Conversation

Notices

1. Embed this notice
   alcinnz (alcinnz@floss.social)'s status on Sunday, 01-Dec-2024 12:56:33 JST alcinnz

   I've been wondering if there is any actual innovation behind this round of "AI" hype besides the sheer scale. But the polarisation (in which I'm painted as anti-, bit of a simplification...) made it difficult for me to tell!

   3Blue1Brown mentioned that transformer models were innovated in 2017, which is what allows us to throw so much hardware at the problem.
   Took 4 or 5 years to bring it to market at said scale, but that seems about right!

   1/1.5!

   • Embed this notice
     alcinnz (alcinnz@floss.social)'s status on Sunday, 01-Dec-2024 13:00:33 JST alcinnz

     in reply to

     Jevon's Paradox aside, I do think Transformer Models (2017) are useful technology!

     Optimizations will still be useful when we scale down! Creating models not pretending to be general-purpose...

     1.5/1.5!

   • Embed this notice
     alcinnz (alcinnz@floss.social)'s status on Sunday, 01-Dec-2024 14:26:44 JST alcinnz

     in reply to

     • Wannabe Astronaut

     @space_cadet I mean then-new techniques which can be useful for solving real problems.

     I'm coming from a computer-science perspective of innovation!

   • Embed this notice
     Wannabe Astronaut (space_cadet@aus.social)'s status on Sunday, 01-Dec-2024 14:26:45 JST Wannabe Astronaut

     in reply to

     @alcinnz
     When you say "innovation", what definition are you using? Because in the circles that I move within, "innovation" means taking some already-existing, tried-and-proven technology, and doing Business stuff to successfully bring that thing to market, and "Make Bank".

     Actual Creativity is different altogether.