This morning, the news broke that Larian Studios, developer of Baldur's Gate 3 and the upcoming, just-announced Divinity, is apparently using generative AI behind the scenes. The backlash has been swift, and now Larian founder and game director Swen Vincke is responding to clarify his remarks.
At a certain level, it is going to be a chore to determine who is or is not slopping up with AI media. Not every asset comes out with six fingers and a half-melted face.
I can see legitimate frustration with an industry that seems reliant on increasingly generic and interchangeable assets. AI just becomes the next iteration of this problem. You’ve expanded the warehouse of prefab images, but you’re still stuck with end products that are uncannily similar to everything else on the market.
And that’s before you get to the IP implications of farming all your content out to a third party that doesn’t seem to care where its base library is populated from.
At a certain level, it is going to be a chore to determine who is or is not slopping up with AI media. Not every asset comes out with six fingers and a half-melted face.
That’s a good point. I think a lot of people are dismissing AI content because there’s this fallacy and desire to believe it’s all “slop”. It’s willfully ignorant to wave it all away like that. Sure, we’ve all seen the stupid stuff, and it’s really annoying, but we absorb the good stuff without even knowing it. Anyone claiming they can reliably spot AI generated images is fooling themselves even at this early stage.
I’d like to know when something is real, especially real art and even pictures of nature, but I don’t think I can.
At a certain level, it is going to be a chore to determine who is or is not slopping up with AI media. Not every asset comes out with six fingers and a half-melted face.
Image/video diffusion is a tiny subset of genAI. I’d bet nothing purely autogenerated makes it into a game.
I can see legitimate frustration with an industry that seems reliant on increasingly generic and interchangeable assets. AI just becomes the next iteration of this problem. You’ve expanded the warehouse of prefab images, but you’re still stuck with end products that are uncannily similar to everything else on the market.
See above. And in many spaces, there are a sea of models to choose from, and an easy ability to tune them to whatever style you want.
And that’s before you get to the IP implications of farming all your content out to a third party that doesn’t seem to care where its base library is populated from.
Thier tools can be totally in house, disconnected from the outside web, if they wish. They might just be a part of the pipeline on their graphics workstations.
Keep a distinction between “some machine learning in tedious parts of our workflows” and “a partnership with Big Tech APIs.” Those are totally different things.
It sounds like Larian is talking about the former, and I’m not worried about any loss of creativity from that.
At a certain level, it is going to be a chore to determine who is or is not slopping up with AI media. Not every asset comes out with six fingers and a half-melted face.
I can see legitimate frustration with an industry that seems reliant on increasingly generic and interchangeable assets. AI just becomes the next iteration of this problem. You’ve expanded the warehouse of prefab images, but you’re still stuck with end products that are uncannily similar to everything else on the market.
And that’s before you get to the IP implications of farming all your content out to a third party that doesn’t seem to care where its base library is populated from.
That’s a good point. I think a lot of people are dismissing AI content because there’s this fallacy and desire to believe it’s all “slop”. It’s willfully ignorant to wave it all away like that. Sure, we’ve all seen the stupid stuff, and it’s really annoying, but we absorb the good stuff without even knowing it. Anyone claiming they can reliably spot AI generated images is fooling themselves even at this early stage.
I’d like to know when something is real, especially real art and even pictures of nature, but I don’t think I can.
Image/video diffusion is a tiny subset of genAI. I’d bet nothing purely autogenerated makes it into a game.
See above. And in many spaces, there are a sea of models to choose from, and an easy ability to tune them to whatever style you want.
Thier tools can be totally in house, disconnected from the outside web, if they wish. They might just be a part of the pipeline on their graphics workstations.
Keep a distinction between “some machine learning in tedious parts of our workflows” and “a partnership with Big Tech APIs.” Those are totally different things.
It sounds like Larian is talking about the former, and I’m not worried about any loss of creativity from that.