|||

Multiverse theory and latent space

The story goes like this: Earth is captured by a technomagical singularity. The argument for our universe being a simulation goes both ways - it’s easier to crudely simulate a lot of universes than it is to actually take over the one we’re in.

So whether or not we’re currently in a simulation, we’re more likely to enter a fully simulated world by choice once it’s high enough resolution to fool our senses. And you can simulate different types of universes more cheaply than IRL (where it’s basically impossible).

So we will diverge realities based on preference. We will also converge our realities sometimes, but that’s less revolutionary because it’s what we already do. This is what Jack Clark calls reality collapse.

_1605601488398450689-Fkg-CjvacAAeYwN.jpg

Reality collapse is an inevitable product of the acceleration of information processing technology. Information can be compressed into a lossy format and it can be decompressed.

You don’t need all the information to represent the information. This is why hashing functions work. It’s also why math and science work: you don’t need a perfect understanding of gravity and relativity to get rockets. Lossy information can be good enough.”

Hashing is also how git works. The branching and merging of realities of code is managed by finding the difference between discrete states and hashing it to keep track of unique time steps in the universe” of the codebase. Okay, so now I’m getting to the point.

In the near term, reality generation tools will be good enough to fool most of our human senses. We diverge into different fantasy universes. Infinite Hall of Mirrors Simulated Realitopia society will be united by its dedication to keeping the simulations running.

We are already doing this. Video game fandoms and Netflix binges and podcasts in your ears at work. I have read over 12,000 pages of rationalist fantasy web serials this year.

We have babies raised by the YouTube algorithm. Now we’ll have kids raised by GPT. Our function will be, more than ever, wrangling and editing huge torrents of information. We will make tools for this. We will use AIs that control AIs that control AIs. We will develop Magic.

People will lose their minds to this, as magicians have always done. I summoned Jesus into the machine by asking it to roleplay. Imagine what real wingnuts will do. The origin of consciousness in the bicameral mind, but for keeps.

Robert Anton Wilson had this thing about the acceleration of information, he called it the jumping Jesus phenomenon.

The amount of information in the world when Jesus was alive is one Jesus. And it’s been exponentially increasing. Two Jesuses, four Jesuses, 8, 16. I don’t know where he was getting his numbers. It doesn’t matter. It’s obviously been growing even faster than the number of minds.

We now have global information processing, to the degree that Twitter is basically a hivemind. Global processing is often short-circuited by our various human fallacies and contradictory goals. It’s amazing that it works at all, but it does.

And that global information processing network can now simulate minds at roughly the same cost as human biopower. You can get a language model to act like a person if you summon one into it. In a hi-rez virtual world, how is that not a person?

Git is software for tracking changes to files and folders. It’s used by software developers to avoid having to call things finished_v4_final_final_v2.

_1605614248289112069-FkhJpjcWQAE5i0O.jpg

If you want to be able to fork and merge universes, you will need something like this to navigate collisions. This is also why popular culture has been navigating the multiverse for 50 years.

_1605615465568444417-FkhKwdlWAAIK62l.jpg

The Everett-Wheeler-Graham multiverse hypothesis has clawed its way from comic books to novels to cartoons to cinema. And it’s done so because people need help navigating the collision of worlds. Einstein’s relativity gave us postmodernism. Multiverse gives us… metarealism?

_1605616258828816385-FkhLemuXkAAMg_e.jpg

You don’t have to take my word for it. You can test this hypothesis right now with ChatGPT. Language is a world simulator. Large language models are mirror worlds. You can perturb them and see into different multiverses. Tactical guide available here.

View original

Up next When will a human level AI be built? American rainbow gatherings I didn’t think I liked America until I found out about international rainbow gatherings. Apparently everywhere else it’s just shanti hippies singing
Latest posts Deep Fates Program The Will Smith test Magic crystals Experimenting with Flux Fill Convert your Twitter archive into training data I love San Francisco Experimenting with Recraft v3 Experimenting with Recraft v3 text Criticism AI dating apps Infinite Keltham Machine Bot swarm Grug code editor Replicate Intelligence #12 Replicate Intelligence #11 Experimenting with Flux seeds Deepfits on Flux Mimic Replicate Intelligence #10 Sleepyhead Replicate Intelligence #9 AI cartoons Four takes on the same prompt Experimenting with FLUX.1 The dithered look Replicate Intelligence #8 Hyperstition Replicate Intelligence #7 Experimenting with LivePortrait Replicate Intelligence #6 What is “AI engineering” anyway?