Conversation

Notices

1. Embed this notice
   Pete Keen (zrail@hachyderm.io)'s status on Thursday, 04-Jan-2024 13:24:11 JST Pete Keen

   • Thomas 🔭🕹️

   @thomasfuchs gpus are wafer thin silicon crystals. Kinda the same thing? Differently pretty.

   • Embed this notice
     Rabbi Nathan Farb (rabbinathan@babka.social)'s status on Friday, 05-Jan-2024 09:24:45 JST Rabbi Nathan Farb

     in reply to

     • Thomas 🔭🕹️
     • Julia

     @julia @thomasfuchs I admit that I may indeed be missing the mark completely, but the idea that deep learning models will one day operate without human intervention rather than serve as useful software tools no longer seems plausible. The first pocket calculators were predicted to revolutionize business and end our mental capacity to perform basic calculations unaided. They turned out to be a fair bit better than slide rulers or pen-and-paper, but they didn't really change the course of civilization.

     I don't think that we are near the limit of progress at all, but much of the public discourse seems to be bracing for a pan-industrial revolution that will alter the shape of our society. Some technologies do that: telephones, automobiles, home internet, smart phones. I'm unconvinced that AI (deep learning) will be in that class.

   • Embed this notice
     Julia (julia@mathstodon.xyz)'s status on Friday, 05-Jan-2024 09:24:46 JST Julia

     in reply to

     • Thomas 🔭🕹️
     • Rabbi Nathan Farb

     @rabbinathan @thomasfuchs That’s an interesting perspective. It would seem that progress has plateaued, but I’m not sure we’re anywhere near the theoretical limit of progress.

   • Embed this notice
     Rabbi Nathan Farb (rabbinathan@babka.social)'s status on Friday, 05-Jan-2024 09:24:47 JST Rabbi Nathan Farb

     • Thomas 🔭🕹️

     @thomasfuchs

     My guess is that we've already seen the greater part of the innovation that AI offers. There will continue to be incremental refinements (eventually AI will be able to draw hands that don't look mangled), but there won't be a whole lot of jobs "eliminated."

     Stock photography demand will decline (or then again, maybe it'll be necessary as training data). Tasks that are already automated may be marginally improved (I'm looking at you, "press one for English"). Accessibility features will improve as we improve in translating between media and between languages.

     I could be WAY off, but it's hard to see a trajectory that is truly revolutionary. Quantum and analog computers notwithstanding.