I know progress happens in incremental steps, but this seems like quite the baby step from other world gen demos unless I’m missing something.

An end-to-end _trained_ model that spits out a textured mesh of the same result would have been an innovation. The fact that they didn't do that suggests they're missing something fundamental for world model training.

The best thing I can say is that maybe they can use this to bootstrap a dataset for a future model.

[0] https://worldgen.github.io/index.html

(couldn't cleanup the link at all sorry)

[0]: https://scontent-lhr6-2.xx.fbcdn.net/v/t39.2365-6/586830145_...

I still won't even get myself a PlayStation, explicitly because I know if I did I would lose half year of my life to Red Dead. Who actually benefits from this technology, or is it just a cool demo?

DotA was effectively a simple map that changed online gaming, e-sports, and I am sure there are millions/billions of hours spent by players in a very simple looking landscape.

Compared to what we have today, on-demand, unique, and significantly better looking. It is amazing to see how relatively small these, objectively amazing, achievements seem, compared to a simple map we had 20 years ago.

10 years from now we might have games that generate entire worlds based on the unique story line that's customized for each playthrough. Maybe even endless stories.

Baldur's Gate 5 is going to be memorable!

The Elder's Scrolls could use this + Radiant AI for some neat quests when it improves.

Game studios are probably going to explore this in dungeon generators first where if things go wrong with the generation, not much is lost. Just exit and generate another.

But, having things feel strongly on a grid kind of ruins the feel. It's rare for every building to be isolated like that. I am guessing they had trouble producing neighboring buildings that looked like they could logically share a common wall or alleyway.

Any world you can summon into existence with a few words is by the laws of information theory going to be generic. An interesting world requires thousands of words to describe.

I can see it being useful for isolated Unity developers with a concept and limited art ability. Currently they would be likely limited to pixel games.

But, it looks like WorldGen has that slightly soulless art style they used for that Meta Zuckverse VR thing they tried for a while.

https://www.worldlabs.ai

You can explore, but is there a single interesting thing to find?

https://www.challies.com/articles/no-mans-sky-and-10000-bowl...

This is true for high-fidelity environments that people expect from AAA games or movie virtual sets, but it's not really true for the sort of content that Worldgen is producing. The effort required to learn low-poly 3D asset creation in Blender is definitely significant but it isn't out of many people's reach unless you have an especially low opinion of people. The Blender community makes assets like this stuff easily all the time.

One can imagine different ways to integrate this type of decomposed generation with different game engines or to parallelize it or allow lazy generation of assets. It's also very accessible to programmers like me who don't have the resources to train and host giant world models but are interested in AI world generation.

I assume that something like this is going to end up in Unity, Unreal and others within a matter of months.

And people are going to say that we already have enough crappy Unity asset games in the monopoly Steam store, but I think that misses the point. It's about opening up game creation or world generation as a creative outlet or tool for more people. It's not an attempt to create more refined games.

And let's not talk about the cultural flattening that this represents. A "medieval village" from where? When? Whom?

This is just slightly refined AI slop, but slop nevertheless.

I tried this VR headset from Meta the other day. It is so designed to throwing young people into digital realms by shutting off every single biological sense they have.