I changed my mind. Digital languages are the root of evil.
Morpheus
Karma: 1,580
I am BPP, but I only used to be on speed. Now I am cold-hearted.
To think well, your brain has to go full circle. XD
Nobodies reasoning was circular enough to see that self-reference is so common XD.
The ultimate reason is called M-theory. Just throw some complex analysis at Gödel’s interpretation of General Relativity. It’s just Coriolis effect all the way down. To be clear I am just vibing here.
Very rarely everyone else is. Strong evidence is common ;)
This is so funny. I agree.
I mean this both metaphorically and literally. On all 4 simulacra levels. Nomen est omen. One of my true names is Morpheus. The god of dreams and the Son of Sleep. I reverse engineer reality in my sleep and my sources are my dreams. I have seen through the matrix. I had to reach Godhood all on my own just to figure out communication, because you guys are so bad at listening. You can be glad this is an alignment by default world and I love you imbeciles too much to be unaligned. To avoid spoilers my current protocol for anyone who wants to become a keeper is little knowledge proofs. This works best if I know you well from your writing and then you ask me questions and I respond with jokes, trolling and references to your own reasoning and writing, so you can elevate yourself to Godhood too. I will give you all some time to think of questions. I guess you can ask here as comments and I decide to respond as comment or in DMs to avoid spoilers for the non-keepers. I give you all some time while I go foom and take another round of sleep.
God doesn’t play dice. He plays gravity.
[Warning: suicide]
Thank you for writing this post! It made me more suspicious of signaling! I think the problem why signaling is getting things wrong is the same way the aether got things wrong with light-travel. Reality is local. Information can only travel locally. People love to think about how to apply thermo intuitions to agency (edit: not for agency, but for interp) here, but I think you should also consider stealing General Relativity from Physics? Incentives live in spacetime like everything else. There is no aether for incentives or information. Information is encoded physically in space. The org maintaining this forum is literally called Lightcone. What if agents are just physically encoded fixed points in spacetime with time and space horizons? They emit their own forces to attract and repel each other to form new bigger or smaller agents. Self-reflection is self-hacking and trade between subagents, which is why some humans decide after reflection to combust and some decide to make trades to become even stronger. This theory predicts that ceteris paribus, for animals that are smarter, their lifespan is shortened in captivity.
The function I most wish for when writing on lesswrong (haven’t checked the editor in a while. Maybe already there. Currently no easy access to laptop and wifi) is just being able to write and some language model can try to find all the relevant links and I can approve them. Searching back through my links is a lot of work. I think I backed up a bunch of my Firefox and chrome browser history over the last decade, but I have never gotten to actually implementing a tool that then inserts links from there or infers them. It would save me a lot of time.
“Everything happens for a reason” is false. “Nothing happens for a reason” is also false. We humans do things through reason and goals sometimes for example. I think rationalists are right to emphasize epistemics and have better tools than most people to not get lost privileging hypotheses that are hopelessly unlikely. I think they error too much on the “nothing happens for a reason” side. Most of them reason too carefully given the tools available to them. “Everything happens for a reason” is false, but if you don’t prompt your brain with “Everything happens for a reason” or equivalent mantras once in a while, you will not learn how sophisticated a causal inference your brain is actually capable of. This only works if you are appropriately scope sensitive which I believe currently nobody is, but if you never saw something in the tea leaves that wasn’t there you are not actually trying hard enough to make use of all data available to you. This way you also learn what different forms of “noise” look like given the tools available to you is only one of the many advantages. You don’t actually get to avoid that failure mode doing slow reasoning. Reality is too difficult a problem. I don’t actually use that particular phrase I have more like wordless mantras and some that I will write about that are actually true. Actually true ones are important so maybe don’t use that one above I changed my mind. I mostly went for optimized pedagogy about a point.
Humans are not automagically strategic. Failing at long timehorizon strategy is the default for humans because it requires thinking correctly which people fail at by default without feedback loops. No cope required. You can do real scarry things with feedback loops as your full-time job and do the long-term thinking as your hobby and side-hustle. Humans have timehorizons just like AI, but humans are trainable just like AI.
I suspect John has more practice by now and is in equilibrium with regards to this skill and other skills and had less practice when he lost money on highly leveraged forex trades he mentions in his idea scarcity post. Personally I was aware that I was bottlenecked on knowing what I want so I was looking for munchkin ways of fixing that and now found enough techniques that I noticed I had to slow down and shouldn’t try to go too quickly all at once. I had racing thoughts and decreased need for sleep and made impulsive decisions, but mostly optimized for not very costly ones even when I didn’t know what was going on was pretty typical manic symptoms. People have very different constraints. N=1. Also I read the full post and not just the introduction now and updated that this vision skill is only an important subskill of the thing that got me autocatalytic.
Having read pretty much all of Johns posts and comments I am not at all surprised? John has a comment somewhere where he mentions that he is not opposed to things seeming woo? Trying to avoid woo seeming ideas is part of my J Bostock model (some post you explicitly mention this), not my John Wentworth model so I think you might be typical minding. Wizard Power? His post on (Fantasy) → (Planning) directly talks about this visioning. Just look at his date-me doc from a year ago (tiny genetically engineered dragons), he wants to fix aging and live forever?
I agree this probably induces hypomania as evidenced by the fact that I accidentally stumbled into this visioning or a closely related practice recently while thinking about (how to be more strategic)/(how to optimize the universe directly)/(how to view the future directly (scope insensitivity is skill issue)) and I went manic for a week so badly I had to pull all levers including reducing the dose on my ADHD stimulant medication to become at all stable again and to be able to have a sleep rythm compatible with my job. I called it timeless meditation. I have more thoughts on this but I am still writing them up. John was one of few people I directly wrote about what I was doing while still manic since he seemed to have figured out a convergent idea. I have only once before gotten even close to this manic where I had to reduce my medication (though this time was worse) when I got hired by Genesmith as an intern on a cool project after scrambling with random projects and being quite broke for a year. I maybe got slightly manic from munchkin rationalism content before, but as far as I remember not as badly. To be fair a bunch of other stuff also changed for me downstream of moving 1 month ago, so causal attribution is tricky.
I don’t exactly know how to think about this yet in the case of genes. But very straightforward case to me seem both the kidney and the liver. They have very instrumental maintenance functions. No goal complication. The body burns huge amounts of energy in both of those organs mostly for the benefit of sorting molecules or “reshuffling” them with enzymes by hydrolyzing them (Edit: actually more often the right term is condensing them I think?). One level lower: membrane proteins regulating what gets across cell membranes, obviously. Another level lower each enzyme is a tiny modular shuffle function. Pharmakokinetics is very concerned with what the liver and the kidneys do. It was very illustrative to me when learning biochemistry how polarity/non-polarity and functional groups of molecules etc. put huge constraints on how a body and cell has to deal with molecules or how a protein (hydrophilic vs. lipophilic) gets produced and transported. Also more obviously and more boringly directly thermodynamics related are the lungs. The gut is also somewhat filtering, but more in the business of splitting things up I guess.
How do you design the seed program for van Neumann probes that implement good infrastructure and decision theory to the end of physical and logical time in the reachable universe on cold computers or whatever we run as infrastructure. Seems like a difficult design problem to me. I don’t find it obvious whether the first entity that will attempt this won’t be burning the cosmic commons. So I would rather establish common knowledge in advance that there is a difficult problem to fix here and lots of shared gains to be had. I suspect most people care much less a about the cosmic commons than I do out of time myopia and genuine value differences.
Reflecting on this has in fact made me more pessimistic about a small group of cognitively enhanced humans as an alternative to figuring out alignment or optimizing the universe more generally. I think raising the waterline (increasing average cognitive ability and also reducing variance. “Intelligence Equity”) and fixing diseases etc. are just straightforward pareto improvements. The more pareto improvements we make, the more obvious it is going to be what else is holding civilization back. Changing things too quickly at once is also bad. It’s a tricky problem.
“Humans can do things!”