Do Our Opinions of Generative AI in Games Really Matter?

General
1

Games That Use Generative AI Are Selling Really Well On Steam Right Now

A quote from the article:

It does feel like, whether we like it or not, the use of generative AI in video game development is here to stay, particularly if those games continue to top the charts despite backlash.

As we’ve continued to talk about LLMs on these forums, I keep noticing that while we have strong opinions on the subject in both ways, paying customers don’t really seem to care as long as they like the game.

Skele-Tom

I made a game jam game, Skele-Tom, completely with LLM generated music (using Suno). I got 70 people to rate my game, and 66 of them left comments. I was very clear about my use of AI music. Despite me using AI generated music, the reviews were generally positive.

As much as I hate AI and I will actively want to punch you in the ballsacks for using it, no matter what I’m about to say next and there’s no changing that fact . . . the main theme coming at me out of left field was a pleasant surprise and added more to this game’s production-value! (even if something made with AI cannot be counted as production, for obvious reasons). This game’s music is good, the title theme as well as the theme(s, mutiple variations apparently which increases the game’s quality even more) in-between stages is nice and catchy with coherent lyrics for AI. The game looks nice, even if (if I understood the credits correctly) all of the assets in this game are premade
It really pains me to say this, and it speaks to the quality of everything else I’ve witnessed in this Jam so far, but this might be the best, most stable experience I’ve played here so far

(Link to Discussion)

Without soapboxing too much.. man. I’m glad you were upfront about it, but it did bum me out pretty hard that the music was AI generated. People have opinions and I respect that and try not to judge. The song was a banger. However, as a music educator and composer myself I’d be lying if I said that knowledge didn’t make me a little sad.

(Link to Discussion)

The music is super cool and captivating. It was holding me in levels that I didn’t want to end just to listen to this piece :smiley:
Could you share this song?

That background track was a banger!!

Thanks to you I will now sing “Skele-Tom” all night :smiley:

The intro is just amazing, the TV effect is great, and the music—although I don’t encourage AI-generated content—is spot on.

the music is a absulute BANGER you should post it on spotify or youtube at least

The Rest of the 26 Comments on the Music

i love the music and the changing camera view! the game is very impressive! also your splash screen is really good.

bonus points for music, so what if its ai it certainly makes atmosphere twice as good
amazing what you was able to achieve in the time of jam solo

Really loved it. felt pretty polished, in my playthrough, The suno song is pretty funny… this will def- be high in the jam list for sure

A very polished platformer with a very charming art-style and catchy music!

Also, I really liked the music — I actually stayed on the menu just to listen to it hahaha.

Music and sounds are really cool, very Halloweeny, the song is just a banger, I really loved it, and it made my day :slight_smile:

The music is something else!

  • Loved the music.

It’s a very creative and fun game, and I liked the music!

That was fun, the music was great and suited the game well despite being AI generated.

I really like the job you did with the music and sound design and visually this was a treat to play.

very nice music

Musics great

Wow, did you produce a full theme song… and and 8-bit version of the theme song for this jam?! I’m impressed!

Awesome music

The soundtrack is on point too, awesome job!

Music and sound effect feet very well the game.

haha, fun game! liked the AI music and the effects!

The theme song was a nice touch!

honestly the music gets you interested at the beginning, but then its AI so that’s a bit of a bummer

epic, loved the music

I loved the music.

Killer music and aesthetic. Great job!

The music was SICK. Very creatively made.

Great Music and sounds.

I like the music in the game and the CRT effect on the screen is very cool. especially the music, it’s an absolute jam .

Granted, this is a single, very limited study with other people who make games, but of the 33 comments I got - all of them were positive about the music itself. Only 5 were negative or disappointed about the fact it was AI-generated. The game itself came in 12th, and so did the music.

image
image286×318 13.3 KB

Does AI in Games Matter?

While I think that there will be No-AI games in the future. I believe they will be bespoke things. Like going to the Farmer’s Market for locally grown food, home-made cheese, or one-of-a-kind pottery. I do not think most people care.

Generative AI is getting very good. And yes, I do think there’s a difference between AI slop and human-created art. But when humans start using AI to create their art - I think the lines are going to blur even more than they already have.

97% of people cannot distinguish AI music from human music. That’s getting into the territory of statistically insignificant. Where those who can distinguish may actually have just been guessing. Either way, that’s up from 95% in the last study done a month ago. (I cannot find the reference but I believe I posted it in another thread.)

The same is going to be true of well-made games. If the game is fun - a lot can be forgiven by the average player.

3 Likes
2

I know no one cares, but it’s still a bummer.

3 Likes
3

Music is easy to produce and sneak in as it’s a completely independent part that just sits there and plays.

Producing consistently good graphics with “ai” is entirely different matter.

4

For now. But I think music is just the leading edge. Graphics are going to get better.

5

Define “get better”.
Game graphics are much much more than just images. And the only thing generative “ai” can deal with is pretty much only image synthesis. Fancy looking things like gaussian splatting are pretty much useless for games.

1 Like
6

There is a difference between good use of AI (optimize code, learn stuff, etc.) and bad use of AI, which is the majority of the use (brainrot, AI “art” for the games or other stuff, etc.) I have used AI, I will admit, but from now on I will try to avoid AI as much as possible, why? Because the AI I used AND ONLY ONCE BECAUSE I WAS USING ANOTHER AI, broke all my weapons, I ask Gemini to make my player and weapons script better, it did, but remove essential stuff of the gameplay and make weapons unusable, It helps me fix some problems, but it created more, AI (especially for not so big engines like Godot) it’s not that good, it can help you sometimes, but it also makes things worse if used badly, you need to give it lots of details, if not, it will not follow your way of writing, for example, I use a weird syntax that is “var_EXAMPLE” the first word in lower case and the last one in capital, and AI will just go to the internet and take whatever it finds and uses it for the code, also, AI is NOT more efficient than regular humans, because AI uses a lot of resources to make code, the code may be better than a human one, but the human one will not waste thousands of liters of water and lots of watts.

Sorry If I didn’t follow the topic of this, I just wanted to point out something’s of AI.

Also following the topic of AI generated music, in my opinion, it’s just easier to make a synth based OST than a real instrument OST, for example, if you ask the AI to make a song based on electronic music, it will find lots and lots of song with diversity and make one, and in my opinion, music using synths it’s not that hard, I made a song using mostly synths, and it turned out pretty good, but, for other genres, rock (which it’s what I like), metal, and similar genres, it’s harder, guitars are not easy to program virtually, and (although, it also affects synth based music), there are more boring generic rock metal songs, (personally, since I’m a beginner guitarist) it’s harder to come up with good guitar riffs, and AI doesn’t create, it takes something as a base, leaving more generic stuff.

Sorry if I got things wrong, this is my opinion and I don’t know everything, point me out something if I got it wrong.

2 Likes
7

The only remedy for deluded and misplaced “ai” usage is education. People need to learn how LLMs or diffusion really works.

2 Likes
8

LOL if I had a crystal ball, I would. TBH, it’s just a feeling. I thought gen-AI music was pretty shit when I started with it. It will never replace a good music producer, but it can provide “good enough” for a number of things. Including getting on the Top 100 Billboard charts. (Which is a separate discussion in and of itself that I don’t want to get into, because I believe taste in music is subjective and ever-changing.)

Yes, but I’ve already seen a gen-AI take a 2D picture and turn it into a 3D model. And even though I’m learning how to create 3D models, you bet your hiney that I will take advantage of that to create a starting point for me and then fix the things I want.

I do not pretend to predict where these things will go, but we are coming close to where the uncanny valley is no longer an issue. I’m not talking about perfection. I’m talking where assets can be made with AI and not look like “slop”.

Maybe I’m wrong.

Yeah I’m a big proponent on not using AI to code. I have made MANY posts on that in this forum. I did not really indicate I was mainly talking about assets. But I’m not ruling out things like GitHub Copilot.

Click on the playlist I linked above and listen to the music. two of the songs use synths for the level music, but listen to the original song then tell me what you think.

9

For what it’s worth, I think your sample is of poor quality. It wouldn’t matter if 100% of the reviews were positive about using AI, because the sample is intrinsically biased towards people who have the exact outlook as the reviewers you already shared here. Something along the lines of “I am disappointed but the result was not bad”, or better.

Your sample is biased because people like me, the creatives and artists, will actively go out of our way to avoid titles like yours. The game could be free to play forever, and it wouldn’t matter. I’m not engaging with it. Thus, you don’t see reviews from people like me, and thus, very few or even no negative reviews due to the use of AI.

This post was originally a lot longer trying to get into the details of why someone like me might avoid it so unilaterally, but I trimmed it for brevity since I’m sure you’ve heard the slew of arguments by now. The fact of the matter is, you can’t use reviews alone as a frame of reference, no matter how good they might look. You can’t even use people coming to you personally and praising your work, and it’s all for that one reason of avoidance.

4 Likes
10

I don’t think you understand what a game jam is. A game jam is a time-boxed amount of time to create a full game. The only people who can rate and review your game are other people who also made a game in that time. I went out of my way to review games in the jam to get people to engage with my game. The most reviews I believe I’ve ever gotten in a jam before is 12. Everyone who beat me in the jam had less reviews, except maybe one game. Because it’s based on percentages.

So I’m being reviewed by people who literally told me they were angry about AI usage. I can’t see what they rated my game, but just because they thought the music was better than they expected doesn’t mean they rated me well on it.

I’m also going up against other people who are trying to compete with me specifically. Different people are going to have different personal rating systems.

There are a lot of people in the game development community who are very vocally against it. I was very adamantly against it myself until I was in a game jam and really enjoyed someone’s music and that’s how I found the music and it made me reconsider.

Yes, my sample size was limited, but I do not think it was as biased as you seem to believe.

First, to be clear, I typically use music that I’ve paid for in games - even game jam games. I don’t even use free music. But I believe you are making a number of assumptions here.

  1. That people are going to communicate that they are using AI in their games.
  2. That people are going to be able to tell if game publishers/developers do not disclose the use of AI.
  3. That creatives and artists avoiding paid games that use AI is going to make a significant dent in sales.
  4. That platforms are going to continue to require AI use disclosure.

Yes, and if you’d followed the links to the further discussions, you would have come across the further things I had to say on the subject, both in response, as well as links to the DevLogs I did about the subject.

Like I tried to communicate, my one example was a small sample size and anecdotal. The real story was the story I linked in the header of the first post. Which is that all the games currently getting trashed for using AI in the press are also still selling like hotcakes.

I appreciate your opinion, and your passion. And my larger question is, does it matter in the grand scheme what we think because games are making money despite being panned.

1 Like
11

Honestly, the better question is if it matters to YOU.

To me, AI use matters quite a lot. I know the tech is magical. You just press a button, review the result, then BOOM, done. Easy!

But personally, I’ve dreamed of working with others to make Monkanics a reality for the past 6 years. Those different viewpoints and skills matter a LOT to the final game and my personal relationships.

Video games are an artform that’s the sum of other artforms. Trying to optimize the art out of gaming defeats the point.

I understand the use of AI for more technical and mundane things, but the art and design must be handled by a human who cares.

While you are correct in saying nobody cared, I still think it’s one of the lamest things you can do as a game developer.

3 Likes
12

Fine. Just don’t at the same time complain about ram prices.

3 Likes
13

No, that’s a very different question IMO. My point is the juggernaut is going to carry on.

Ah, but I think of coding as an artform. However vibe coding makes “good enough software”, and people who don’t know better do not care. If they can put me out of a job they will.

And this is where we get into nuance I think most people gloss over. If an artist can get a better result out of an LLM tool than a non-artist, and it becomes indistinguishable from what could have been made by a human, and the end user cannot tell… does standing up for principles have any effect? Maybe.

Perhaps. But I figure it’s more about what we as humans care about. And ultimately we are not fighting against malice, but apathy.

Fair enough. But in a capitalist market prices are always going to go up.

1 Like
14

I saw this lecture on the history of programming and it seemed that a common pattern is: Progress happens → People who are invested in previous solutions call it a crap solution. For example many insisted on continuing using binary when Assembly first appeared. Assembly was lazy, cheating, stupid etc. etc. Then, some time later, people instead insisted on using Assembly when other languages appeared, as those newer languages were lazy, cheating, stupid etc. etc. I think that it speaks of your open mindedness that you are open for trying to use AI seemingly more out of curiosity than that you actually like or want to use it.

I think people have naive ideas about artists. Most game dev artists are not some Rilkes running around being too pure for this world. They are just regular workers. There are many exceptions ofcourse but if i were to complain at companies for mistreating gamedev artists i would start by looking at the industry itself. Churn rate through the roof. Comparatively poorly paid - it is one of the worst paid fields to be an artist in i believe. Burnout common. Limited agency. But instead artists complain at some hobbyist using an AI tool to not have to do everything from scratch.

1 Like
15

LLM chatbots are a solution for what problem exactly?

3 Likes
16

I’ve dabbled a few times trying to get good graphics (3D). Finally gave up and I just go to artist sites to buy these as they are much better, already rigged, optimized and ready to go. No messing around with them for substantial amounts of time trying to fix the AI delusions, optimize and other errors, then having to go rig, etc.

In code writing I’ve tried it a few times. For something VERY SIMPLE it usually gives an acceptable answer. Anything slightly more complex or out of the norm and it fails miserably. Code doesn’t work, requires more time to fix than it would have taken for me to just research examples and write my own, breaks previous code and the funniest one of all… It said to add “this” to the beginning of the code, then add “that” procedure and then finally in your main look you should (literally this is what it said) “oops something went wrong”.

Music is one area I have not tried but after failure after failure, and knowing that I’m trying to create games and hoping one day to offer for sale… that I just don’t feel right screwing over artists doing the same thing in favor of AI instead.

4 Likes
17

I never said they are the solution to any specific problem. They are tools you can choose or not choose to use. Just as Assembly when it first appeared.

For someone who needs some assistance with some of the many things needed for a game, they can be great. One example found right here is by making some music for a small indie game.

I mean, AI has in many tests and studies done things such as outperforming human doctors at diagnosing illnesses. No doctor seems to have cried out in pain because of that. The reactions seems to have been more along the line “How can we best use this tool?”. But suddenly when it comes to gamedev AI are all 100% garbage at everything?

I even saw a guy a while back trying to fix a bug for more than a week then finally have AI solve it for him in five seconds, still claiming that AI suck.

It feels almost religious.

2 Likes
18

You were sneakily implying they are the “next step” in dev tool evolution: binary > assembly > low level languages > high level languages > (LLMs). This is fallacious. LLM is a statistical language model, not a language.

Yeah, it’s mostly garbage. Ironically, it’s always the people who have very little actual production experience proclaiming it’s useful. Someone who is inexperienced in any field the “ai” is currently being peddled for, simply doesn’t have a broad enough horizon to gauge its real usefulness. If you have trouble implementing conway’s game of life and some “ai” bot farts it out in a second, you’ll surely be amazed and think is enormously useful. That’s far from objective valorization though.

But its uselessness is not the main point of contempt. Unlike medical applications of machine learning, all typical ML applications that may be used for games (code, visuals, music) are based on training datasets that contain enormous amounts of stolen copyrighted data. Generative “ai” breaks laws. It’s a criminal activity on a mass scale.

The funny thing is - even with all that stolen data - the usefulness is marginal… and there is very little fresh, human-created data left to steal. As proompters continue to slopify the internet at ever growing rates, and human creators get de-incentivized to supply fresh data, generative models have nowhere to go but in-breed. The only possible trajectory is a slow grind to a halt… unless financial bubble pops first and flushes the whole thing quickly down the tubes.

3 Likes
19

Wouldn’t your review be biased anyway?
You must admit that for you there will never be a case where AI game/asset-set would review well. You could never provide an honest review.

Right now AI is sloppy. But if (hypothetically) AI could produce a product that is better than what could be produced by humans would you insist on human content?
Not hard to ignore quality and go with the human produced content when it comes to game assets but then…
If a robotic surgeon using AI was safer and more successful at that craft than a human which would you prefer for your ailing family member?
Both questions are hypothetical only and are somewhat rhetorical and meant for a little self discovery as the underlying ethical considerations are the same in each case.

3 Likes
20

Well it may have been unclear if that is how you interpreted me. I only meant to say that they are new, not that they are the next step in language evolution. But i guess it could be, as it translates human language into code. Another example could be early resistance to modern user interfaces. But that may also sneakily imply something. What do i know?

I agree that the AI business is hyped up and that the training data sourcing is unethical but i disagree with that it must necessarily be entirely useless because of that. Also, i think you are wrong about it only being so for art. Absurd amounts of not publicly available medical studies etc have been used as training data. I believe there has been several lawsuits about that. But for some reason the medical industry as a whole seem more intent on trying to find a suitable niche for gen-AI than nitpicking and listing flaws.

1 Like