Everything is born out of nonequilibrium thermodynamics Physics Intelligence Big Bang Intelligence Big Crunch You should think in probability distributions over probability distributions over any data Circular levels of generalization: Mathematics of our physical universe -> Mathematics of the space all possible universes -> Space of all possible mathematics -> Space of all possible philosophical constructions -> Space of all possible thoughts -> Space of all possible qualia experienced by the space of all possible minds implemented in the space of all possible universes i have this hope that some form of future AI will one day help us get unstuck when it comes to quantum gravity and i wanna work towards that beyond current ML analysis in CERN beyond current LLMs producing stuff that is not mathematically coherent beyond current initial attempts at neurosymbolic systems (connecting LLMs with Lean or maybe systems like DreamCoder) beyond current physics inspired ML architectures for predicting dynamical systems beyond current AI assisted design of experiments etc. so i'm currently collecting and learning everything about the intersection of AI and physics that is relevant to this Screw building machine Einstein, machine Von Neumann or machine Feynman. Let's build machine Grothendieck. Do you ever think that it’s completely mindblowing, that we observed the nature around us, inscribed runes into sand, somehow turning it into rocks capable of computing arbitrary logic, then observed nature again, and tried to encode its laws into that rock, creating imperfect simulations of physics into which we add our dreamed up stories, then speedrunners reverse engineer every single detail of those imperfect simulations, similar to physicists reverse engineering nature, and discover ways to exploit the physics for speed, and invent new tools on top to optimize that speed even further? If "free will", defined in some way without breaking the laws of physics and biology, is somehow localizable or not in the laws of physics/biology, seems to have zero effect on anything in the physical world, because the laws of physics/biology that govern us and everything are still the same whether we know some kind of equation for free will or not, if it exists, and if we somehow localize some kind of "free will" law or not doesn't change that fact at all. It would just cause us to have more detailed models of the universe, which is good. To first approximation, my mind has roughly 3 philosophers inside it: - Science and engineering philosopher, who wants every model to be implementable in our physical reality to be real, who wants models that predict physical reality empirically, and the degree of realness of a model depends on the degree of ability to predict that the model has in the domain that it's trying to predict or/and control - Platonist philosopher, who loves the existence of all abstract concepts, mostly loving mathematical structures, and seeing them all as real in a giant platonic space of all possible structures - Epistemological anarchist philosopher, who loves swimming in the space of all possible philosophical assumptions, in the space of all possible qualia, where anything experiencable is real Survival of the stablest is the ultimate imperative across all scales and domains of reality Wikipedie je fakt totální zlato lidstva, úžasně obecný zkompresovaný kolektivní vedění, s pointerama všude když člověk chce větší hloubku, miluju ji Wikipedia is truly a total goldmine of humanity, a wonderfully general compressed expression of our collective intelligence, with pointers to everywhere when you want more depth, I love it so much The only constant is change = Everything can be described by differential equations Effective Curiosity is about maximizing the total understanding of the universe in the form of having as comprehensive and as empirically predictive models of as many physical phenomena as possible Loss function that maximizes stability of as many coherent structures as possible over time My biggest goal is to collectively understand the mathematical structure of the universe with our brains, AI systems, and other technologies My religion is accumulating as many empirically predictive models about the universe and intelligence as possible The difference between mathematicians and physicists is the difference between masturbation and sex I want to accumulate as many mathematical theoretical results about the equations and mathematical structures that we use in physics and in AI as possible All possible trajectories in the phase space of all possible configurations of particles in all possible universes I believe it is highly likely that we will only have better and better approximations of the universe over time, and never a full model, because we're limited finite computational observers in comparison. Epistemologically my favorite definition of truth is empirical predictive power, so the degree of truthness of a model depends on how well it predicts empirical data in its domain. Ideally the model is described mathematically. And it has to be falsifiable and not a tautology or too vague. No model that we have and that we will ever have will be able to predict all possible empirical data that we can measure with absolute 100% accuracy, because we're limited observers, with limited measuring and modelling brains, with limited measuring and modelling tools and technology, with finite computational resources, etc. relative to the whole system, the whole universe and everything in it with all its complexity, that we're trying to model and predict. My nerd mind cannot comprehend that normies don't think that figuring out the mathematics behind intelligence and behind the universe isn't the most fascinating thing in the whole reality Categorical quantum machine learning https://en.wikipedia.org/wiki/Quantum_machine_learning https://en.wikipedia.org/wiki/Categorical_quantum_mechanics https://arxiv.org/abs/2402.15332 https://ncatlab.org/nlab/show/quantum+field+theory https://youtu.be/ZSmORp3Bm2c?si=ddkDhYZ7YVRHUxPC " My favorite AGI system would model the equations present at all scales of physical reality and do applied and pure math with them. It can be from scratch model or a system of models, with singular or hybrid architecture, in neural or neurosymbolic or other paradigms. But for now we could connect all specialized math/physics/natural science AI/ML systems to LLMs as tools, like AlphaGeometry, FermiNet, AlphaFold, etc., and give the LLM a bank/database/graph of all equations, proofs, mathematical definitions, heuristics, optimization algorithms, etc., that it could execute in Python/R/Mathematica/Lean/etc., that we have so far in applied mathematics, natural sciences and pure mathematics. " Digital omnidisciplinary scientist We haven't even scratched the surface in terms of AI discovering novel groundbreaking scientific theories How to create, automate creation, or build entities that create something that is as out of distribution as possible, while still making sense to human pattern recognition machinery, and either giving us all sorts of interesting emotions like art, or being scientifically useful like scientific discoveries, or mathematically interesting like mathematical results, or philosophically interesting like philosophical theories? The meaning of life is to learn all the mathematics of the universe and of intelligence and become it I can't wait for the physics of alien intelligence AI is physics of adaptive information processing with limited resources Which psychedelics do category theorists and string theorists do? Scientific method in the limit applied to everything As comprehensive, as simple, as predictive model of everything as possible i want to run a neural network on a computer running on top of turing complete fluid dynamics governed by navier stokes equations https://arxiv.org/abs/2507.07696 the future of intelligence is hidden in fluid dynamics https://arxiv.org/abs/2304.02637 https://x.com/burny_tech/status/1966336302485496058 The free energy principle made simpler but not too simple https://arxiv.org/abs/2201.06387 https://www.youtube.com/watch?v=PNYWi996Beg I am sometimes thinking, if we want AI to discover some very out of distribution novel highly abstract mathematics, maybe one way could be doing an open ended search in the space of toposes? Topos theory is a branch of category theory that generalizes notions of inclusion and logic, which are traditionally based on set theory. It studies different mathematical universes, toposes (topoi), with their own laws of how mathematical objects within them behave, an example of such universe is sets, but there are many more. https://x.com/burny_tech/status/1967090536784834714 https://www.youtube.com/watch?v=gKYpvyQPhZo https://www.youtube.com/watch?v=o-yBDYgUqZQ Topos Theory for Generative AI and LLMs I found a nerd snipe on the intersection of AI and topos theory, paper from 10 days ago, but i dont know how legit/rigorous/practical/etc. it is Topos theory is a branch of category theory that generalizes notions of inclusion and logic, which are traditionally based on set theory. It studies different mathematical universes, toposes (topoi), with their own laws of how mathematical objects within them behave, an example of such universe is sets, but there are many more. https://www.arxiv.org/abs/2508.08293 A mathematical reality cathedral made of axioms governing mathematics and mathematical physics governing the universe with all applied mathematics across all scales of reality at all levels of coarse graining I want to build and update a unified probabilistic model of how the entire world works, and use that model to make predictions and decisions. Effective Curiosity Maximizing the total understanding of reality by building models of as many physical phenomena as possible across as many scales of the universe as possible, that are as comprehensive, unified, simple, and empirically predictive as possible. Intelligence and fundamental physics, which are subsets of this, are the most fascinating to me. I think modelling the whole world in fully unified way and in total accuracy is impossible, even for all of science with all our technology, because we're all finite limited agents with limited computational resources and time, limited modelling capability, we get stuck in local minimas, from various perspectives, and so on, and all we have is approximations, that predict reality to a certain degree, but never fully all of reality in perfect accuracy. " My current primary subfield I'm exploring is reverse engineering the black box of deep learning with interpretability. It endlessly fascinates me how these systems can do stuff like folding proteins, or just human language generally, without us fully understanding how they do it on mechanistic level, because so far there are many empirical results, and little scientific theory explaining it, which is slowly emerging. I'm curious about many other types beyond just deep learning. More generally, intelligent systems fascinate me, with many biological ones. I'm fascinated by what makes our brains work, what makes our society work, what makes animals generally work. I wonder how does all of that emerge. I'm fascinated how we learn and do math and physics. I wonder what is happening in the brains of people that come up with very transformative scientific breakthroughs. I scientifically wonder about the technical intersections and differences between all types of systems with capabilities that we associate with intelligence. And even more generally, I wonder about the mathematics of the universe. It's absolutely mindblowing to me how you get all this structural complexity in our universe from the standard model of particle physics, general relativity and so on. So many endlessly fascinating mysteries that I'm infinitely curious about. So much to learn. So much unknown that we still didn't figure out. " Derive the standard model Lagrangian and general relativity Lagrangian from as much scratch as possible by starting with the foundations of mathematics, axioms underlying mathematics, axioms and definitions in various subfields of mathematics, relevant theorems, axioms and definitions in physics, theorems in physics, and so on. Diffusion models are nonequilibrium thermodynamics turned into kind of intelligence " A gigantic synthesized map in great length and depth that includes all these maps in it: Map of all STEM, philosophy, intelligence, and all mathematics used Map of all knowledge Map of all mathematics Map of all philosophy Map of all philosophy under STEM Map of all technology, and all mathematics used Map of all physics, and all mathematics used Map of all fundamental physics, and all mathematics used Map of all intelligence, and all mathematics used Map of all AI, and all mathematics used Map of all computer science, and all mathematics used Map of all cognitive science, and all mathematics used Map of theories of everything Map of everything from first principles: foundations of math, philosophical assumptions Map of all mathematics from the simplest to the most advanced Map of everything from first principles: from the fundamental scales of reality to all emergent sciences, philosophical assumptions, and all mathematics used Map of all physical objects from the smallest to the biggest, both natural and engineered, and all mathematics used Map of all physical objects from the least complex to the most complex, both natural and engineered, and all mathematics used Map of what all mathematical structures are an abstraction of and a concrete instance of Map of what all concepts are an abstraction of and a concrete instance of Map of all theory Map of all practical systems Map of all applications Map of all theory in AI, and all mathematics used Map of all practical systems in AI, and all mathematics used Map of all applications in AI Map of all theory in physics, and all mathematics used Map of all practical methods in physics, and all mathematics used Map of meta theories of everything Maps of all the history of all the fields Map of all ways of how these maps include eachother as subsets (submaps) Map of all connections between all these maps on a meta level as connections between nodes, like which math is used in which subfields of what sciences, or analogies between fields " Universally approximate the function of the whole universe's structure across all scales and levels of granularity Current ML/AI systems aren't really trying to be replicas of human thinking like many people seem to think. They're mix of all sorts of insights that we got from neuroscience, optimization theory, probability theory and statistics, computer science, physics, applied mathematics, pure mathematics, philosophy, cognitive science, psychology, empirical random playing around with random stuff, etc. into one complex system. Neuroscience - Neuronky začaly jako "hmm neurony jdou popsat Hodgkin Huxley modelem, ale vlastně taky jdou hodně simplifikovat jako graf s vertices jako neurony a edges jako synapses, what if we added more neurons and...". Nebo reverse engineering neuronek, obor co se nazývá mechanistic interpretability, je v podstatě digitální neurověda. Optimization theory - Gradient descent přichází odtuď a je to nejpopulárnější učící algoritmus neuronek, a dokonce je původně z astronomie. Applied mathematics: Neuronky jsou aplikovaná lineární algebra, víceproměnná reálná analýza, teorie pravděpodobnosti,... A přes tools těhle metod jde dělat tuna triků pro efektivnější tréning. Probability theory and statistics - Neuronky jsou technicky statistická metoda, existuje statistical learning theory, nebo existují bayesovy metody, atd. Physics - Hopfield networks jsou prekurzory k neuronkám co známe teď, a jsou spin glass system, za co Hopfield dostal nobelovku z fyziky. Diffusion modely jsou vlastně díky nonequilibrium thermodynamice. AdamW optimizer do gradient descentu přidává "momentum". Flow matching učící algoritmus se učí velocity fields. Na chápání neuronek se teď víc a víc hází statistická fyzika. Pure mathematics - Jsou pokusy o pure math modely deep learningu, včetně toho categorical deep learning. Nebo je geometric deep learning co používá na analýzu deep learning architektur teorii grup. Computer science - Např turing completeness je též relevantní, existují architektury jako neural turning machine. Neuronky jsou algoritmy a computational complexity je taky důležitá pro efektivitu. Optimizing concrete software and hardware matters. Etc. Cognitive science/Psychology/Philosophy - Etika se řeší v alignment problemu a control problemu. Kognitivní vědy řeší problém toho jaký fyzikální systémy mají "pravou" inteligenci, reasoning, nebo co je to vědomí, a jak tyhle slova zadefinovat (uh to je rabbithole). Empirical random playing around with random stuff - Tak některý improvements vznikly. Dost tvoření modelů je v podstatě alchemie A je toho milion víc. :D i want to be computronium mining the last evaporating black holes for energy resisting the heat death of the universe trillion trillion trillion trillion trillion trillion trillions years from now and no conservatives will stop me from doing that https://x.com/i/status/2005059007044825261 https://youtu.be/uD4izuDMUQA?si=mrWK35vEjgXmpWJ- https://youtu.be/TBikbn5XJhg?si=lNtrPaHefbTygCkB https://www.youtube.com/watch?v=lAJkDrBCA6k AI is the best field for me because it uses almost all different scientific disciplines in terms of constructing AI systems. And at the same time almost all fields use some form of AI in some way (in both better or worse ways, both where it is appropriate and where it is not, for both good and for evil, unfortunately). Everything is causal networks To truly identify a causal relationship, you have to do an intervention https://youtu.be/9suqiofCiwM?si=dYfg4PG45Ay4-zUR The set of all mathematical structures The set of all physical objects And all the relationships between them (instance of, generalization, includes,...) And all the types (scientific fields they're part of, mathematical fields they're part of,...) Different perspectives give us different forms of predictive power over different subsets of incomprehensibly complex reality https://youtu.be/pO0WZsN8Oiw?si=tF_gPvVjEhIliY-a The brain is whatever most advanced technology we have at this moment https://youtu.be/pO0WZsN8Oiw?si=tF_gPvVjEhIliY-a Maybe *future* AI systems won't have singular consciousness but a ton of tulpas since they have continuous persona space that they will be navigating and spawning tons of nested conscious subagents I wish to understand everything on the most fundamental microscopic level, and all the emergent patterns that arise from those first principles, in both natural sciences and formal sciences But there is finite time and capacity with this brain, so specialization is needed, rip https://youtu.be/M2iX6HQOoLg?si=3BOF4cevrFldthwE Blaise calls a function on top of some collection of matter a spirit. Joscha calls the higher order causal pattern a spitit. Interesting. I want to collect as many concrete mathematical models and what they use from pure math that predict as many various things in physical reality as possible Physics of intelligence: General physics theory of learning of features and circuits and their formation across all architectures, learning algorithms, data, substrates, systems, etc.