment of Imagination 27. pro
Odgovor korisniku/ci @hexagr @GlitchesBrew
Updating beliefs is good and useful. Updating goals is slightly more precarious and should be done with great care if at all.
Reply Retweet Označi sa "sviđa mi se"
Stephan Bridger 27. pro
Odgovor korisniku/ci @stellahymmne @GlitchesBrew
I believe those concepts are intractable, thus cannot imagine anything that shouldn't be done with great care.
Reply Retweet Označi sa "sviđa mi se"
Dan Solitz 27. pro
Odgovor korisniku/ci @hexagr
How do you integrate beliefs into AI operating system? And if AI could believe would it generate a faith?
Reply Retweet Označi sa "sviđa mi se"
Stephan Bridger 27. pro
Odgovor korisniku/ci @dansolitz
This is the problem. How do we build the AI to become good, regardless of faith? The AI alignment problem is about humanity as much as it is about AI.
Reply Retweet Označi sa "sviđa mi se"
Dont Worry About It 27. pro
Odgovor korisniku/ci @hexagr @GlitchesBrew
For sure, but it's easy to go from updating beliefs to hop-scotching to whatever is politically/socially/financially expedient
Reply Retweet Označi sa "sviđa mi se"
Symora ꙮ 27. pro
Odgovor korisniku/ci @hexagr
the difference is that humans eventually die and get replaced
Reply Retweet Označi sa "sviđa mi se"
Psi Operator 28. pro
Odgovor korisniku/ci @hexagr
Depends, how secure are those updates? Are they comparable with my setup? Will they break my workflow. A true power user will never update without checking first.
Reply Retweet Označi sa "sviđa mi se"