Hacker News Reader

Comment by CuriouslyC | Hacker News Reader

I don't think AI will ever really solve "making a fun game" because:

1. Games are highly complex systems, so for example adding a fun mechanic from one game to another game can actually make that game less fun, and AI has no way to know, you just have to playtest it.

2. Games are evergreen (people still play and love the original Zelda/Mario/Tetris) and social (you want to share your love of game XYZ with others), so there's literally no reason for anyone to play a purely derivative game, when you can just play the original.

Eventually AI will be able to make perfect assets and all the other stuff AROUND making a fun game, but the game design will always be a human thing. At best, AI will come up with a bunch of game design experiments that a human will need to playtest and provide feedback on.

View on HN

You may be right, at the same time a lot of "AI will never solve X" takes have aged pretty poorly, and I'm not confident what's fun about existing games cannot be codified and shown statistically somehow. Some genres seem more accessible to AI than others

It's hard to say for sure, I think we can at least confidently assert that games are going to be more challenging than other forms of creative expression like Books, Music, Art, etc because people's tolerance for derivative games is lower than their tolerance for derivative movies/music.

Agreed!

And I genuinely hope there will be some categories of games/music/books/art where human input is still needed and preferred. Currently I'm already a handful of bulletpoints down on the list of stuff I thought AI would never be able to do, and it's been kind of destabilizing

Certainly yours and mine, given we're both (presumably) adults with experience. I'm not sure if the next generation of kids is going to care much given the popularity of platforms like Roblox.

It's not that kids actually love the lazy derivative games on Roblox. What they love is the rush they get from feeding the gambling addiction the Roblox Child Casino was carefully designed to develop in them. Cheap games on Google Playground won't be nearly as attractive to kids unless Google also fills it with anti-features like push notifications to nag children into getting another fix when they start having fun outside of Roblox, or adding a ton of limited-time rewards to fill kids with anxiety and make them feel like shit when school, homework, or family gets in the way of earning more Robux.

What Google Playground really needs is their own currency, designed to hide the true cost of every purchase, that children can earn by making popular games. Then Google can profit from child labor the same way Roblox does. https://www.classaction.org/news/roblox-lawsuit-alleges-enti...

i think the difference here is that humans can't even write down what makes a great game great. we don't know. you have to play it.

if we can't even write it down, how is an llm going to be able to ace it?

I feel like this is true with most media such as music and movies too - generally consumers don't understand exactly what they like.

I imagine AI will be making semi-complex indie-style games before we see an AI movie worth watching

I don't think anyone, given better alternatives, would choose games like these. There's a reason low effort output of LLMs has been dubbed Slop... Slop isn't fun or interesting to consume.

> a lot of "AI will never solve X" takes have aged pretty poorly

"Make a fun game" or "create good art" is the cognitive equivalent of trying to solve the halting problem. These are some undecidable algorithms that cannot be solved by just throwing more GPUs at it.

Now, if you were to ask an AI to make an "addictive game", I'm sure it will eventually be able to.

That’s fine though?

I’m having a lot of fun creating an “improved” clone of a very old school game I loved.

It takes lots of back and forth but there is definitely (for me at least) to play this derivative game since it adds a lot of levels and stuff that weren’t available in the original.

Without AI I wouldn’t be able to do anything; but with it it can make some changes, I test them and we iterate on it.

So for me AI solves “making a fun game”. It doesn’t “one shot GTA7” if that is your goal but that also sounds boring

this is probably a personal problem, but sometimes I get turned around on 'fun making a game' vs 'making a fun game' and I have the idea that ai elides this distinction even further

You misunderstand me, AI solves the "making" part, but not the "fun" part, which is 100% you.

Yea not the "making" part even (at least not yet). Every single demo/sample of a vibecoded* game thing is a piss poor "counter strike clone" or "gta clone" type of app that has bare bones everything (and i mean PS2 level or worse) and plays like absolute ass down to the basics of character control ans basic gameplay mechanics.

*: I don't consider a project you built with an LLM to be "vibecoded" if all the important ideas are from your head, you know what the code is supposed to be doing (domain model is in your head) and you guided the project every step of the way. Otherwise we're all vibecode. So yea, I don't doubt you can create amazing games with AI, but "with" as in "together", not as in "Claude, build me a Minecraft clone, make it fun, no mistakes."

seriously, games require balance and playtesting, and synthetic users don't count.

Reply on news.ycombinator.com