There are so gosh-darned many free plugins for UE related to AI now, with yet-another-monthly fee for a paltry amount of “credits”. I think these sorts of things are neat to experiment with if you are new to AI, but do you really see serious lone-wolves, indie studios, or AAA++++ studios using a third-party platform and paying?
First you have model/agent/harness control issues. Most of these plugins are opinionated, which is to be expected of course, but it’s an important factor nonetheless.
Second, you are funneled into their ecosystem. Their models, their workflows, etc. Again, it’s to be expected, but also again, an important factor.
Personally I’ll stick to having control over my pipeline, if I’m going to be pair-developing with an AI fwend. I’ll choose the model(s) I want. The harnesses I want. The workflows I want. If I need it to be marketable, well then I’ll slap a UI on it, generate a website that has a list of features every similar platform has with a pricing table… including a free tier that gets gobbled up immediately because these platforms are always, always clumsy when coming out the gate.
While this may sound harsh, it is quite realistic and objective I believe. Most of us already pay for some sort of AI this-or-that, or have a very capable local, free, and secure environment with (again) models, workflows, and so on, of our choosing. Our opinions, our specific needs met. These platforms lack the necessary context tracking, domain experience (not data points in their training - actual tangible experience), and capabilities to creatively problem solve or develop novel features and functionality.
Can it handle the basics? Sure, I don’t doubt it. Can it handle all the basics, and have them bound together in a cohesive and logical way that is required of game entities, data, and so on? I’ve yet to find one that can, reliably, without requiring so much human-in-the-loop (or beyond) correction that… well, you may as well do the darned thing yourself!
So here is a challenge for both this, and all the other platforms like it: let’s see an end-to-end game created with it. Not just cherry-picked examples of feature X, Y, or Z; But all those features and functionality, and scenes, and effects, and… all the things that make up a game. We aren’t there yet - these platforms currently can whip up a quick material, or a class, or a scene… and generally do well at that one small and specific thing. But not all the things together.
I’ve gotten responses from other platforms on this front, and while marketing pitch leans towards “our AI will help you throughout your whole workflow/pipeline”, when you get down to it, they ultimately admit it really only does a few things. This is totally okay, mind you - it’s just that people tend to get excited and see the shiny thing and end up with misaligned assumptions.
Now where I take issue is, if at the end of the day these platforms are just a pretty UI, with an opinionated stack, that does one-off basics (as in, material creation, class creation, etc), what are we really paying for? If we strip away the UI and opinionated guts, wouldn’t we just end up with any other AI assistant that we’re already paying for - but we have to click the buttons? When I’ve tugged on the string, this is where I end up: we are paying you guys $9 or whatever a month, passing you our information and internals from our project, to have an agent in the editor do stuff.
I think these platforms are great for those indie types that are just getting started, have no intent to learn and master the various domain skills/knowledge required for game/interactive media development, and have money to burn. And hey, maybe that’s your target demographic. And that’s totally alright.
Take all this as honest, tangible feedback from a real person with actual experience over many years across many domains. It might prove helpful, as surprisingly I’ve talked with or consulted for others that managed to miss some of these thoughts in meetings/pow-wows/townhalls/etc…
“Oh wow. It is just a button clicker when you strip the distractions away. We need to pivot and reposition our goals if we want to provide the scene with something truly useful and valuable…”