5 ms·
> max out at "knowing everything" LLMs know nothing but are great at giving the illusion that they know stuff. (It's "mansplaining as a service"; it is easier
by WorldMaker 5mo ago
> max out at "knowing everything"
LLMs know nothing but are great at giving the illusion that they know stuff. (It's "mansplaining as a service"; it is easier to give confident answers every time, even if they are wrong, than to program actual knowledge.) Even your first case seems wildly optimistic. The second case is a lot of "maybes" and "we don't know how but we might figure it out" that seems like a lot to bet an entire farm on, much less an entire industry of farms.
We sure are looking at a shift in the job market, but I don't think it is a fork in the road so much as a Slow/Yield sign. Companies are signalling they are willing to take promises/hope to cut labor costs whether or not the results are real. I don't think anything about current AI can kill the software development industry, but I sure do think it can do a lot to make it a lot more miserable, lower wages, and artificially reduce job demand. I don't think this has anything to do with the real capabilities of today's AI and everything to do with the perception is enough of an excuse and companies were always looking for that excuse. (Just as ageism has always existed. AI is also just a fresh excuse for companies to carry on aging out experience from their staff, especially people with long enough memories/well schooled enough memories to remember previous AI booms and busts.)
But also, yeah if some magic breakthrough makes this a real "buggy whip manufacturer moment" and not just an illusion of one, I don't mind being the engineer on that side of it. There's nothing wrong about lamenting the coming death of an industry that employs a lot of good people and tries to make good products. This is HN, you celebrate the failures, learn from them, and then you pivot or you try something new. If evidence tells me to pivot then I will pivot, I'm already debating trying something entirely new, but learning from the failures can also mean respecting "what went right?" and acknowledging how many people did a lot of good, hard work despite the outcome.
- anon84873628 5mo agoI'm skeptical of LLM "reasoning" but they sure as hell know a lot. That's what the embeddings are: a giant semantic relationship between concepts.
- koonsolo 5mo agoI agree with you, but a big drawback is that the accuracy or confidence of their output can't be estimated. So they surely know a lot, but you are never sure if the info is correct or not.
- anon84873628 5mo agoThey can estimate confidence based on distances in that state space. But yes, it gets tricky.
- wiseowise 5mo agoEncyclopedia and Wikipedia know a lot too. Knowledge isn't much of use on its own, it's about how you use it.
- anon84873628 5mo agoWell Wikipedia can't write an essay for me, and LLMs can. I'd say they are quite adroit at using their knowledge. I mean, is Mythos finding all these vulnerabilities not evidence enough? Does AI Studio not clearly understand React and use it artfully?
- WorldMaker 5mo agoEmbeddings are still mostly just vectors into n-dimensional K-means clusters. It isn't "knowing" two things are related and here's the evidence, it is guessing two things are statistically likely to be related, based on trained patterns, and running with it without evidence. It has no "semantic understanding" as we would define it. It's just increasingly good at winning cluster lotteries because we've increased the amount of training data to incredible heights.
- anon84873628 5mo agoCan you explain how you "know" two things are related? If I ask you the similarities between a cat and a dog, is your answer based solely on an understanding of their genetic phylogeny and how those genes express traits? Grouping vectors in concept space is exactly how you create semantic understanding. The proof is in how good they are at creating semantically valid text. The fact that it took massive amounts of data is irrelevant. That just shows how much knowledge is encoded in all our language. It takes humans a ton of training to know things too.
- WorldMaker 5mo ago> is exactly how We don't know that. It seems like great hubris to declare we know how the human brain works. You are asking me to explain how we know things and then telling me we've already figured it out in the same breath, and that's hilarious. It doesn't take massive amounts of language data to train a baby human. It is almost entirely just: "Look. Here's a cat. Can you say cat? Cats go meow." "Over here, your aunt has a dog. Dogs go woof." There's generally a flood of non-lingual contextual data in such moments such as sights, smells, sounds, movements, touch but that also only further underscores how different LLM training is from anything we'd consider human learning. Our memories aren't just "conceptual spaces of linguistic topics", they are complex sensory maps where a smell can remind you of the first dog you ever met. There is so much of our human knowledge that is not and never been encoded in most of our languages. The fact that LLMs take massive amounts of linguistic data is relevant, because it shows how far we still have to go in barely scratching the surface of how the human brain seems to work. (Which again, we know only the barest details. Anyone who tells you they know 100% of how the human brain operates so far tends to be a snake oil salesman.)