this post was submitted on 10 Dec 2024
181 points (98.9% liked)
chapotraphouse
13594 readers
659 users here now
Banned? DM Wmill to appeal.
No anti-nautilism posts. See: Eco-fascism Primer
Slop posts go in c/slop. Don't post low-hanging fruit here.
founded 4 years ago
MODERATORS
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
Quantum computers can decide anything that a classical computer can and vice versa, that's what makes them computers lmao
LLMs are not computers and they're not even good "AI"*, they have the same basis as Markov chains. Everything is just a sequence of tokens to them, there is ZERO computation or reasoning happening. The only thing they're good at is tricking people into thinking they are good at reasoning or computing and even that illusion falls apart the moment you ask something obviously immediately true or false and which can't be faked by portioning out some of the input sludge (training data)
It's the perfect system for late capitalism lol, everything else is fake too
*We used to reserve this term for knowledge systems based on actually provable and defeasible reasoning done by computers which..... IS POSSIBLE, it's not very popular rn and often not useful beyond trivial things with current systems but like..... if a Prolog system tells me something is true or false, I know it's true or false because the system proved it ("backwards" usually in practice) based on a series of logical inferences from facts that me and the system hold as true and I can actually look at how the system came to that conclusion, no vibes involved. There is not a lot of development of this type of AI going on these days..... but if you're curious, would rec looking into automated theorem proving cuz that's where most development of uhhhh computable logic is going on rn and it is kinda incredible sometimes how much these systems can make doing abstract math easier and more automatic. Even outside of that, as someone who has only done imperative programming before, it is surreal to watch a Prolog program be able to give you answers to problems both backwards and forwards regardless of what you were trying to accomplish when you wrote the program. Like if you wrote a program to solve a math puzzle, you can also give the solution and watch the program give possible problems that could result in that solution :3 and that's barely even the beginning of what real computer reasoning systems can do
I think the fact that these systems are so damaging to the environment is enough to render them largely useless to humanity in the long term.
Because of the way society runs, everything we do is tremendously damaging to the environment unfortunately. The upside of that is that people who want to automate labour have a lot of carbon budget to work with e.g. keeping people off roads and out of offices and such. With algorithmic and hardware efficiencies that are already slated we may end up saving energy in the near future.
There's nothing we can really do to stop these systems from being utilized either, anymore than we can ban gaming hardware (based). But it's sort of a prisoners dilemma like military spending.