Abstraction's End
home page
original site
twitter
substack
about
Linked TOC

Welcome


This is the canonical website for !c{my}{(Mariven)} writings. Trim the URL and you'll find my old site; while it was powered by Notion, this site was coded from scratch by me so that I could manually code up all the hypertextual features I want (see the dropdown key in the upper-right corner) while also having $\LaTeX$, images, and so on The downside of this is that, since I had to learn HTML/CSS/JS for this purpose, it's kinda rough. Especially if you're not using a modern browser on desktop. .

Mainly, I write about the structure of intelligence and the nature of abstraction; my "official" focus is mathematics—specifically, category theory and mathematical physics—but you won't find much here (except as examples given for other points, of which it's a very rich source). My goal is to find the truth of reality that I might be as effective as I can at doing what is right, and, currently, this means making sure that artificial intelligence is pointed towards creating a better future rather than a blank one.

You won't be able to pin me down in any specific school of thought, as I keep my identity loose and my thought eclectic, but you'll find it to be mostly congruent with a sort of sub-additive utilitarian !c{sentiocentrist}{prioritization of all sentient life} I am not an Effective Altruist—instead, you can call me an Ethical Actualist, taking it to imply many of the same viewpoints and conclusions, but not the community or culture. morality, a !c{critical realist}{we know reality only indirectly,
but sometimes veridically} metaphysics, a subjective Bayesian rationalist epistemology, a !c{metaphysical naturalist}{there is only matter acting according to
natural law, taking on various guises} ontology, and something between a cognitivist and functionalist theory of mind. The only thinker who I consider an influence is Kant—though I think he was wrong about most things, he was attacking the right problem (what are the necessary conditions for us to be able to know anything at all?) with the right strategy (determining the structure of cognition, and of reality's coupling to it)—and the only thinker who I seem to agree with unusually often is Yudkowsky; he says few non-obvious things, but nobody else seems to think in the same way, which is concerning.

As a perfectionist writing at immense levels of abstraction, technicality, and volume (I have hundreds of thousands of words of rough drafts and scattered notes, many of which would themselves just be considered complete essays by most people) , it takes me a pretty long time to write; being a grad student on top of that makes it far worse. But I do believe I'm discovering some genuinely new and exciting things here, so it's worth it.

A list of my main writings follows. The blocks that are filled in lead to completed essays; the others are the ones I'm currently working on. Whether complete or not, you can hover over each block to get an idea of what it's about.


Main


A program for exploiting the conceptive nature of cognition in order to solve any solvable problem that can be cognized.
Every Canvas a Mirror
Posted on August 29, 2022 (ver 1.0). Card imageMinjeong An, A Study on Pink Lens Effect
An exploration of the ways in which concepts fail to bind to reality, and of reality's failure to fall into natural kinds.
Non-Kindness
Posted on October 9, 2022.
A formalism for intuiting the convergent structure of optimization.
Worldspace
Status: significant parts complete (29k words, 75%)
Alt. title: Speeding Up Physicalism. What do the Gibbs paradox and the Doomsday argument have in common? They're reality telling us to abandon our basic conceptual principles. But the channel goes both ways: here's how we can read the nature of reality off of the way it forces us to think about it.
Making it Make Sense
Status: prototyped (30%)
(no preview available)
The Game
Status: several large, disparate rough drafts (50%)
Brains are structured prediction engines, and "the self" is a manifold of grooves they fall into. Nobody teaches you how to be a self, though—your concept of self is not just specified but structured in its construction, and this determines everything about your persona and life strategy.
Concepts of Self
Status: idea concretized, outlined (20%)


Branches


Quantifying the logical structure of the AI alignment problem to determine what exactly makes it hard to solve, and how it needs to be solved.
AI Safety is Team ∀
Posted on May 15, 2023.
Genetic influences on personality traits are neither deterministic nor ignorable. Instead, they're like a poker hand.
Hands We're Dealt
Status: prototyped (30%)
Which tastes louder, honeydew or cheddar? Which feels deeper, a shaving cut or a tension headache? A potential method for Turing tests against modern LLMs.
Global Synesthetics
Status: under construction (30%)
The consequences of an action depend on the scale of its implementation in a systematic way. Tools for locating yourself in the space of agents (a la TDT) and "renormalizing" the gauged morality of your actions.
Moral Renormalization
Status: under construction (30%)


Reference


An exhaustive report on our scientific understanding of the entire process of visual perception. A key reference for my neuropsychological metatheory.
How We See
Status: mostly done (75%)
A first-principles approach to developing general relativity from differential geometry from spatial intuition. A key reference for my metalogical metatheory.
Whence Relativity?
Status: under construction (30%)
A systematic port and expansion of earlier writings that attempt to dissect the foundations of probabilistic inference.
On Inference
Status: almost done (90%)