‘We shall take the brain from your poor rotting body and keep it healthy in a special nutrient bath. We shall have it connected to a machine that is capable of inducing in it any pattern at all of neural firings and is therein capable of bringing about for you any sort of total experience that it is possible for the activity of your nervous system to cause or to be.’

When I was in college, my aunt, who runs a neuromechanics lab, proposed the idea of creating a virtual reality simulation of standing on the edge of a cliff, and using it to study how people coordinate their muscles to balance themselves. There are a log of degenerative neurological conditions which can affect a person's balance, and so the idea with this is that such a simulation might be able to help someone more quickly learn to recover balancing ability because of the perceived sense of urgency (without them having to actually stand on the edge of a cliff).

I've been seriously wondering if something similar is happening to me right now. Perhaps it's a byproduct of being an employee at OpenAI - increasingly the face of obsolescence for millions of Americans - or perhaps it's a function of spending two years throwing my heart and soul into building a company that I believe in, and becoming increasingly jaded by a startup ecosystem that, so far as I can tell, seems almost entirely uninterested in actually making money.¹

My wife and I have recently been tinkering on a side project to build our own connectomes. We got inspired by the Weight Space Programming blog post from a few days ago, and before that, the viral fruit fly connectome that a bunch of people on Twitter were using to do things like control robots. We both spent a couple of years in college working on computational neuroscience - back when using your brain for learning seemed to matter - and we both have a soft spot for brain-inspired models of intelligence and connectomics.²

That thought that a lot of smart, high-achieving people seem to have in the back of their head, about how their life is actually just a simulation, or maybe we're in the afterlife but haven't realized it yet, is becoming increasingly hard to ignore. I think if I had a lower degree of reflexivity I would probably have gone psychotic by now. Not to be melodramatic. It's just that the part of my brain that tries to rationalize and quantify my life is going haywire, and getting mega-fucked with by the part of my brain that insists on giving me agency.

There is the base rate $P(\texttt{sim})$ which comes from the fact that my own life seems to be going in a very unusual trajectory compared to baseline. The majority of people I grew up with are now doing very predictable things, whereas I'm working on building a machine god / humanity's next stage of evolution.³ So calling this very unusual thing that I get to do with my life $\theta$, in Bayesian terms:

from which it is clear that as $P(\theta)$ decreases (meaning, the relative unusualness of your life increases), $P(\texttt{sim})$ must increase commensurately. Thus, an infinitesimally-unusual life must only be possible in simulation.

For the last few years I've been able to ignore this thought by assuming that there is probably some major issue in my math here (let's say $P(\texttt{sim} | \theta)$ is not constant - I have not yet formalized this in Lean). That really broke down for me when I started messing around with making connectome megakernels and realizing that, actually, simulating all of the ODEs to model a human-level brain is probably attainable on hardware that I could run in my closet.

Which of course leads to the annoying implication that maybe I am actually a brain running in a simulation in someone else's closet somewhere. And perhaps my brain was itself designed by some super-intelligent AI model. In which case, given that my life seems fairly comfortable (although not extravagently so), I must be on the right side of Roko's Basilisk.