I appreciate the academic and rigorous effort to measure the bullshit that LLMs generate: machine-bullshit.github.io/
I just wish we would do away with the rhetoric that we're just aiming to make these models better.
How about no? How about admitting that the premise is poisoned from the start?
How about saying that holding everyone's labor, creativity and expression hostage for the sake of deskilling people and making them precarious is a horrible idea?
#AI #antifascism #antiAI
reshared this
Aral Balkan
in reply to Alex Ștefănescu • • •This.
But they’re afraid that if we do that, the whole system might unravel because it’s bullshit all the way down.
The Sleight Doctor 🃏
in reply to Aral Balkan • • •Yep. Also, chatbots cannot possibly get any better than they already are, because they're polluting the internet with error-strewn slop - and that's their training dataset! They've already reached peak performance and it's all downhill from here, no matter how much more money is thrown at them.
The analogy I've used before is that it's like feeding a chicken its own eggs. It'll work for a while...but eventually the chicken will die from malnutrition.