decision theory

preface: my arguments below are eccentric and self-serving. even though they are flawed and may not be true, i choose to believe them and it makes my life better.

my mind constantly drifts to the thought of our reality being a simulation. the two barriers i see are energy and cognition.

to put it in perspective, the entirety of humanity’s current energy needs is less than a trillionth of the sun’s output. less than a trillionth. that single fact is genuinely absurd to think about. try to imagine what we could do with even just a millionth of the sun’s energy. it makes the energy problem go from a fundamental barrier to an engineering one, one that a sufficiently advanced civilization could solve to run a simulation such as the one we may be in right now.

now let’s consider compute. given the current trajectory of LLMs and world models, it seems difficult to argue that human-level cognition won’t be fully simulatable within the next few centuries.

it begs the obvious question: if we can see a plausible path to building this kind of simulation, why would we assume we’re the first reality to get there? i find it impossibly hard to believe that we are the first reality where all of the factors have aligned to make simulations like this feasible.

perhaps we are just some advanced versions of AI with some incredibly complex reward function that we internalize as conscience. maybe this world is one of many RL rollouts inside some high-fidelity world model. its possible we’re all just a collection of ai agents for the sake of prediction. the scariest part of this for me is the idea that my continued existence may depend on the utility i provide to whatever simulation i’m a part of. this is a thought i haven’t fully made peace with, and maybe shouldn’t.

or maybe I’m wrong. and this isn’t a simulation. maybe the 1 in 100 trillion chance that we have evolved to intelligence is real. in which case, i’ve never felt a stronger urge to live for myself.

humanity has been around for a couple hundred thousand years. if we thought about the age of the entire universe as the volume of the pacific ocean, humanity’s time here is 1/16,000,000th of a single raindrop. that number is supposed to make you feel small, but i find it does the opposite. when nothing persists, what’s left is only what you actually lived. i particularly enjoy reading this excerpt-

in 100 years, we will all be buried with our relatives and friends. strangers will live in our homes which we fought so hard to build, and they will own everything we have today. our descendants will hardly know who we were. how many of us remember our grandfather’s father? after our death, we will be remembered for a few years. and a few years later, our history, photos, and deeds will all fade away. we won’t even be memories. maybe one day we will stop to analyze these questions and understand how ignorant and weak the dream of obtaining everything really was. i’d change all of that just to live and enjoy the walks i’d never taken. the ungiven hugs and kisses to our children and our loves. that is what life is all about.

i have come to peace with the fact that i will not matter after the next 100 years. it reminds me how the insignificance of myself should be the very thing I cherish.

both of these arguments shape my decision making philosophy. if i’m in a simulation, i feel an obligation to push some frontier. and if not, i feel an obligation to live every minute for myself; to maximize my own serotonin and dopamine. every decision i make should do one of those things, and ideally will achieve both.

rishabh