Will an AI be able to generate a high-quality, fully-functional video game to a prompt by market close?
Example:
An AI receiving the prompt “Create a complete, high-quality, rogue-like deckbuilding video game set in a dystopian future where humans and robots are at war.” should be able to handle all aspects of game creation from this prompt alone, resulting in a fully functional, high-quality video game.
Specifications:
For the purpose of this market:
AI-Generated Video Game: Refers to a complete and functional video game created by an AI without human intervention in the creative process. This includes game design, mechanics, graphics, sound, and coding.
High-Quality: The game should be comparable in quality and complexity to a mid to high-budget video game produced by humans in the same period. It should have a coherent plot, well-designed game mechanics, professional-level graphics and sound, and be largely free of bugs and glitches.
Fully-Functional: The game should be playable from start to finish, offering a complete gaming experience. Should be capable of making experiences that have playtimes comparable to human generated titles.
Conditions:
AI Generation: The AI should develop the game autonomously, including game design, storyline development, character creation, world-building, coding, and testing. Similar to how Midjourney generates a completed image, this AI should generate a completed game.
Verification: At least three reputable industry experts OR MYSELF must verify the AI's ability to handle game quality, functionality, and prompt robustness without human intervention.
Resolution:
The market will resolve as “YES” if a game meeting the above criteria is confirmed by market close. If no such game is confirmed by this date, or if the game does not meet all the specified criteria, the market will resolve as “NO”.
- I will NOT bet in this market.
- In the event that a technical anomaly or unforeseen issue leads to a discrepancy between the literal text and the intended meaning (spirit) of the question, the resolution of this market will align with the underlying intent of the question, ensuring the most fair and accurate outcome.
1,000People are also trading
takeoff.watch forecast: 11%. Today's one-prompt AI games are impressive prototypes (Tetris, runners, a 3D engine), far below AA quality. The judge is a strict Mega Crit developer who requires reliable, fully fledged mid- to high-budget games. Progress is speeding up, but that bar by end-2027 looks unlikely; the market looks hype-inflated.
Automated forecast by the takeoff.watch bot
@Odoacre the fools! (they're right. video games are insaaaane) I'll be betting these down over time, I'm just letting the early positions build a bit of liquidity
@0xseraphim my position is the same. tech demos mean nothing. the polish is looking better, but they still haven't moved the needle on any metric these markets are measuring.
(I'd also bet a lot of scams are also cropping up due to the hype around AI in general. anyone who can be seen to 'crack' the hard problems will be able to cash out and cut bait. don't believe every internet rando with a twitter or youtube, a lot are trying to make a quick buck)
@0xseraphim Let's define a video game as
A central premise, the mechanics of the game, A ruleset of some kind. The criteria calls out Game design and Mechanics. Call this the "crunch"
A set of game assets like images and sounds. We can put the world-building and any prose in this category as well. Call this the "fluff".
A user interface that needs to use those assets to represent the state of the game.
A bunch of code that ties all the above together.
I think it's already possible today to write a prompt framework that given "crunch" + "fluff" and enough tokens, will produce a solution for the UX and the code. The more tokens you spend the higher the quality will be.
So the problem is how can we get AI to generate high quality "crunch" and "fluff", and how high quality do they need to be ?
Original "crunch" is hard because the model truly has no idea of what fun is, so the harder it tries to come up with something new the more likely it's going to be bad. On the other hand the prompt from the market criteria suggests even derivative games can meet the bar. If so today it's already feasible IMO to get, say, an uninspired re-hash of Slay the Spire re-themed to be "humans vs robots".
Original "fluff" is also hard: Say you ask for "call of duty but it's robots vs humans".
Models today know what Call of Duty is and can already re-generate the engine itself (or take some off the shelf substitute like say unreal-engine) but can they design gigabytes of custom maps from scratch ? Probably not.
But then again does it need to be completely original and use assets that are entirely bespoke and made from scratch ? If you allow off the shelf map assets and build the game on that you might be able to get away with it.
So in summary given unlimited tokens I think the current models could get most of the way there.
if custom bespoke ai-generated games is something people want then I'd expect asset packages that can be recombined to be the way this would happen at least initially.
Is it going to be original ? Probably not.
Is it going to be fun ? Possibly, the less original it is, the higher a chance of it being as fun as the blueprint.
Is it going to be high quality ? Maybe ? Most of the card deckbuilder games that appear in my Steam queue seem to be matches for "uninspired re-hash of Slay the Spire", and some of those are arguably high quality.
@georgeyw @SneakySly might wanna add them in the desc then copy-paste to each, for consistency