Building an AI game studio: what we've learned so far

I'm not even certain that product can exist, at least not anytime soon. But AI is in a hype cycle right now so everyone wants to jump on the gravy train.

Of course a lot of people don't want it. A lot of artists, developers and creative people want nothing to do with it, for obvious reasons. I don't want it. I think AI-driven tools only serve to commoditize and diminish human creativity. I've never seen anything generated by AI that I find interesting beyond the sense of superficial aesthetic or vibe. But I'd never prefer it over something made by a human being.

But no one cares what the muggles think, and criticizing AI right now gets you denounced as a crank in much the same way as being an anti-vaxxer or flat-earther. There's nothing to be done but hope the fever breaks before too much damage is done to the world.

So... I think maybe the same could have been said about writing once upon a time. "What if everyone could write, we'd flood the market with poorly written books!". Well we went from no one being able to write to almost everyone being able to. But just because everyone can technically write, it doesn't mean that everyone is a _writer_.

I believe that the same goes for AI tools for games. Yes, more people will technically be able to produce games. But I think that will shift the focus; it won't be enough with an impressive tech demo in the future, instead you have to connect to the human side of people when you build games.

We now see it as a natural thing that everyone can read and write. We don't want to go back to a time where it was only for a select few elites. If we turn things around and imagine that the technology exists that makes it possible for anyone to build games; then should we keep that from them? In that world, why would we want to gate keep game creation to only people that have time and money to go to a game school or equivalent? You might think "well everyone can learn on their free time", but that's not necessarily true.

I think we will see more human, and more personal experiences that touch us deeper, because they no longer can just be about the technology (since the technology will be commoditized). That's what I'm personally excited about and why I think it makes sense to work on this.

I do agree thought that it can feel overwhelming to look at all this in aggregate. There are already hundreds of thousands of games, why do we need more? But maybe looking at things in aggregate is not the right way to look at it. There have been countless conversations between people throughout history. Does that mean that a conversation is meaningless? I don't think so, and I think (some types of) games will move into this space too; something more personal, something we don't count in aggregate, something that is between maybe smaller groups of people, but more meaningful to those groups of people. At least that's something I'd be excited to see.

Thanks for giving these interesting examples.

I hope that this doesn't come across as moving the goalposts, but for me it's a given that popular Youtubers can multiply the success of a game, and that this depends to a large extent on luck. For me, the crux of the matter in discussions about game development profitability and survivorship bias is not how much luck influences the maximum possible success of a game.

For me the relevant question is: If you develop a carefully crafted, fun game based on a game concept you have reason to believe will have a decently sized target audience, will you make enough money for a living with some kind of predictability, or does even this depend on luck?

Now when looking at The Void Rains Upon Her Heart, I'd like to know the sales figures before the jump in popularity. You can filter Steam reviews to the period before the game's popularity surged in June 2023. At that time, there were 561 reviews (or 338 reviews for copies sold via Steam). Using the common sales estimation trick of multiplying the number of reviews with 50, we get approximately 28,000 (or 17,000) owners before June 2023. I think the price fluctuated between $9.99 and $6.99. Assuming that after Steam's 30% cut and taxes there are about $3 left on average for each sold copy, then 3 * 25,000 = $75,000. According to the game's credits on MobyGames it was made by a single person. Depending on how long he took, maybe it is reasonable to assume that the game was en route to profitability anyway?

Nah, the standard will just change.

A "AAA game" will be the ones that have that extra it whether that's something resulting from human curation (the person firing off prompts who knows how to tune them and knows what other people want) or style (the person assembling things to be more cohesive) or editing (the person making sure the AI puts out an absolutely amazing story instead of a meh one) or whatever.

Basically, look at what a AAA game was 10 years ago, 20 years ago, 30 years ago compared to today, and then extrapolate the other direction. ;) The person who wanted games with the quality of 30 years ago has countless options, yet people still pay for the ones they think are the creme of the crop.

Emphasis on "if you could simply type in English" (or insert native language here).

Sometimes the AI misinterprets what you want. Sometimes this is due to lack of sophistication on the model's part, but I think the bigger issue at hand are the shortcomings of natural language. Even when working with an intelligent human being, sometimes you have to communicate your thoughts/ideas through pictographic means.

It's something I've thought a lot about lately. No-code/low-code is about as old as GUIs themselves, yet it's never become the de-facto way of creating software. However, for everything else we do we use a GUI. No one creates illustrations on the computer programmatically (unless you're into generative art like me). You probably aren't doing your taxes in a terminal. You get the idea.

This is a segue into my other critique of prompting as a UI. A lot of UX people think it is the ultimate UI, and natural language the premier way of communicating jobs to be done. I disagree. I think symbols and diagrams communicate more with less. Concerning language, I think that for whatever reason programming languages are the most natural way (that we've currently found) to communicate intent in a programmatic context, markup languages are the most natural way to communicate intent in a layout context, and that's the reason why despite 50+ years of GUI innovation programmers still use text-editors. The old guard has all aged out of the profession, the new guard still uses text-editors.

So that all being said, I think the most incredible AI demo I've seen so far is the tldraw stuff done by Steve Ruiz, namely Make Real, and that's because you're communicating with the AI in a pictographic / symbolic context, and because it's a whiteboard, there's no limit to the symbols at your disposal to communicate your intent. The limit is AI's ability to interpret what those symbols mean.