Hating AI

[BlasteroidAli]

AI...as if us humans need more things to make us even lazier!

Song about AI in the style of New Model Army. https://suno.com/song/9a6b7463-4912-4f58-aed1-a12f02feb4dc

[C64_Critic]

Start serving them up a healthy diet of zipbombs:
https://idiallo.com/blog/zipbomb-protection

[newmodelarmy]

AI...as if us humans need more things to make us even lazier!

Song about AI in the style of New Model Army. https://suno.com/song/9a6b7463-4912-4f58-aed1-a12f02feb4dc

Cool song, thanks for sharing. Very spot on!

[ActRaiser]

Start serving them up a healthy diet of zipbombs:
https://idiallo.com/blog/zipbomb-protection

That's brilliant!

[k8track]

I really despise A.I. 99 percent of the time, except for that one time when I got it to render Gary Coleman as Galactus. That was awesome.

[wur]

I'm a computer scientist by trade, but I hate AI. I'm not saying it can't serve a useful purpose, but the manner in which AI companies are vacuuming up the data from all the legitimate web sites without regard is distressing.

So here’s an idea:

Put random cookie recipes in your reviews in a color that matches the background and a tiny font size (so normal readers can’t see it other than a gap). This will be stripped of all the font metadata add be blindly trained on.

Maybe us technically inclined readers can all pitch in for prizes for each time someone finds your review in a particular LLM model—wouldn’t be hard, just iterate thru a list of old games until a cookie recipe comes out. (The prizes should be nothing big — maybe a 6 pack of craft beer or something)

For each model, you can put up a notice on the front page—“as seen in LLaMa version 5.1!!” Then link to the proof of this)

Might be a way to have a little fun and a few laughs with this phenomenon at least.

[Miles Prower]

They'll claim that AI art is theft because it uses existing art as a reference, but those same people have never had a problem stealing another artist's characters and style to make money off fanart commissions. But now that I can get a picture of Cloud Strife riding a unicorn to McDonald's for free, it's suddenly a sin against humanity.

I know exactly what you're talking about. So many people draw a picture of Sonic and parade it around on Twitter or Blue Sky like it's theirs. LMAO! These people can only make things that are from something that already exists and is well established. You're not in 4th grade anymore drawing doodles in your notebook, pick up a pencil and paper and draw something that is wholly unique and original.

[Stalvern]

The issue isn't that Cloud Strife is a copyrighted character. The issue is that Stable Diffusion or Midjourney or whatever is itself a mountain of people's works collected and profited from without permission or compensation. People who don't know what they're talking about call it plagiarism when it spits out fake Ghibli pictures, but the actual problem isn't that far removed.

My biggest problem with it is that mindless and useless auto-generated slop is already crowding out actual content. If I search for a picture of a horse, I want a picture of a horse, not a fake picture of a fake horse with an extra hoof, and not even a fake picture of a fake horse with all of the right parts in the right places. I don't want a fake article written by nobody and telling me nothing when I want to learn something about horses. Even if I did want junk like this, why wouldn't I just prompt it myself? I don't want an Internet of worthless fake garbage that exists to piggyback ads on for minimal investment, but that's what we're getting.

[m0zart]

I'm a software architect, and I've designed and implement AI solutions a lot over the last decade. But in all of our cases, it was focused on very industry specific use cases, and limited to customer data that we already owned (via contracts) for that particular use.

Yet I also have come to hate AI. But I've come to hate it for other reasons. Product managers here at my company have been floating around this idea of replacing developers with AI. I wish they would listen to my warnings. It's not as easy as it sounds when the problem itself is very complicated. Sure, it can come up with quickly developed solutions for small problems in almost any language, but it's terrible at making large working systems like the ones we make.

Even to get it to achieve a proximity to the intent, you REALLY have to spell things out for the engine. And frankly, our product managers barely know how to state requirements at a high level. I can't imagine them interacting with AI to the degree necessary to spell out issues and then test to make sure they were met. Talk about setting yourself up for failure!

[m0zart]

My biggest problem with it is that mindless and useless auto-generated slop is already crowding out actual content. If I search for a picture of a horse, I want a picture of a horse, not a fake picture of a fake horse with an extra hoof, and not even a fake picture of a fake horse with all of the right parts in the right places. I don't want a fake article written by nobody and telling me nothing when I want to learn something about horses. Even if I did want junk like this, why wouldn't I just prompt it myself? I don't want an Internet of worthless fake garbage that exists to piggyback ads on for minimal investment, but that's what we're getting.

Yes that is happening. And it's causing AI engines themselves to "learn" from that data, meaning that the AI engines are producing in some cases less accurate content than they were prior.

In a way it's pretty comical.