The conversation/free-for-all around the role of automated “AI”-based game development rolls on with a few thoughts from Tom Hall, co-founder of id Software and one of the creators of the original DOOM, who says he’s (Commander) keen on the prospect of “ethical” uses for such tools in gamedev, but worries that reliance on them “will homogenize games, sort of like AAA games are now”.
Speaking to Sektor.sk, Hall said he was “excited” by “how AI could be used ethically to be more of a core element of the game, so it’s almost like a game that you’re playing and it’s playing you, in a sense, or it knows what you want. It could generate things for you, or enable different gameplay, it can adapt much more seamlessly to what you’re doing, or just sensibly create more game content.”
But he added: “I don’t want it to just willy-nilly be procedural, everything AI, and just not have any crafting to it, because that will homogenize games, sort of like a lot of AAA games are now. They’re just kind of like I attack the monster, oh, it’s attacking, I’ll roll out of the way. It’s all kind of the same stuff. And that’s what I don’t want to happen to games because of AI. I want it to enable us to make cooler things, and more amazing things, but there still needs to be a sense of craft.”
Is there any ethical AI, all they do is take data people posted online and then profit off it. With the original creators not getting a say if their data gets used or any profits derived from it.
If your training is from in house sources possibly, although then you’re limited on sample size.
We did this in a previous org. Basically, we had a bunch of user-generated data, users would then classify a sample of that data, and then we’d train our model on those classifications.
I don’t see how it would work in game dev though, unless they’re using AI to customize an NPC’s behavior based on the player’s actions (i.e. teaching an enemy to block player attacks). Generating models and whatnot would just have too small of a data set to work with.
arguably no?
Though Getty did introduce their new AI today that was only trained on images they own the copyright to. Arguably, still not ethical, but at least it’s things they own the data for.
Can you verify their training sets or do you have to take their word on it.
I didn’t dig too much into it, but my guess would be no.
Even if you could verify, it’s still an ethical grey area as it’s taking works they paid photographers to generate new works potentially without crediting the original photographers? Their own website tells people they have to credit the original photographer, and I’d be surprised if the AI lists all the works it used to create it.