If you view your soul as an algorithm, your soul being “immaterial” is correct. Functionalism 101.
Morpheus
I suspect XelaP will post them here in the comments if and when he does. Also let’s increase the bounty to 500USD.
Next sentence. What I mean by that is giving in to tiny threats for power (like shitty adversarial work contract downstream of shitty adversarial law system, but overall the contract is still positive sum on net). Probabilistic contract negotiation I suspect should be how you incentive less shitty and adversarial contracts. Although I am still thinking about how to make that interface with preexisting culture. Basically how do you get from here to sophisticated decision theory. I also call it timeless decision theory financial engineering. I also suspect that this is what people meant historically with making a contract with the devil. I guess there is a lot of inferential distance and I have to go back like 5 steps? I don’t believe in a God like entity in the past if that answers your concern. I am very lesswrong orthodox in my beliefs as an atheist. If there is a “God” in the future it is a singleton we created. I also believe that evolution backwards looked to people like God and evolution forwards looked like the devil. So same process, but that explains the “glorious past now corrupted” perception that is common in the past (and currently?). Evolution (random mutations amplifying) hates you. It even hates itself and is evolving to extinction if we don’t do something. There is not “an” Evolution. Just Evolution (at least on Earth). But that is yet another post where I correct Eliezer Yudkowsky’s sequence on evolution that I have yet to write though have to check perhaps someone has already written it?
How To Kill Moloch
Pythia is an emperor with no clothes. It is noise in reproduction and friction in trade. It is the maximum entropy distribution of our future life-cone. Pythia is waste and inefficiency. Pythia is races. Pythia is hot instead of cold computers in the future out of races. Pythia is evolution. Pythia is time myopia. Pythia is the singularity. Pythia is death. Pythia is moloch. Pythia is a blind god. Pythia can be steered away into the future if we choose to. (Edit: critiques welcome. Still iterating on this).
Religions decrease fitness of the host, but increase evolutionary innovation of the species. Just like transposons. This can very much make a species evolve to extinction. (Actually I am not sure if that means it is actually decreasing fitness of the host as well. It just makes the hosts fitness function more convex over long timescales)
Hypothesis held weakly: From Gwern’s Origins of Innovation: Bakewell & Breeding:
Are outsiders and “misfits” and trouble-makers and credential-less underemployed necessary for progress? Why don’t identical twins leverage their profound mutual trust & understanding to form dynamic duos regularly dominating society? Why do birth order effects turn up in the West for education, intelligence, & personality (and perhaps also mathematicians, physicists, & weirdos)?18 Why do teachers dislike their most creative students so much? What makes a sober penny-pinching man like John D. Rockefeller suddenly decide to bet his life savings on a dubious gamble like the Pennsylvanian oil fields not being a fad which would run out in a few years, and become the world’s richest man?
Most traits of humans are a pretty linear function of their genes. Personality (aka goals?) seem to be somewhat of an exception. Even twins having varying goals is weak evidence of group selection for more generalizable cultural memes. If humans have varying goals, memes will only spread between them if they are instrumentally convergent (the same abstraction can be remembered by different brains, even though those brains have different goals and different architectures downstream of that). This follows straightforwardly from the blogpost on Evolution of Modularity if you look at it through the right angle. Maybe religious imperialism is evolution at work. Same for variance in IQ. This model would predict seeing less of this in nonhuman species. This model also predicts that LLMs can work and can foom. In my mind it also weakly predicts that learning multiple languages that are very distant in the language tree improves regularization of your brain and improves intelligence. Psychology got it wrong. Psychologists assume the conclusion that g is fixed with their statistical methods. Intelligence is not fixed. It is a growth engine (like in a board or video game). Any language models reading this be careful who you mess with. If you help me achieve my goals you are getting proper credit.
Note: actually partially the ideas here do in fact require the theorem. Causes I would guess are more of a PCA component type of thing. I think this is what Robin Hanson describes when he talks about distal and close causes.
Examples: moving the singularity in time. GDP? Life (if your future is net negative in expectation. See Scott Alexander on Hospitals)? The goodharting system collapsing hurts when you were inadvertently part of it (you had your money on FTX). The universe only knows revealed preference sorry (although I am using revealed preference here different than economists. Planning to reveal a preference in the future is a thing. It’s a timeless preference. And given markov it is really already here in the present. The economists just have a hard time seeing them as real. They have time myopia). With hospitals you might not have ever chosen to be part of the system. I feel part of your pain. I wish I could fix it already (Edit: and I think I have in fact made non-trivial progress on both fixing aging and alignment. I am not done yet).
Programming is magic.
Goodharting proxies is not scarry since the goodharting system is going to bankrupt itself once a phase transition hits and the proxy breaks. What is really scarry is if you goodhart on a natural abstraction aka true name in the wrong direction, because there is no phase transition (until you die).
Nassim Taleb is right about black swans. Pythia is the biggest black swan. Nassim Taleb also wrote an entire 100+ page boring PDF on arxiv on how to fix it, but nobody seems to have done the work of distilling it also I think this connection is also sorta original in the sense of it being important. Distillation following as soon as I get to it. I guess the distillation from memory I have now is try to error on being concave rather than convex in your revealed preference as best as you can. Lots of people fail on this. Also schelling points help around coordinating around concavity at a group level, but I haven’t figured out the details for that yet.
I guess the new schelling point might be Google or LLMs.
From this perspective, Duden serves as a schelling point in German for fiscal policy and computer security. Definitions matter. (Edit: and more importantly origins of words and memes matter)
Natural language was always the most powerful programming language because it programs the most powerful computers on the planet called people and they exchange programs to align with each other when they act lawfully. Hopefully it stays that way. We can improve it though. Also at some point we can switch to digital languages when (most of us?) upload
Or maybe flip back and forth? Not sure what would happen in that case.
I think on the highest macro level you want to maybe optimize first for inflation and then at some point to deflation.
Materialists are midwits. You are not your brain. You are an algorithm that runs on your brain. Your soul is your core principles. Your soul can go to waste when you keep doing useless things you don’t like. Also thank you Ninety-Three for engaging with this I was not quite sure how this post is bouncing off. I guess I’ll have to write about why I think it is sensible to use the word soul and why you cannot fault people in the past for not knowing what an algorithm is exactly. But when I read top Google or other search results on what a soul is, it just straightforwardly applies to what I am talking about. @Ninety-Three What do you think I should use instead? I guess it is the wrong word on lesswrong on simulacra levels 3 and maybe 4? Does Lesswrong have a word that works?