Good comment. Progress is based on team work, nobody is genius enough to do it alone. And I agree management is stupid. Just look how Meta is fumbling with LLMs, how the Chinese Qwen and DeepSeek do laps around them. It was a clear management issue and the poor guys became the scapegoat.

I doubt the ML engineers were not good enough since they were hired by Meta and worked there for years. More plausible is that they overspent on VIP hires and then decided to cut from other ML roles. I am curious to see in 6 months if Meta can catch up. Anyway, the majority ML work nowadays is not training models, it's prompt and agent stuff, this type of work is complementary to model training.

I think you confuse desire to lower costs with desire to increase profits. While the first might lead to increased profits on short term, companies don't care about costs as long as profits expand. If your company reduces costs 50% by using AI instead of human labor, my company uses both humans an AI and expands 10x more. In the end your company might lose.

Why make this about me? I asked a legit question. It is easy to find anything online, and see the original. Most of it is officially from the source, but there are a lot of "piracy" sites too. So if infringement was the goal, it would be easier to just get the original. It is fast, free and perfect fidelity. On the other hand generative AI will not replicate works closely, it costs money and is slow. Imagine generating bootleg Harry Potter, it would not work out, it would cost a lot, and take a long time. So generative AI is the worst tool you could have for infringement. We use it to create other things not substantially similar works, because perfect replication is already easier and cheaper.

A good take, I agree. Let's say we generate 100 pics privately, and as most gen AI images, only see them once and promptly move on. Is this private use infringing? Even if it doesn't look substantially like anything in the training set?

How come? If we wanted to infringe on works would just .. you know .. download them? Using AI means we Don't want those original works.

Even so, infringement was supposed to be based on "substantial similarity", not "training". What they mean is to drop the similarity test and replace it with a simple "causal link". A broadening of copyright that will hurt creativity from now on.

We have been conditioned by the all mighty Algorithm to produce slop that will get views on their sites, because they "own" all the public attention. Platforms exploit both ways, up and down, like ISPs under non-net-neutrality rule. We need to force these platforms to prioritize their own selfish interests less. Why do you think we can't find a decent app in App Store? Because in-app payments make more money for Apple. So they tanked the whole App Store to milk us of money at every corner. Why can't we find what we look on Amazon? It's on page 7, first 6 pages are full with shit they want us to buy instead of what we are looking for, or the best deal. Same for web - why are the top 10 links junk? More money for ads.

In fact articles like the one posted here are how humans organize to generate human slop. Someone telling people what to think and do. Then they all parrot the party line.

It's AI slop even if it is written by a human, if it comes by algorithmic feed. Because people who get good positions in those feeds cater to the algorithm intentionally by making slop. Most slop is written by SEOs and professional writers fighting for the scarce public attention.

I think the majority of AI text people read is their own AI text. There are already 1B users. OpenAI alone has 800M. Seeing clear AI generated text in the wild is rare, seeing your own chat is frequent. I think most AI usage is a private affair. Almost all LLM text is read just once. If people complain of AI slop, then it's their own fault. You have to chip some original ideas in, can't expect the model to be original. But there are also times when the model surprises you with a really useful idea or connection. I see it like a jam session, your ideas influence its ideas, its ideas influence your ideas. Together they get in a place neither of them would have gotten alone.

> it's most of the time useless crap from start to finish.

Maybe.. you forgot to put any ideas in the prompt. That would explain it.

AI has no skin, can't be sued, punished, shamed, or imprisoned. But managers on the other hand ... hmm they are also safe. Never mind

AI will read the texts you provide, write generic text on any topic, if you don't bring anything original don't expect anything but slop. On the other hand, if you have plenty of original and useful ideas to communicate, why not use AI as your wordsmith?

That explains why management loves AI so much.