Conversation

Notices

1. Embed this notice
   Tony Toon (deepthaw@social.sdf.org)'s status on Sunday, 14-Apr-2024 04:26:00 JST Tony Toon

   • Thomas 🔭🕹️

   @thomasfuchs Meanwhile it took my parents like fifteen minutes of vigorous work and then they just had to wait about 18 years to get something that pretends to think (but doesn’t) and to deliver bullshit non-deterministic results while using absurd amounts of computational and environmental resources living in their basement.

   • Embed this notice
     Tony Toon (deepthaw@social.sdf.org)'s status on Sunday, 14-Apr-2024 04:31:21 JST Tony Toon

     • Thomas 🔭🕹️

     @thomasfuchs You’ve not seen my coffee cups

   • Embed this notice
     Brian Drell :veritrek: (drell@hachyderm.io)'s status on Sunday, 14-Apr-2024 12:38:11 JST Brian Drell :veritrek:

     • Thomas 🔭🕹️

     @thomasfuchs I want a fully deterministic answer engine. Like, all it can ever do is provide verifiable facts. This is a really hard problem, but it’d be immensely more useful than a hallucinatory LLM in providing useful information.

   • Embed this notice
     Andrew Feeney (andrewfeeney@phpc.social)'s status on Sunday, 14-Apr-2024 23:02:17 JST Andrew Feeney

     • Thomas 🔭🕹️

     @thomasfuchs Not disagreeing with you at all, but isn’t the “non-determinism” just an illusion created by a pseudorandomly generated parameter? Like any neural network, the same model should always give the same result for the same parameters right?

   • Embed this notice
     Arne Visscher (zomgwtfbbqkewl@hachyderm.io)'s status on Monday, 15-Apr-2024 00:27:58 JST Arne Visscher

     • Thomas 🔭🕹️

     @thomasfuchs very succinctly put.