https://lemmy.ml/post/31490862 pretty interesting article linked in this post, tl;dr researchers tried to get AI agents to run a simulated vending machine (which, let’s be clear, is a solved problem and can be done with a normal algorithm better and cheaper) and it didn’t go that great. Even if some of the test runs actually managed to earn money, they mostly devolved into the AI becoming convinced that the system doesn’t work and desperately trying to email someone about it (even FBI, one memorable time). I think it illustrates quite well just how badly things would go if we left anything to AI agents. What are the odds anyone involved with pushing autoplag into everything actually reads this though…

Yeah, that video was rough. I wasn’t actually expecting him to be super lefty or anything, but the “both sides bad and should talk more instead of screaming” take was still a blow.

Good news: I’m sounding less and less like a conspiracy theorist nowadays. Bad news: see above.

Ah damn, I have it and enjoyed it a lot - it’s fun! The card assets have always been obviously generic, store-bought assets… but at least they were still made by humans. I remember the first AI controversy in that game, I think it was centered around the main menu’s background image, where the castle turrets and towers kinda didn’t line up.

I don’t know how I feel about this, on one hand this is such an incredibly niche product that I can understand wanting to save every scrap of money, but it sets a bad precedent going forward.

Name and shame? Which game?

Musk got banned in Path of Exile 2 for cheating. I’m not sure what angle to take here, but you gotta admit that it’s a bit funny/satisfying. (how does such a busy [assume I’m making air quotes with my fingers] guy have time to play video games? why is he so obsessed with status that he’d try to cheat his way up the leaderboards, and not for the first time either?)

I’ll bet quite a lot that they’ll learn the exact wrong thing from this and decide that they just weren’t republican enough.

So how many ChatGPT 4s have they precariously stacked up on top of each other this time?

A seemingly small detail that really bothers me about this story is that those glorified chatbots were given backstories like they are actual humans - in their bios they have things like “an acoustics engineering student at AGH [local university]” or “passionate about queer issues”. And… can a chatbot have a student id? I somehow doubt that AGH has ever issued an id for a bot. Can a chatbot be “passionate” about real issues that real queer people face (especially in a country like Poland which, while it’s been getting better on LGBT+ rights, is still one of the worse places in the EU to be queer) and should it pretend like it has faced these things too? That seems like a horrible idea.