Yeah I mean, that sounds reasonable. There is a big difference between generating all your game assets with AI and using Claude to refactor methods and write docs.
post
Jesus christ can we please stop trying to automate art and content creation
Tools like SpeedTree won‘t get dropped because the efficiency gains are enormous and the downside negligible.
I mean, if were talking about efficiency tools for artists to use rather than straight up automatically generating the assets (not sure what those tools would be at the moment, but I've not been following what the AI industry actually releases for awhile because it's always seemed a bit useless), then the result should be an increase in the output of those artists rather than replacing them with statistical amalgam.
It's to cover things like code completions built into the development tooling, or asking Claude/ChatGPT questions.
It covered everything before, and unless you're a solo dev, not using anything at this point within the entire game studio, is pretty much impossible.
And even as a solo dev truly trying to not use any AI even googling anything and reading the AI summary at the top would qualify you as having used AI and required disclosure. It was unenforceable and everyone would lie or were lying to themselves.
This is a much better policy and more enforceable.
As long as it is used as a tool that has human refinement as part of the process it would be comparable to CGI replacing background matte paintings and motion capture replacing manual manipulation of CGI to make movement. Gollum worked because of the blend of technologies that were replacing existing practices, but as a new approach and not a cost cutting measure.
The problem is entirely about using the output directly as a replacement for humans to cut costs, not having another tool that can be used as part of a process. This is coming from someone who absolutely hates LLM and genAI slop, which is taking the horrible approach.
That is, in fact, not the problem in its entirety.
The move to CGI didn't require stealing the artwork of the matte painters. The move to mocap didn't require raping the land of all its water. The move to either didn't require all of the world supply of computing power, leaving it only affordable for the world's richest. The move to either didn't create a corporate circle jerk that the damn near whole world economy was propped up by.
So the thing is, the server farms for doing Toy Story and ILM server farms for rendering CGI were and are massive, but they were built over time and focused on a particular purpose. The same thing can be done with localized models of LLMs and genAI.
The giants who are buying all the parts and choosing to strain the grid and add polluting energy methods in order to stuff absolutely everything into massive all in one models are a related, but distinct issue, with different goals that encourage shitty practices. Yes, if a game company used the LLM and genAI slop producers tools that is a negative. If they use homegrown or at least dedicated models that are closer in scale to what is already used for CGI then it isn't automatically a negative.
It is like advertising. A little is fine, because awareness is needed for people to know something exists. Massively invasive methods of jamming advertising into literally every moment of the day is a problem. ChatGPT and OpenAI are the latter and a problem. Or how Nestlé doing literally anything is horrible even though other companies do the same thing without being nearly as horrible.
I would prefer to know what AI tools were used, so that I can avoid the ones using the AI slop machines that are a negative.
If you don't like automation in video games then I trust you're opposed to IntelliSense-style refactoring, IDEs in general, and in fact that you work through every single instruction executed by the computer in your own head.
And there's the crack in the door needed to slither the rest of the way in. First they come for the copy writers and concept artists. Then it'll be writers of "boilerplate" code (I know it's already happening). Soon enough the disclosure will quietly disappear and we'll be completely drowned in slop.
The Hottest of Takes:
If we're talking artistic credibility (as opposed to job security, plagiarism, and environmental impact), I want anti-AI people to uninstall their desktop graphics applications like Photoshop and GIMP. If you depend on buttons, value inputs and algorithms to get the art you want out of the machine, as opposed to using an easel and scanning your work into the PC without minimal touch-ups after the fact, then you're no better than the person typing book-length prompts to get what they want. If you animate with key frames instead of hand-drawing every frame, you're likewise just as credible (or not) as the prompt jockey. Hell, if you at any point use CTRL+Z, CTRL+C, CTRL+X or CTRL+V, you're as artistically incredible as Paulie Promptnuts.
Just to be clear, I don't think any of those things. But if you're dismissing art on the basis that AI was used at some stage in its development, you should be thinking those things.
As I am replying, you got 9 upvotes and 9 downvotes; looks like the perfect "storm" to put my hot take too.
We got psychotic people on both side, when you get this grade of polarization people usually lose the perspective.
AI is a technology, an human logical entity like math: AI works on very advanced (probabilistic) math. Math is not the evil... but an actual evil does exist.
There's a difference between a LLM chatbot that runs on your local GPU... and one in the cloud.
The chatbot on your GPU is "trapped" by your questions, your needs, your choices.
Today the chatbot on the cloud will tell you that Elon Musk is a controversial person, tomorrow it will tell you Elon Musk is the savior of the Earth and you're not worthy to kiss his feet.
People seeing absolute evil in AI, are against you running your chatbot locally, on your PC.
People enthusiastic about AI will accept any "gift" (or AI GF) Elon Musk will give them.
no,
if someone learns art on digital systems they can grab a pencil and do the same on paper. maybe they'll be annoyed by lack of the undo button, maybe they'll have to learn colour mixing, and how materials interact with each other, but the core ability to make art is fully transferable between ditigal and paper
same goes for animation. btw. your statement about it doesn't make much sense, keyframes are a concept used by both digital and traditional animators. and if you meant frame interpolation then it's a brute force calculation of the most average of averages given two data-points, 90% of the time the animator has to go back and fix it, developing their animation skills that they can then take to paper and do just the same (would just take way longer)
now what will a prompt typer do without their AI?
fuck all is what they'd do. the only transferable skill from that would be idk writing image descriptions for the visually impaired
Speaking as someone who dabbles with all kinds of art.
What are you talking about? I can’t even begin to make the art i make digitally on paper and i am using a drawing tablet. I heavily use blended layers and may zoom in to the pixel level to meticulously get something perfect.
My used to be art teacher (pencil drawing, digital drawing and tattoos) started lesson 1 with affirming that digital painting is not drawing and you can not use a drawing tablet to draw. They affirmed its its own thing and much closer to painting.
You could raise one point for eye hand coordination holding a pen like item to draw lines there are transferable skills there but for all my digital work, my paper drawings suck, and i don’t even want to think about that guy i know that was actually good on paper, but disliked using a drawing tablet so much he used a mouse to “scratch” drawings on pc instead.
If we are considering experience broader creative techniques there is also the very common (and very fun) art of making collages.
People used to do this with magazines and you needed to be good with scissors. Doing it online eliminated that skill yet public internet images and photoshop actually turned that hobby in an actual career. (I know many magazines where every cover is a photoshopped collage of pictures)
I like to to point this one because in my Own dabbling with ai generated (diy comfyui -not an llm prompt) the creative energy i put in feels much close to making collages, and a little bit like coding. You start with something basic and then you gradually change and mold stuff into intention.
My actual critique of ai generation is actually something much more basic.. its just not as fun, its carving a rock that can change shape and sometimes it just sucks.
i meant the knowledge of line weight, colour theory, perspective, shapes, how light works, where shadows fall, how to treat your values, and the million other things that you can take to paper, that i have taken to paper as a ditigal artist
drawing ditigally is one way of drawing sure, but the skills you get as you draw are 90% transferable to any other medium - you just need to learn the new tool first
obviously it won't be the same, but you'll still be able to use the knowledge and skills you've obtained drawing digitally - to help you put the idea in your head onto the paper much better
and AI prompter would be on the same level as someone who's never touched a pencil before
I feel like your being incredibly specific about how it applies to one art form.
Ai generation is not drawing. Just like drawing is not woodcarving.
Id like so see pencil or brush skills transfer to photography. Which is a good example because i have taken photos with intentional effect to make them look paintings.
Someone who works abstract without light play does not need to know light theory to be an artist.
If you want to do make the same art on a different medium sure knowledge is transfered but that really depends on what you are trying to do which is different for every creative individual.
In my experience there is knowledge overlap between making photoshop collages and building advanced ai generation flows.
These tools also complement eachother. Generate a specific something, photoshop it into the final product.
overlap between drawing and photography? easy, composition, colour theory, framing, in fact you could take everything (aside from gear knowledge) from photography and apply it to drawing or painting
pretty much all mediums of art have the same basic core - concepts that can come in handy no matter what other medium you grab, shapes, colours, compositions - you learn those by working with any medium, and you can take them everywhere else
The same concepts that also overlap with designing ai genres art. which is my pointW
Framing is set manually, different generation zones and is quite literally composition.
Adding skeletons with control net to fix a pose, again composition.
Choosing one of 100 different generated sky backgrounds is easier if you know color theory and recognize realism or lack therefor.
Pretty much all mediums of art have these indeed. So does comfyui. Ergo creatively designing ai generated images is art.
let me know when AI actually makes something new instead of mass plagiarising actual artists' work (whilst burning through tons of water)
because at the most generous you could call what AI does a "collage of anonymous artworks guided by a user's input", but in reality, it's just stealing, and you don't even know who's art you took
Easy.
Don’t use a public ai model you download from the internet or use other copyrighteded work to train your own model.
You dont really need or even want that much data to get a focused result. The technology is open source and gaming hardware can handle it.
Just because 99.99% of what you see is people doing something dumb and unoriginal using tools that are designed by evil corporations does not mean the concept of what there doing is wrong and that there are not any people who do it right.
The problem right now is, try to find artists who do it right within a climate where everyone will discredit your work out off slop bias.
Consider that most people who do art do not have a public artist identity and just do it for their own private passion (best art) those that use ai definitely have no incentive to share their projects with the world.
Which is a shame making sure that the only works you will see is slop engagement bait
ah, i assumed you were talking about public models. the sort of stuff that you specified now is fine in my books. might be a good idea to mention that you don't mean the slop making AIs from the start, since that's the first thing that comes to mind when someone says "AI art"
I used autocorrect to write this sentence, which is a language model trained on copywritten works. It just so happens to have been developed in the 2000s instead of the 2020s.
that language model doesn't regurgitate the media it was trained on. it just makes sure you don't misspell your thats
There are definitely lines being arbitrarily drawn around AI, and there's not a sensical flow to what any individual might believe is an acceptable AI use case. Nobody has really sat down and made a full documentation of every modern AI use and weighed the benefits and detriments they have on society. I think many of the outward haters of AI are likely just as ignorant as the blind defenders of it.
top 81 comments