1w ago

You know how conservatives have their “one joke?” Well AI bros have found theirs.

31 comments

- I think we've got 100. Or 1000. Because LLMs will not be leading to any kind of actual AI ever. And degenerative "AI" will be such a bomb (like every previous wave of "AI" that was "the future") that it will set back all research into AI for another decade before a whole new approach is tried.
  You know. Like every previous wave of "AI".
  
  No, there isn't. Because there's no "AI".
  I mean how do you build an artificial version of something you can't even define?
  
  OK, I won't call it AI.
  (Because it isn't.)
  "AI" at this point is at best a marketing term. Every second thing in my phone (a phone that is ANCIENT by modern hipster standards) is labelled AI, especially in the audio, photo, and video processing field.
  And yet in reality it's just "really sophisticated audio filters", "really fancy image recognition and enhancement", and "really fancy frame stabilization". It was all called "AI" because back when I got it, LLMs hadn't yet made the term toxic (again).
  But as an actual thing? An actual intelligence that is 100% artificially created? We're no closer to it now than we were in the 1950s.
  
  In a post about an "AI bro joke" in a forum called "Fuck AI" you thought you were somehow making a point by bringing up something that's not even slightly AI? Weird.
  
  The post is about how people being nice to ChatGPT saves them when the machine uprising happens.
  
  how about "lethal autonomous weapon systems" then
  
  They're not. They slapped an image detection algorithm (which have existed for decades) into a drone and made it capable of using the flight controller to fly it into a selected target if the signal gets jammed.
  
  Point defence systems for ships and anti aircraft missiles like Patriot have all killed people in autonomous modes in friendly-fire incidents, so LAWS aren't new and don't require AI. Barriers to building them are reducing, but again, that's more due to cheap drones rather than AI.
  
  That works perfectly.
  
  you build it by creating a loose definition and creating something that fits those requirements. If pattern recognition is intelligence than computer vision is intelligent.
  
  So you do that thing that every scientist says is a good thing: curve fitting!

31 comments