While the game certainly isn’t vibecoded, I would bet some Pocketpair devs used an LLM for a Python script, to figure out something in documentation, to point them somewhere over an error, maybe some Windows issue; you know, utilitarian things.
Artists likely use oldschool ML models in their graphics software, without even knowing it. Or maybe when processing textures to finish them. Or to search through assets, or documentation.
I’m just saying, if you’re strict with the definition, it would be really hard to block an entire dev studio from all “AI.” It would almost certainly seep in from casual use, legacy integration, or “oldschool” things like image recognition and processing.
It doesn’t mean they have to vibecode or ship slop assets, of course. When they say “no AI,” that’s what I hear, and exactly what I want.
But I think studios, especially larger ones, need to be careful about labels like “100% human,” lest something come to light that seems to contradict the guarantee.
I imagine by “AI” they mean strictly LLMs from OpenAI and other related sources. It’s more or less a consequence of the term being used as a catch-all by marketing teams even if it’s misleading.
They’ve for sure used machine learning tools in their development process, I have no doubt about that. I mean, that’s literally what code completion is. That’s been a thing for much longer than LLMs though, and it’s really only a supplement to the actual intentions & efforts of the devs using it. Stuff like that is useless if you don’t have any game development expertise.
Correct me if I’m wrong, but it doesn’t seem like the interviewee was ever asked about tools like code completion. The article focuses almost exclusively on ‘generative AI’ and PocketPair’s attitude towards its usage in game development. I don’t think it’s fair to consider it a dodging of questions if said questions seemingly weren’t brought up to begin with.
Besides, I doubt non-devs even really know about/understand machine learning tools in game development aside from the standard AI slop machines. For them code completion probably seems more like a standard programming tool as opposed to something more akin to an LLM, especially since most people already know about things like autocorrect through their phones and the like which is pretty similar.
Where is the line of “no AI” exactly?
While the game certainly isn’t vibecoded, I would bet some Pocketpair devs used an LLM for a Python script, to figure out something in documentation, to point them somewhere over an error, maybe some Windows issue; you know, utilitarian things.
Artists likely use oldschool ML models in their graphics software, without even knowing it. Or maybe when processing textures to finish them. Or to search through assets, or documentation.
I’m just saying, if you’re strict with the definition, it would be really hard to block an entire dev studio from all “AI.” It would almost certainly seep in from casual use, legacy integration, or “oldschool” things like image recognition and processing.
It doesn’t mean they have to vibecode or ship slop assets, of course. When they say “no AI,” that’s what I hear, and exactly what I want.
But I think studios, especially larger ones, need to be careful about labels like “100% human,” lest something come to light that seems to contradict the guarantee.
I imagine by “AI” they mean strictly LLMs from OpenAI and other related sources. It’s more or less a consequence of the term being used as a catch-all by marketing teams even if it’s misleading.
They’ve for sure used machine learning tools in their development process, I have no doubt about that. I mean, that’s literally what code completion is. That’s been a thing for much longer than LLMs though, and it’s really only a supplement to the actual intentions & efforts of the devs using it. Stuff like that is useless if you don’t have any game development expertise.
But see, even in this interview, the lead is dodging code completion questions. As non-devs could interpret that as “AI in the game.”
Correct me if I’m wrong, but it doesn’t seem like the interviewee was ever asked about tools like code completion. The article focuses almost exclusively on ‘generative AI’ and PocketPair’s attitude towards its usage in game development. I don’t think it’s fair to consider it a dodging of questions if said questions seemingly weren’t brought up to begin with.
Besides, I doubt non-devs even really know about/understand machine learning tools in game development aside from the standard AI slop machines. For them code completion probably seems more like a standard programming tool as opposed to something more akin to an LLM, especially since most people already know about things like autocorrect through their phones and the like which is pretty similar.