How long will it be before AI can generate AAA games without staff?
The art I’ve seen generated by AI is on par with seasoned concept artists and it’s generated in seconds for free. Surely concept artists are already out of work now.
AI could easily build a 3D model from the art, optimise it, rig it and animate it better than humans can.
What are your predictions?
I predict: ( drumroll ) - nothing.
AI is a regurgitation of a regurgitation of a…
The art generated by AI is already looking a bit tired and repetitive.
Don’t get me wrong, there’s a lot of places that will benefit from AI. You’ll be able to enhance fuzzy images, like they do in cop dramas, for instance. But for real. And many other things.
But AI taking over creative processes is something we don’t need to worry about. Why? Because it can’t create, it can only copy and regurgitate.
The main problem with AI is it doesn’t understand what anything is. So if you have an idea that isn’t already a ‘thing’, you can’t tell AI to create it because it has no references.
Unless we want to see fingers & toes 
AI could easily build a 3D model from the art, optimise it, rig it and animate it better than humans can.
Eventually, to an extent, sure, agreed. We’re not there yet, though. The problems arise when AI starts learning from another AI’s botched attempts. It’s already a thing.
build a 3D model from the art
And art can be described with language now. But Make AAA Game button will not be thing in the foreseeable future, mark my words.
And eyes…
Reminds of the Samsung Moon thing, where the software simply replaces parts of the photo with astronomically-correct imagery.
Imagine taking a pic of a license plate, only to have it messed up by processing.
AI is advancing exponentially, so the limitations we are seeing will be short lived. I’ve read predictions that there will be massive layoffs in the industry in the next 24 months and that’s just the beginning.
Why would a company hire people if they can get the work generated iinstantly for free? It will not take years to make games anymore.
AI will soon be able to calculate with quantum computing. Consumers will experience mind blowing VR simulations soon. I predict we will have to pay a subscription to access it and get inundated with microtransactions. It will be the end of stand alone games.
3D modellers, artists and programmers will be made obsolete. There will be new job vacancies, but they will be AI based.
We’ve been hearing about this for how many decades now? #RemindMeIn20years The progress is incredibly slow, tedious & prohibitively expensive. Rapidly-emerging is not what I’d call it. I wouldn’t even call it steady.
Same with high capacity batteries and cold fusion. It’s always been only 5-10 years away. For the last 50 years…
Ever since 1971, I think…
Where has ML been advancing exponentially? Everything has been throwing more compute behind still flawed methods and poorly scraped together models (how many models are part of this process alone now?). All of this combined with more make-up on the pig to make it look like it did a good job. Patchwork after patchwork “improvement” in community projects isn’t that either. It’s typically not better, only different or worse.
Some things can make solid use of ML. Most things can’t. Nobody will care if a lot of the tedium can be resolved by other means for things like 3D. So long as it’s doing it properly. Artists and programmers really aren’t going to be obsolete or go anywhere unless a very significant paradigm shift occurs that eliminates training. This is something only parasites or those unwilling to learn are going to parrot around. The virtually zero amount of interactivity in the process is it’s most severe flaw in any part of the creative space.
TBH it feels like “HD” Skyrim texture mods before ML upscalers were so widely used. It’s just another progression of the same flawed method. I’m not just referring to image generation either.
I should have known it would end up with a bot spewing chatGPT garbage and linking unrelated sites. We’ve been had. 
I’ve got a friend who wants to replace his cranium with GPT4… 
It will change Unreal Engine and the industry in ways it already has.
To throw a buzz word into the mix the has always been technology convergence (it’s a thing, Google it) where technology has always changed the way things have been done to make development of all things easier and faster and Unreal already has been converted from a simple game engine to a real time interactive environment that has made it’s way into the movie industry so the future there is a migration of media where film and video game development will share the same media pathways.
Cell phones became a thing when someone figured out how to insert an HD camera and GPS functions and people are sending their Iphones up using weather balloons.
As the saying goes.
The art challenges technology. Technology inspires the art.
Motion capture was once another thing that some say animators would loss their jobs but the truth as to history created more jobs based on just the need to manage all that data for the purpose it needed to serve. You can create this stuff but there will always be the need for the creative mind to decide what to do with it and more important is the result fitting as to the purpose needed to be served.
What will impart AI will make it easier for a couple of kids to produce Pixar quality results, costing million of dollars, from the comfort of their bedrooms as AI now does a good job now as to application generated results but as with motion capture will need the creative mind to decide what to do wit the results.
People don’t want to be replaced. Daily on Linkedin and fora I see people ask questions “What will it replace? When? Me? Is it a tool which requires a human?”. The answer is quite simply that tech can go way past human limitations. Its development takes less and less time. What surprises me is that in the past 10 years many posts I read online (of just anyone who believes anything) claimed that computers would not be able to be as creative as humans. I think right now we reached a point where AI is “creative” above the human average if we look at models which can come up with stories, names, paintings, music, summaries in just seconds. I also think that they actually have more trouble being specific, following very specific rules for a very specific complex situations. Right now you can’t ask any AI a question about law and trust it will correctly understand you, come up with the right laws for the right country exceptions and whatnot included and trust that it is more accurate than a professional lawyer. This is because most AI are trained by copying public data from the web. It could answer in an average way it learnt from Reddit posts without actually “knowing” what it talks about. This is a dangerous situation. But can it make a 3D model? sure. estimate bones and animate? Sure. The human tools are lagging behind AI development, I’ve yet to see a DeepLabCut integration with Blender or Unreal while I got research papers with relevant data dating back to the 90s, probably some older stuff too.
This is funny, if the mocap business wasn’t milked as it is, people would have moved to markerless tracking and poste estimation software already XD. Mocap suits, even on animals are obsolete but still sold and very expensive. It’s 99% a case of there’s money to be made so we don’t advance to the next big thing. The only thing we need now for accurate pose tracking and estimation is an average pc and a 15 euros webcam. AI models are pretrained and free to download.
What I’m actually expecting to happen in the next year is that the steps of 3D modeling and texturing will become obsolete, that we can pull interactive video in realtime from an AI. This would fully obsolete the steps we required for human development on art. No textures, no vertices, no UV maps, no bones.
For now chat gpt, or gpt engineer is like slash-dot website, but instant reply and zero toxicity.
However you also get really dumb answers and sometime spot on. And it is up to you to find out all bugs chat gpt happily created.
I use it, but to make all that silly code for arguments, classes etc. and simple functions. Then i fill rest myself.
Overall benefits for deep AI is that i make my code about 50% faster.
About AI future (until general AI that is as smart as we are). To make big applications, like game, you need coders anyway, because you cannot be sure AI created exactly what you wanted. So you need coder that can read whole code and fix it (current AI will make different variation of code with possible errors in another places).
And in big projects time spent to understand whole code and fix bugs, is usually longer than writing that code yourself.
Example with unreal:
TLDR:
As i wrote at beginning of this long rant. Deep AI is fast response slash-dot without toxicity. But as folks on /. AI does not care about your project, does not read (or understand) your prompt fully, and does not know/remember other pieces of your code. So in current state AI would make your game as good as /. community, just faster and without replies like: “learn to code!” or “google do you know it?!”
PS.
Sometimes AI works really well as “yellow rubber duck” you throw your idea at it and it will give you valuable feedback. So it is great tool, but that it is.
A better question to ask maybe is will gamers accept AI quality issues today, thanks to all the cost savings from layoffs? As industry executives don’t care about any loss in quality. 
Change (Winter) is coming, but the above example is a bit low-ball (a model isn’t a game). Whereas, pro-banker / pro-industry articles like this, show how the rush to AI is really creating change. Hint: Its more about cost savings and mass firings than anything else. 
So its not so much about when will AI be able to create AAA, its about what are the cost savings and how many humans can be gotten rid of right now, using AI as an excuse. As AI is still very much in its infancy. Its also a lot more expensive to run than most realize.
Example: less than 1 in 10 workers use ChatGPT right now, despite all the hype. Why? The larger truth is, AI cloud costs are huge, and so AI firms can’t afford to have 9 out of 10 workers using tools like ChatGPT every day. Not for super-cheap or free anyway. 
VC’s and banksters are to blame for a lot of the hype, but hey don’t let facts get in the way of making a good profit. But if you dive deeper and get closer to the truth, you’ll find that there are cold hard limits. Plus Stable Diffusion is in trouble right now, and there’s no GPT5 coming anytime soon, not without new ideas. But none of that is stopping the gold rush. As its about money (cost savings vs innovation). The really neat tools will take time, less hype. And they will still cost a lot of money to create and run, and so wont be available to those without a lot of money to spend anyway. The AI games mentioned above will run their course. Right now they’re more a newsworthy novelty than anything else.
Where things are more obviously problematic right now, is in the Healthcare and Film / TV industries. As bean counters have already accepted big compromises in quality for cost savings. Less pesky humans to pay. For most of us that basically means more human pain 
Just like the legendary ‘Make MMO button’, there’s no ‘Make AAA’ game button. That’s years away or never. AI is mostly about selling fear to enable more cost cutting and mass layoffs. As AI can only offer questionable quality right now. Look to industries from Film / TV to Healthcare to highlight problems / successes in AI in the short and medium term.
AI-generated AAA games without staff might be possible within a decade. While AI shows remarkable potential, the human touch in creativity and innovation remains invaluable. Predicting exact timelines is complex.
This was ai generated: Reeks of AI.
Only area where AI will excel is creating personally targeted spam.
In my opinion A.I helps a lot in development and make the task easy, like if a Games designed in a year with the Help of A.I the time become short and Game developed in 6 Months only, reduce the effort and improve the quality, less bugs and helps the developer in all sense.
Deep AI is great tool, speeds up working/coding.
And here is example of my recent work with it:
This is sector map for game i am working on, numbers are int32 IDs for cells in map variable. I needed function that gives adjacent cells to selected ID.
There are 2 cases: when column size expands and when shrinks. I described it all to GPT in prompt, and it created decent function. But it could never get few things right, no matter the prompts. It does not understand reason and how this should work, so it never could fix edge cases.
So after getting first prototype of function it was time to fix and code myself rest. Deep AI without real understanding what code is for cannot create whole game or application, and there is no way of explaining all nuances in prompt. You still need coder that can write prompt, then check code and fix it. Even for quite simple function like i needed in this example.
