Wacky!
riskable
joined 3 years ago
Don't call it theft. Theft is what happens when the original owner doesn't have that thing anymore.
Nothing was "stolen". AI just lowered the value of generic "thumbnail artists" and probably increased their value as artisans for those willing to pay for a premium service.
That's what generative AI does: It makes stuff of mediocre quality that's "good enough" for the masses. However, if you want something better than that, you have to pay a professional.
My hypothesis is that Baumol's Cost Disease combined with AI is going to result in the collapse of capitalism. Where bespoke service work (e.g. hiring a thumbnail artist) is going to become more expensive over time while generic stuff that doesn't need to be high quality is going to become cheaper and cheaper, to the point where it's basically free. We're almost there, actually.
For shore, you pelican't fish with them around.
You are the problem!
You know what you are, actually? You're like someone with an abusive spouse. "Yes, they beat me bloody from time to time—at the gas pump—but they make me feel good from time to time too. I love them."
Yes. This is how real estate markets work (or don't).
In other news, people who bought gas guzzlers when gas was cheap are now taking devastating pump receipts.
No doubt about tit, they're not just good, they're the breast in Texas.
No. LLMs and AI image models are fundamentally, a different technology. They use completely different mathematical models, data models, and target different platform spaces.
As a simple example, to train an LLM you need to feed it lots of text and not much else. To train an image model, you need to feed it billions of tagged images. That is, images that have every stupid little thing in the image "tagged" with a bounding box and text describing what's inside said bounding box.
Also, these terms might not mean anything to you, but LLMs use transformer decoders (autoregressive causal retention) while image models use diffusion networks. The way they work is different at a fundamental level. Text is 1-dimensional and an LLM acts as a word prediction machine whereas image models start with chaos (random garbage pixels) and refine it over and over again in n steps until the chaos starts looking like the prompt.
The way prompting works is also different.
Someone cut above the rest‽
Grimy already pointed out that this article is non-scientific nonsense but there's something else it's missing: The efficiency of AI over time.
Qwen-3.8-27B is due out next week and it's supposed to be on par with Claude (Sonnet, latest). That runs on a regular PC GPU.
Big AI isn't the only game in town. They're just what makes the non-tech news.
The open weights stuff gets drastic improvements every three months or so and there's breakthroughs in efficiency every six months (ish). We're only at the baby steps of AI tech and the advancement is behaving like a chaotic Moore's Law that leapfrogs itself regularly, then stagnates for a bit, then there's another big shift.
If this keeps up, the efficiency of AI will become something that makes using huge data centers for anything but training a waste.
Yes, but on the other hand, they get the mail before anyone else and those guys deliver!
The fresh water thing is wrong. The study that came out showing AI using shittons of water was including the cooling pond water at power plants. It was... 76% of the water use? I forget the % but it was so much that it completely changed the statistic from "holy shit" to "oh, who cares, then?"
Also, most of the mega huge data centers under construction are everywhere. Not "villages". Meta's big one is actually in the middle of nowhere.
Agentic AI isn't just "if else" stuff. It's way TF more complicated than that. It's so fucking complicated, they're terrified of implementing it at my work because they fear they won't be able to understand what went wrong when something inevitably goes wrong (LOL).
The ontology thing isn't really a thing. That's just what outsiders are calling some internal programming they're adding to LLMs so they don't hallucinate "obvious shit". The actual issue there is that, yeah: You can feed the LLM output back into itself four fucking times over to double-check it but that uses 4x as many tokens!
The thing you're missing is that token usage is exploding. It's like the world of AI has collectively decided that it needs that 4x token usage but they can't figure out a way to do that economically, so they're just sort of whistling while looking away from their balance sheets while at the same time getting seemingly endless loads from private equity idiots who think "AGI is just around the corner."