Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

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

braindump.me

Post reply on HN
Print view View original

291–300 of 312 posts

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

#291

Post by mathteddybear »

johnnyanmac wrote: post #279
somenameforme wrote: post #274

Earlier quoted context omitted.

No I don't, and this is one of the reasons why. When you can work online and get paid in $, that dollar goes so much further in the other 95% of the world. What people who don't travel much may not understand is that it's not just the exchange rate. What really matters is PPP - purchasing power parity. $100 might translate to e.g. 8000 rupees, but those 8000 rupees go far further in India than $100 would in America.…

>So by living outside of America and getting paid in USD you basically just multiply every dollar you make, by a very large amount. That $168k easily becomes a $million+ in the overwhelming majority of the world That's a nice sentiment for those who live in the not-US. But sadly most game development is in fact located in the US and Japan. For the US, many of the developer scene is in fact in higher cost of living ar…

As someone who moves out of Bay Area back to home country and looks forward to spice up early retirement with small-time gamedev, I find the economics of indie development not that simple.

Generative AI notwithstanding, in order to make a game you will hire some artists, and the best artists are "getting paid in USD" already, to use the words from this discussion thread. They may be talent living locally, but they work on-line, too.

Thus, successful indie games usually have budgets "in USD" and sales "in USD". I mean, making one is still a first world problem.

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

#292

Post by wasteduniverse »

Jasper_ wrote: post #204

Spot the fun logic bugs! Here's a great example of where AI requires some careful attention to detail: const tentsFacingNorth = getAllUnitIds() .filter(unitId => getUnitString(unitId, "unitStrings/tags")?.includes("tent")) .filter(unitId => Math.abs(getUnitRotation(unitId) % (2 * Math.PI)) This code only works if getUnitRotation() returns values between -pi and pi, with north pointing up. This isn't an unheard of con…

[dead]

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

#293

Post by kkukshtel »

I'm very into the idea of using AI to change the types of games we make, but I think that "prompting the game to do stuff" is barking up the wrong tree. As the top comment says here, you can develop non-intelligent solutions to do what is going on here in probably the same if not less time. This is more like "end of chain" AI work and doesn't center AI as necessary for the game itself. If anything this is using LLMs as a shortcut for keyboard shortcuts or something, which is fine, and likely necessary work to get to something new and interesting, but already makes the category error at the top of constraining what is possible based on the need to fit the LLM into the rules box.

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

#294

Post by kjkjadksj »

johnnyanmac wrote: post #232
Bjorkbat wrote: post #156

I think a part of the reason why I'm kind of dismissive of AI when it comes to creative endeavors is that I haven't seen it do something that isn't already possible using a "not-intelligent" solution. For example, AI website builders. It's impressive as a tech demo that you can prompt an AI to make a website for you, but otherwise I don't think this is going to be disruptive to the web design industry since it's alre…

>AI doesn't eliminate the complexity, it just hides it. At best it'll get you maybe 80% of where you want to go and you can get to the remaining 20% by hand. At worst, the complexity is so thoroughly hidden that addressing it isn't an option. It simply does not allow you to manually edit the outputs, or makes it unreasonably difficult to do so. That's the thing about games. It's "easy" to make a website, it's very ha…

You are comparing an apple to an apple pie. The “easy” website is going to be something that looks like static html. The “easy” game will look similarly low tech: maybe a platformer that looks like it was written for an atari. 3d modern games are not the only way to make games although thats been a popular way for some time.

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

#295

Post by kjkjadksj »

jarjoura wrote: post #95
dkarras wrote: post #26

Earlier quoted context omitted.

It is not just about marketing though. There is this belief, especially among fledgling indies that if they make a game that they think is good, and if they are able to market it well, it will be a success, even a hit. A hit in gamedev is not easier than creating a hit in the music industry really. Even the veteran composers are not able to replicate (or even achieve) creating a hit. Possibility of success depends on…

Hate to break it to you, but the pop star success formula is actually meticulously detailed and very repeatable. The producers and song writers know exactly how to do it. It’s not about challenges in finding success, so much as, at least in music, artists don’t want to follow a formula. They want to create songs and melodies that are personal and meaningful to them and their fans.

Artists absolutely want to follow a formula. Read a little bit about songwriting. How common chord structures are across popular music. Music theory in general is literally formulaic. Beethoven or Radiohead both follow formula.

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

#296

Post by ghostzilla »

hluska wrote: post #263
ToucanLoucan wrote: post #227

Earlier quoted context omitted.

> So... I think maybe the same could have been said about writing once upon a time. I mean, you say that as though it is not increasingly year over year more and more difficult on balance to make your living as a writer. That it hasn't been a famously difficult thing to do since like... I mean good god, I remember writers complaining about this when I was a kid on forums back in the early 2000's and at that time it w…

You come across as being extremely condescending. And I’m sure you make some good points, but I can’t find them behind the tone. It’s a shame because again, I’m sure you make good points.

On the internet, no one hears you being subtle. (Torvalds)

I'll add my own view: when you watch a movie, read a book, listen to a song, play a game... you CONNECT with the mind of the person who made it. When there is no mind, or the source is a dead, statistical amalgamation of countless fragments of other minds, there is nothing you'll want to connect to, nothing you'll want to squander precious hours of your life on.

And while you may be curious to see, once maybe, a movie such an imaginary AGI-LLM has created from your prompt, no one else will have the slightest interest in seeing it. And vice versa. Which means there would be absolutely NO MONEY in that market. There would be no market.

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

#297

Post by johnnyanmac »

kjkjadksj wrote: post #294
johnnyanmac wrote: post #232

Earlier quoted context omitted.

>AI doesn't eliminate the complexity, it just hides it. At best it'll get you maybe 80% of where you want to go and you can get to the remaining 20% by hand. At worst, the complexity is so thoroughly hidden that addressing it isn't an option. It simply does not allow you to manually edit the outputs, or makes it unreasonably difficult to do so. That's the thing about games. It's "easy" to make a website, it's very ha…

You are comparing an apple to an apple pie. The “easy” website is going to be something that looks like static html. The “easy” game will look similarly low tech: maybe a platformer that looks like it was written for an atari. 3d modern games are not the only way to make games although thats been a popular way for some time.

I can clarify. an "easy" website can still do its job and get you traction, because the website design and architecture is not going to affect your product as long as you follow the simple rules. The website is not your primary product, it is an ad to your actual product.

There are a few outliers like Vampire Survivors, but for the most part a simple game will not penetrate this market, and it is the product you're selling. Unless you're going to try to be like that dev that makes a simple game every month, for years on end, you're probably won't succeed making "simple" games alone (even then, you can argue that the primary audience is not a traditional gamer by that point).

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

#298

Post by Hammershaft »

kibbi wrote: post #283
Hammershaft wrote: post #271

Earlier quoted context omitted.

Sure! Pinning down some shared criteria for quality and obscurity might be a challenge, but I can give it a try. I'll only include games that I've played recently with an overall steam review score of >=%94 and less than 1000 cumulative steam reviews. Deadeye Deepfake Simulacrum: https://store.steampowered.com/app/1545990/Deadeye_Deepfake_... Tandis: https://store.steampowered.com/app/1297210/Tandis/ Nidus: https://s…

Surely you're not claiming that a game with fewer than 1000 reviews is "languishing"? Any indie game with 1000 reviews is very successful. Picking your first example, Deadeye Deepfake Simulacrum has about 460 reviews, roughly 20,000 copies sold, an estimated revenue of about $120,000 while it's still in early access, and apparently a developer team size of 1. I'd guess that this game will net the developer several hu…

Deadeye deepfake Simulacrum has been in development for ~4 years and is still getting updated, so even at this estimate we're looking at a take home pay of 30k - 40k per year before taxes.

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

#299

Post by Hammershaft »

somenameforme wrote: post #281
Hammershaft wrote: post #271

Earlier quoted context omitted.

Sure! Pinning down some shared criteria for quality and obscurity might be a challenge, but I can give it a try. I'll only include games that I've played recently with an overall steam review score of >=%94 and less than 1000 cumulative steam reviews. Deadeye Deepfake Simulacrum: https://store.steampowered.com/app/1545990/Deadeye_Deepfake_... Tandis: https://store.steampowered.com/app/1297210/Tandis/ Nidus: https://s…

I'd drop the max reviews down by quite a lot. Sales:reviews tend to be around 60:1. So a game with 1000 reviews has sold tens of thousands of copies. We can look at something like the PS2 for some reality checks. It sold 1.5 billion games through 4000 titles. So an average of 350k sales. And of course that's going to follow something like a typical 80/20 ratio, which would leave us with a median sales in the ballpark…

The only source I can find for a 1:60 review to sale ratio is this:

https://gameworldobserver.com/2022/11/15/how-to-count-game-s...

This article explicitly mentions that that ratio only holds for hit games, and mentions that the ratio for smaller games is roughly 1:20, which is still a high ratio compared to figures I've heard in post mortems.

Assuming a game with 1,000 reviews that takes ~3 years for 1 developer to make and support that sells for $15 at a 1:20x ratio. We're looking at 210,000 after steam's cut and before taxes, this developer at top of our review range is making ~70k a year before taxes that are calculated on one year of income. Many small indie developers lose ~35% of their income to taxes alone on game sales and so even under nearly ideal conditions for this 1000 review max constraint we have a developer earning Note that many of the games I listed are highly rated but have less than 200 reviews

EDIT: I often reccomend Tandis to people, it's a great puzzle game that builds geometric intuition, and it's designed to distill that intuition into a roughly 5 hour play time.

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

#300

Post by hatenberg »

stale2002 wrote: post #103
hatenberg wrote: post #60

So ... does every single AI demo put copyright and intellectual property infringement on the box? Just asking? How on earth can an entire ecosystem be so damn tone-deaf and lack self-preservation?

> tone-deaf Maybe. I'm not sure why that matters though. Better to be hated than not even noticed. > and lack self-preservation Arguments like this confuse me. Copyright infringement is everywhere. Not just among companies, but also among individuals. Any time someone draws fan art of the Mario and posts it on this internet, this is illegal copyright infringement. But basically nobody ever gets sued for that kind of…

They are baiting the mouse with the star wars items
Post reply on HN