"We did it, Patrick! We made a technological breakthrough!"
A place for all those who loathe AI to discuss things, post articles, and ridicule the AI hype. Proud supporter of working people. And proud booer of SXSW 2024.
AI, in this case, refers to LLMs, GPT technology, and anything listed as "AI" meant to increase market valuations.
Yes, I'll just trust the AI to help me fuck around with grid voltage levels.
With google search results becoming so poor, I guess I need to look into kagi or duck-duck. Pain in my dick, motherless goatfucking, horsehit-happy asshole, corporate varmints gotta fuck it all up for more profit. I'm tired, yo.
This assumes people know how AI works, and that's not common knowledge.
Dude, like a handful of people have any idea how it works, it was completely a black box until recently when they started deciphering how it “thinks” before outputting in any given language. And before anyone argues, yes, researchers have actually started to assess how it processes stuff now, they’ve recorded the binary processing that wasn’t language based and upon suspecting they found the processing akin to what we’d call thinking for ai, they saved the streams, presented it to other ai in droves and asked them to interpret it independently of the other system that manufactured it, and indeed it matched up with the processing of requests, but was not in any human language, it’s indeed machine code, but not like Fortran or cobal or hexadecimal codes that we are used to dealing with, it has its own language. So no one has a library or Rosetta Stone yet to interpret this, and as of now you have to “trust” other AI to tell you what it means. Which obviously isn’t a good idea at all.
Edit: the paper since some of y’all don’t believe it, and frankly I don’t blame you. It’s new research. "Do Sparse Autoencoders Capture Concept Manifolds?” https://arxiv.org/abs/2604.28119
not hexadecimal codes lol. you are clueless.
but also, the point was that for most even a general level of understanding is missing, making them easily think that it is the all knowing machine that cannot be wrong
You have no idea what you're talking about and you'd feel a lot better if you could just accept that. This story you are telling is absolutely nonsense fantasy, stitched together from a bunch of different actual things to create a narrative that tells a completely untrue and fantastical story
Why do you have the exact number of upvotes that the person you're replying to has downvotes. Seems like an unlikely coincidence. Also weird how you don't actually address anything in the comment. Almost like some bot brigade.
if someone tells you some computer data can't be represented in hexadecimal codes, that's a clear sign they don't even know how computers work, let alone LLM. its also futile to compare the LLM internal representation of data to programming languages, its like comparing birdsong to building blueprints.
Hilarious conspiracy brained shit.
Wrong. Here’s one of the papers on the research. "Do Sparse Autoencoders Capture Concept Manifolds?"
https://arxiv.org/abs/2604.28119
and a video on the paper
https://youtu.be/_G1NBI3AW5Q
Your writing shows that you have no idea how to interpret this paper at all. You clearly have no historical concept of the development of machine learning, neural networks, and the fundamentals of the domain. You are trying to read a cutting edge paper that relies heavily on the reader having deep domain expertise. The video you linked is ALSO not for laypersons as it doesn't actually explain the basics of the domain. You can't just dive into the deep end of a pool without knowing how to swim and then tell everyone you were lost at sea.
This isn't even what I meant by "how it works". I mean that it's basically a machine that outputs whatever next text that would be the most probabilistic given the previous text and input.
So the basics of how it works. Not exactly how it works.
Most people don't know the basics of AI and how it works in that sense. They hear AI and think "I Robot" and think the computer is actually talking to them and giving them something that a "brain" has "reasoned" about, which is not the case.
I think you missed the key part of what the person you replied to said: common. It is not common knowledge. Yes some people who have done extensive research and spent hundreds of hours on this stuff might have a somewhat good understanding of how LLMs and GenAI work. Do you think just your average person using Google has any idea how it works? No, no they do not.
You want to cite some sources because like a lot of things people say about AI this too sounds like bullshit?