MarioGPT Uses AI to Generate Endless Super Mario Levels for Free

The secret to Mario games is a new gimmick introduced every level. They&#x27;ve gotten better at this over the years, and Mario 1 + Lost Levels is the worst example of it. That makes it a great comp for GPT-3 which can churn out an endless supply of flavorless brick &amp; pipe levels and still feel vaguely Mario 1. Were this tool to live up to the hype of &quot;indies punching above their weight&quot;, it would need to design novel platformer mechanics. The work it&#x27;s doing isn&#x27;t the hard part of platformer level design.<p>For another example of why this isn&#x27;t commercially viable, look at what happened with Super Mario Maker. In that game _humans_ are given a fixed set of Mario doodads with which to build levels. But Nintendo kept the secret sauce for themselves - the ability to create new doodads. What follows is millions of derivative Mario levels unworthy of their own game. Even if you trained MarioGPT on the rich set of level data available in Mario Maker, you would not have an algorithm that makes commercially viable Mario levels.

This is neat and all, but procedural 2D level generation can be done really well just with simple heuristics, see Spelunky from 2008. And that can be built into the game and computed efficiently on the fly, not requiring an internet connection.

I&#x27;m surprised there haven&#x27;t been more games like Cloudberry Kingdom. It has fantastic level generation with a bunch of settings to play with. Players can even have various movement abilities which the level generation considers. If you got stuck it provided an AI to follow.<p><a href="https:&#x2F;&#x2F;store.steampowered.com&#x2F;app&#x2F;210870&#x2F;Cloudberry_Kingdom&#x2F;" rel="nofollow">https:&#x2F;&#x2F;store.steampowered.com&#x2F;app&#x2F;210870&#x2F;Cloudberry_Kingdom...</a>

Are the levels playable, or just static lookalikes without moving parts?<p>Incidentally, there&#x27;s a nice example of a text representation of a level in the source code (requires scrolling horizontally, which isn&#x27;t totally obvious from the GitHub UI): <a href="https:&#x2F;&#x2F;github.com&#x2F;shyamsn97&#x2F;mario-gpt&#x2F;blob&#x2F;main&#x2F;mario_gpt&#x2F;level.py">https:&#x2F;&#x2F;github.com&#x2F;shyamsn97&#x2F;mario-gpt&#x2F;blob&#x2F;main&#x2F;mario_gpt&#x2F;l...</a><p>Some parts are recognisable, for example the flag pole (which is typically at the end of mario levels, I believe).

Port Mario to WebGL, integrate this model, tell you friend he gets 100 bucks when he beats the level ;)<p>On another thought: this could probably replace the chrome dino pretty well

It&#x27;s interesting because this is using GPT-2 (<a href="https:&#x2F;&#x2F;huggingface.co&#x2F;distilgpt2" rel="nofollow">https:&#x2F;&#x2F;huggingface.co&#x2F;distilgpt2</a> specifically) which you can just fine on a reasonable GPU.<p>But I&#x27;m not convinced the results are any smarter than a randomized procedural generation (I&#x27;m sure using it for text generation instead will yield sub-par results).

I wonder if ML will find its way to maths as a assistive intuition.

<a href="https:&#x2F;&#x2F;arxiv.org&#x2F;pdf&#x2F;2302.05981.pdf" rel="nofollow">https:&#x2F;&#x2F;arxiv.org&#x2F;pdf&#x2F;2302.05981.pdf</a>

Can&#x27;t wait to see someone hook this into an AI that plays the levels and put it on Twitch

Retux and Wario Land like games make better levels for sure because of the puzzles.