Conversation

Notices

1. Embed this notice
   Walter van Holst (whvholst@eupolicy.social)'s status on Sunday, 15-Jan-2023 22:45:42 JST Walter van Holst

   It is deeply ironic that generations of SF writers struggled with plotlines about AIs that would be unable to lie and the first AI-like thing that actually materialises mostly excels at lying and deception.

   • Embed this notice
     Alexandre Oliva (lxo@gnusocial.net)'s status on Sunday, 15-Jan-2023 22:45:41 JST Alexandre Oliva

     in reply to
     lying requires intent to deceive; I'm not ready to assign intent to computing machines
   • Embed this notice
     Alexandre Oliva (lxo@gnusocial.net)'s status on Sunday, 15-Jan-2023 22:49:42 JST Alexandre Oliva

     in reply to

     • Luis Villa
     hallucination is an interesting way to put it. me, I conceive of it as generalization (interpolation and extrapolation) uninformed by common sense as to what's reasonable to generalize. we make generalizations all the time, and we often even mistake our conclusions as facts, but many of us have some sense as to what's reasonable and socially acceptable to make up and what isn't. the machines may eventually learn that too.
   • Embed this notice
     Luis Villa (luis_in_brief@social.coop)'s status on Sunday, 15-Jan-2023 22:49:43 JST Luis Villa

     in reply to

     @whvholst hallucination is even weirder than that; it’s not intentionally deceiving, it’s just making shit up, which … I feel like is even less explored in scifi? Malicious AIs are all over the place in fiction. This blithe “sure, you asked a question, I can’t say no” is weirder than that.