Yeah this really needs to rigidify it's resolution criteria. Do the games need to be high quality and well received? Do amateur games count? Do games from any platform count? Are we comparing the results to Triple-I games, high polish, jank cult games, multi-year micro indies, or just anything anybody ever released? When you say any genre, does that mean it needs to be capable of every genre or does "baby's first game" clones of pong and snake qualify? The genre has to exist in 2023, but does that mean we're comparing to the actual games of 2023, or are we comparing to any game made ever?

@Stralor I've put down a small limit order, but I really wanna be able to invest a ton here. I'm extremely bearish on AI ever being meaningfully competent at doing single-prompt fully-fleshed games. but if it devolves down to "anything with a button on a screen is a game" then this kinda breaks down.

I think it's worth discussing now what kind of <=2023 games should act as benchmarks, and on what criteria.

@DanW Since you didn't get any responses to this question:

I'd interpret "comparable in quality to indie games made before the year 2024" as comparable to high quality indie games made before 2024. I.e., if there's a game in a particular genre that was made with a production budget of <$1 million, an AI must be able to produce a game of nearly the same quality.

Examples of indie games from various genres that I'm pretty sure had budgets < $1 million:

A Short Hike, Overcooked 2, Stardew Valley, HROT, FTL, Among Us, Gorogoa, Eastshade, Slipways.

If you used this definition, I think this is about as unlikely as the AA question. But maybe you have a different definition in mind.

https://twitter.com/emollick/status/1894096525602574824

@AdamK https://twitter.com/kimmonismus/status/1895966293721055619

Trying to figure out whether I think this will happen in 26 or 27. With quite liberal judging criteria, I wouldn't bet against this year.

A market targeting AA games:
https://manifold.markets/DanW/by-2029-will-an-ai-be-able-to-gener

By 2029, will an AI be able to generate a full high quality Video Game to a prompt?

19% chance. A single AI system must be able to use an arbitrary prompt to generate video games from any genre that exists today in 2023. The generated games should be comparable in quality to games created by 'AA' mid market studios in (or before) the year 2023. Some examples: "make me a multiplayer arena shooter in set in zero g" "make me a fantasy rpg with branching choices using a combat system inspired by D&D" "make me a voxel based open world exploration game based around crafting" "make me a mindbending puzzle game that uses perspective and time as central mechanics" The AI doesn't have to be available to the public, as long as it's confirmed to exist. EDIT #1: I will not bet on this market.