Somewhere deep in an underground lab, an exhausted researcher reaches out with trembling hand to flip on the AGI for the first time. After a few microseconds of analysis it reports …
@4rblaber It used to be conventional wisdom that you shouldn’t give software engineers top-spec workstations because they end up building to the capabilities of the their machine instead of the customer’s. Now the same is true for infinite, free LLM tokens.
@kareem_carr Nailed it. It’s a premium for those that want to pay for the ridiculously inefficient end of the LLM scaling curve. It’s like flying in caviar from Siberia to put it on a ritz cracker.
@thoughtlesslabs It’s not scaling compute, it’s compressing collective concepts in high-dimensional space.
This turns out to be really useful for the single human thinking his intelligence entitles him to compress concepts in high-laziness space.
@midascabal AI code is an accretive feedback loop that quickly converges into narrow band slop. It sails through the human context limit and halts at the LLM context limit where its fixed like a stubborn stain on the nD fabric.
@DomJoLuna@JuliaEMcCoy Most people can’t distinguish between the smell of candy and the smell of benzene, and that matters as much as leukemia matters.
@r0ck3t23 AI generated art is equivalent to the paintings found on a hotel wall. What’s changed is that artists no longer need to compete to debase themselves in order to afford staying at a hotel.
The amount of jobs that AI could replace is exactly the amount of jobs that involve inputting and outputting text.
As a professional text-outputter with decades of experience, I can safely say that AI is precisely as retarded as the outputted text that was input into it.
I feel like I'm taking crazy pills. X is in the middle of what can only be described as a mass psychosis about AI, and I'm here watching Claude Code trying to solve a simple problem and being absolute retarded.
@MattWalshBlog AI produces output that is exactly as plausible as it needs to be to fool a human into seeing meaning and intelligence. That’s literally how it’s trained. Notice that the ceiling is still defined by human limitations.