Yeah so I have a v loose feeling that, in technology, whatever you encounter first or do a lot of is “normal” and then you bring that as a template to whatever you make next.
- Excel then iTunes
- Pokemon then Facebook (gotta catch em all)
- Web indexes then foundation models (foundation models are the vast “aaaall the data, uncategorised” machine learning models underpinning GPT-3 and DALL-E)
- Minecraft then …what?
A whole generation grew up in Minecraft. Peak years: early 2010s. I keep an eye out for when those expectations meet the world, and in what form.
I don’t think the “metaverse” is related – too high fidelity, too corporate. (The lo-fi metaverse would be, if that were bigger…)
Maybe the “cozyweb” trend? Cozyweb features in Venkatesh Rao’s 2019 internet map as collage-y, a cut-and-paste aesthetic; friends hanging out in the corners of the internet out of the full glare of the indexers. I’ve previously named this diaspora into sub-Dunbar spaces as virtual private neighbourhoods and maybe it’s a recapitulation of the DIY, small-group-social experience of a Minecraft server. But I’m not 100% convinced.
OR: how about voxels?
Voxels. 3D pixels.
In particular: 3D pixels that are chunky enough to see, pick up, and make things with.
I don’t really care about voxels as an invisible underlying data structure of 3D virtual environments or scans; I do care about voxels that you can see. Like icons that are meaningful to the human and the computer. Fat voxels like Lego bricks.
Minecraft players will have imprinted on fat voxels. I should, they will say to themselves in their deep unconscious, be able to sculpt the terrain of my computing environment. It should be obvious how to do that merely from the aesthetic.
There’s more and more 3D around.
So I look out for signs of voxels in the 2020s.
I haven’t seen much tbh but occasionally there’s something from an unexpected source.
Charter cities and their foundational law.
Charter cities are semi-autonomous, special administrative zones, carved out from their host country to run their own legal system and economy, often from a libertarian perspective.
CAVEAT: I am intellectually intrigued by the idea! It’s like a philosophical thought experiment from Ancient Greece or science fiction that people live in. But morally it’s icky. You never hear of socialist startup societies. [UPDATE: Of course you do, I was being lazy. I mean in the new charter city movement.] The narrative is always “freedom” aka let people with existing power/money do what they want. However: I don’t believe any charter cities actually exist yet, so it’s ok to speculate about them.
There’s a proposed charter city on an island off the coast of Honduras called Próspera. Here’s the website.
What would you do if you were designing a whole legal system from scratch in the 2020s?
Well – I recommend reading this whole piece: Scott Alexander at Astral Star Codex did a deep dive on everything published about the new city, and here is the Prospectus on Pr’ospera.
Here’s one small bit that caught my eye.
Property law today is based on 2D land area maps, with a hodgepodge of rights to light etc to deal with 3D, and surely that’s partially because - historically - flat maps are easy to draw? And now we can deal with 3D maybe properly law can change the fundamentals to think about it differently?
But I wonder if the folks involved played Minecraft growing up.
Voxels don’t have to look blocky.
GANcraft by NVIDIA (2021): Unsupervised 3D Neural Rendering of Minecraft Worlds.
Minecraft landscapes auto-transformed into forested hills, lush meadows, and beaches – check out the videos.
So I imagine a future VR experience that works a bit like this:
- You’re exploring a beautiful landscape. You say “See voxels” and the resolution downshifts to Minecraft-like blocks
- You edit the voxel map, a hill here, a tree there, changing the type of voxel for that building. You don’t need to be precise, it’s quick, a sketch
- Then: the prompt. The AI that will transform the voxel map is a generative AI configured by a prompt, so you give it a short paragraph about what you want this place to be: green fields, rolling hills, open skies, trending on artstation.
- Exiting edit mode, your blocky landscape upsamples to photorealism once again, voxels prompted into a raytraced sunlit 3D virtual world.
I don’t see why it should be any harder than that?
I know there are 3D edit tools that allow precision, but I feel like fine control is maladaptive in this situation. You want to be able to make something gorgeous, and easily, and have full creative expression. That’s what voxels provide, plus the application of AI which - thanks to the prompt - has all the almost-infinite variety of latent space.
(How soon? Using Diffusion Bee, the desktop version of Stable Diffusion, it takes 30 seconds on my M2 Mac to synthesise an image. So we’re ~8 Moore’s Law doublings away from realtime synthesised video at 10 inferences per second – I assume we can get to 40fps with non-inference tweening. 12 years till interactive hallucinatory VR! Gonna be wild. That’s an upper bound. I’m sure there are many shortcuts and optimisations which will get us there sooner.)
Oh: and when you see somebody else’s landscape, you want to be able to see how they did it, and have it graspable so you can copy it yourself. Voxels as View Source.
Dunno feels like VR (and AR, MR) will be a thing in the not too distant future.
So it’s worth thinking about ways that it can be a read-write medium. The environments that users make won’t have clean 3D models and precision-placed curves – instead they’ll be glitchy, piecemeal, ill-fitting, sketched. Accreted not architected. But they should also be easy and fun and decent looking and also, somehow, collaboratively constructed. Like playing Lego together, not working in some kind of prissy 3D Figma.
It would make for a neat, super customisable, multiplayer VR OS.
It strikes me that Minecraft blocks would be a good “language” for future users to have with the computer, and there’s nascent cultural readiness for it, and perhaps these fat voxels are somewhat under explored. It would be worth some R&D spend to figure it out.
If you enjoyed this post, please consider sharing it by email or on social media. Here’s the link. Thanks, —Matt.