So here's the thing, generative AI is a technology, and it's created by humans, and I have no doubt that enough smart humans working together will easily be able to develop a technology that can generate a crochet/knitting pattern by feeding it other people's stolen crochet/knitting patterns without paying or crediting them. That's not a hard thing to do. What AI is doing is not actually hard.

The biggest problem with generative AI is that doing it efficiently is in fact very, very hard, and almost certainly always will be. The reality is that it's an absurdly inefficient way of doing anything, even just for model inference alone, nevermind the training data. Some of that inefficiency is being masked by the fact that they stole all its training data they used to create it. Some of that inefficiency is being masked by the fact that they're giving it away at a loss in order to get people hooked. Some of that inefficiency is being masked by the fact that the entire thing is a game of financial musical chairs that they won't stop playing because none of them want to be left holding the fucking empty bag without a chair when the music stops.

But none of those things reflect the true cost of generating stuff with generative AI. The wastefulness is the point. Imagine the efficiency if each person looking for a crochet/knitting pattern would be able to download the same pattern as someone else previously already did with only a few handfuls of bytes of internet traffic, not millions of tokens of furiously churning datacenter GPUs. Imagine if we had a simple text search engine where we could search effectively for the crochet/knitting patterns other people, real people, have already made, instead of having thousands of different people ask their own personal slop machine for slop a few dozen times until it gets it sort of right.

At a time when we are trying to save the fucking planet from burning down, they are very intentionally and very deliberately shoving this garbage, inefficient technology down our throats, replacing older, more sensible technologies, breaking those technologies to make them seem even less sensible, selling it like it's the second coming of god, and powering it with fossil fuels to further destroy the environment. This is not an accident. Hey ChatGPT, can you think of any of the richest industries in the world that would be happy to team up with some of the other richest industries in the world to increase demand for fossil fuels, even though we're not supposed to be using fossil fuels anymore, and can you tell me how much money they all made last year? I bet it's a fucking lot, isn't it? How can they keep making so much money when we're not supposed to be using fossil fuels anymore? Mysterious. /s

We're being fucked, the planet's being fucked, and the future's being fucked, and the people responsible are laughing all the way to the bank, then jumping on their yachts, and heading somewhere very similar to epstein's island, where they will party and do drugs and plan how they're all going to set up their new mansions on Mars once it gets so bad here on Earth that they have to leave us behind to burn. This is not really about AI, this is about people who don't consider most of humanity worth caring about or in any way consequential to their plans. We do not exist in the future they are building. We should probably find that very upsetting.

I don't seek them out, but I hear that LLM monstrosity patterns are largely a thing of the past.

The most plausible theory I've heard is that people realized LLMs are bad at writing patterns and so pivoted to feeding them other people's patterns to reformat or maybe grade for new sizes.

The new sizes may have stitch counts that don't quite work, but they don't seem to make the same ridiculous infinite increases/decreases that haunted the first generation of LLM-generated patterns.

Still, I doubt that an LLM could give you a reasonable pattern for any garment without directly ripping off an existing pattern.

Give him time. It sounds like he's working on the discrepancy right now.

Good point. With AI news we'll be able to trust it more because the message will be consistent across multiple sources and delivered reliably, compellingly, with conviction, and targeted to the current desires, fears and prejudices of the viewer. All of this will help to maintain a sense of trust, and shape your worldview.

Human beings are, for the most part, not mentally equipped to deal with this. We already see a significant part of the population lost in constructed realities. That's the tip of the iceberg.