Permutation City

So far especially enjoying the focus on different levels of simulation resolution. The autoverse is a biology simulation with simplified atoms (or does that make it physics). While the simulation where people are being copied into operates at the level of organs. Thinking about how that compares to our current simulations... I think there is very little effort to build up out from atoms, right? Light simulations do simulate raycasting - a move towards the actual physical process. Would world simulations trend the same way - it would just be so much computing power! And what about LLMs. There's influence from neurons, and certainly about going from flexible, lower-level objects to produce emergent behavior. But the fact that it's operating in/on language... kind of by default means its higher-level? It's baking in language concepts unless it generalizes to the lower-level concepts from them...

Interesting resonance between this book's idea of throttling the tick speed at which the simulation of a person is run, and the Maths in Anathem, where monks live deliberately without outside contact. In both cases there is time shear to think about. Used as a proxy for general concerns about the speed at which information hits us, and whether it is in sync with what is best for us. I feel like "tick speed" as shown in the book has not been as much of a concern in real LLM systems. Obviously latency is a big deal, but I think a big part of it is that the lack of autonomy they've had - now with the agents with longer timelines it's easier to imagine. But I guess the book - focused on uploading yourself as life extension - is also just focused on a different, less task-oriented structure. For life extension continued consciousness is the goal, even if spread out over a long, long timeline.