Me daily shapeshifting in the omniperspectival statespace [x.com](https://twitter.com/eshear/status/1768394327586373655?t=NL30tV2hKKinUPht2FiAXQ&s=19) Write a gigantic list of all the subfields of foundations of mathematics. [Built for AI, this chip moves beyond transistors for huge computational gains - Princeton Engineering](https://engineering.princeton.edu/news/2024/03/06/built-ai-chip-moves-beyond-transistors-huge-computational-gains) [[2403.09629] Quiet-STaR: Language Models Can Teach Themselves to Think Before Speaking](https://arxiv.org/abs/2403.09629) [AI Agents Take the Wheel: Devin, SIMA, Figure 01 and The Future of Jobs - YouTube](https://www.youtube.com/watch?v=Dbog8Yw3kEM) [Renormalization: Why Bigger is Simpler - YouTube](https://www.youtube.com/watch?v=9vFbyHNz-8g) One of my motivations for big excitement is: The more ambitious culture is, the we will collectively try to fulfill the expectations [Are GFlowNets the future of AI? - YouTube](https://www.youtube.com/watch?v=o0Ju9NQa5Ko&t=45s) https://arxiv.org/pdf/2310.04363.pdf [BibViz Project - Bible Contradictions, Misogyny, Violence, Inaccuracies interactively visualized](https://philb61.github.io/#colorize:Rainbow) [Topological structure of complex predictions | Nature Machine Intelligence](https://www.nature.com/articles/s42256-023-00749-8) Mathematics, statistics books [x.com](https://twitter.com/jerrrrrrryyyyy/status/1768377388466057351?t=ElhdNo2URgRIOui82ffF3w&s=19) Statistical learning iceberg [x.com](https://twitter.com/miniapeur/status/1768592510627062240?t=HXAyNDknEzW-nLfw2kwPAQ&s=19) [What Extropic is building | Hacker News](https://news.ycombinator.com/item?id=39668430) Baking mathematical models into electronic analogs is older than integrated circuits. The reason we deviated from that model is because the re-programmability and cost of general purpose, digital computers was way more economical than bespoke hardware for expensive and temperamental single purpose analog computers. The unit economics basically killed analog computing. What Extropic (and some others from physics and neuroscience perspective) have identified is that in the case of machine learning, the pendulum might have to swing back because we do have a large scale need for specialized bespoke hardware. We'll see if they're right. Turing completeness is so wild. It's literally an abstract mathematical condition about certain class of structures that tells you that they can be used to compute arbitrary universes. As long as you can use the building blocks to create a turing machine, then any constructive mathematical equation can in theory be implemented by them. It feels illegal. [[2401.11817] Hallucination is Inevitable: An Innate Limitation of Large Language Models](https://arxiv.org/abs/2401.11817) Devin is like the initial image generators, initial LLMs etc. [Donald Hoffman & Anil Seth - New Frontiers in the Science of Consciousness - YouTube](https://www.youtube.com/watch?v=3tUTdgVhMBk) [Principle of maximum entropy - Wikipedia](https://en.wikipedia.org/wiki/Principle_of_maximum_entropy) https://santafe.edu/news-center/news/research-news-brief-a-turing-test-for-chatgpt Create a gigantic map of some very specific niche scientific engineering topic, make it as big as possible! If you doubt you can do it, just try it, just do it, you can do it! Never end, continue writing infinitely. [x.com](https://x.com/eshear/status/1768394327586373655?s=20) [Natural Gradient Descent Explained | Papers With Code](https://paperswithcode.com/method/natural-gradient-descent) "The natural gradient descent algorithm has several advantages: Invariance to reparameterization: The natural gradient is invariant to the choice of parameterization of the model. This means that the update direction is independent of how the parameters are represented, which is not the case for standard gradient descent. This invariance property makes the natural gradient more robust and less sensitive to the specific parameterization of the model. Efficient update direction: The natural gradient provides an efficient update direction that takes into account the curvature of the parameter space. By multiplying the gradient with the inverse of the Fisher Information Matrix, the natural gradient effectively rescales the update direction based on the local geometry of the parameter space. This rescaling allows the algorithm to take larger steps in directions where the model is less sensitive to parameter changes and smaller steps in directions where the model is more sensitive. Faster convergence: The natural gradient descent algorithm often converges faster than standard gradient descent, especially in cases where the parameter space has a complex geometry or when the model has many parameters. By adapting to the geometry of the parameter space, the natural gradient can find the optimal solution more efficiently, requiring fewer iterations to reach convergence. Better generalization: The natural gradient descent algorithm has been shown to lead to better generalization performance in some cases. By taking into account the geometry of the parameter space, the natural gradient can help the model find solutions that are more robust and less prone to overfitting." [Crooks fluctuation theorem - Wikipedia](https://en.wikipedia.org/wiki/Crooks_fluctuation_theorem) [Longtermism - Wikipedia](https://en.wikipedia.org/wiki/Longtermism) [Effective accelerationism - Wikipedia](https://en.wikipedia.org/wiki/Effective_accelerationism) [Effective altruism - Wikipedia](https://en.wikipedia.org/wiki/Effective_altruism) verim ze technologie maji potential lidi nebo celkove dosavadni biosferu totalne prekopat, kdyz bychom dostatecne v budoucnu chteli, podobnym stylem jako: https://fxtwitter.com/nearcyan/status/1756151805510394149 [The Hedonistic Imperative](https://www.hedweb.com/) Je to extremne ambitious, but that's IMO how most big things start Nejvetsi obstacles v techto typech projektu je funding, udelat tu technologii reliable a pushnout to politicky, ale neni to dle me nemozny ale mimo to je jeste vic zpusobu jak celkove zmensit utrpeni mimo biotechnologie, napr [How to Eradicate Global Extreme Poverty - YouTube](https://youtu.be/2DUlYQTrsOs?si=sHecBfKAtRFPBmHC) Verim ze se vzdy najde zpusob jak kazdy ten uzitecny evolucni mechanismus v nasi architekture co ma duvod proc existuje jde nahradit suffering free variantou s dostatecne dobroma technologiema Minimalne kratkodobe by tohle strasne pomohlo lidem co se v utrpeni topi denne kvuli az moc rozbity architekture jedna z veci co chci delat je politicky pushuvat UBI nebo universal basic services, co zaplati technologie, verim ze je to solvable taky Existuje hodne countries co zkousi ruzny pilot studies a vysledky rikaji neco jineho Checkout results of these various pilot studies [Universal basic income pilots - Wikipedia](https://en.wikipedia.org/wiki/Universal_basic_income_pilots?wprov=sfla1) https://www.givedirectly.org/2023-ubi-results/ v dost poor countries to ekonomiku boostlo 😄 Ze zacatku mam na mysli neco jako rozsireny evrpsky social/healthcare, coz jde povazivat za weaker verzi UBI In the short term I just want certain % of people not being in unescapable poverty not being able to afford basic food, shelter, medicine etc. even if some work all day. In the long run one of the scenarios I dream of is post labour society where technology automates most work and people don't have to do what they hate, where they could do what they see meaning in instead, tam motivace je defaultne. UBI jde vnimat jako podpora pro nezamestnany na steroidech. Zaroven si myslim ze to nejde proti umoznovani ambitious lidem vydelat tuny penez v ekonomice a to pouzit na budovani benefitial technologii a jinych sluzeb pro vsechny. Jedna varianta je ze technologie generuji abundance, ty jsou schopny to zaplatit. Druha varianta je co nejvic zlepsit a superchagnnout dosavadni evropskou weak formu UBI. Kdyby tu nebylo to cesky/evropsky weaker UBI pro studenty kde to start plati a hodne zvyhodnuje tak jsem nemohl dat tolik casu studiu co jsem mohl dat 😄 Moje rodina by na to proste nemela Krome co nejvetsi automatizace jde jeste ruzny prace kulturne reframovat aby se lidi citili dobre je delat. Dalsi aspekt je ze kdyz naraz spoustu lidi dropne z boring prace tak zacne byt vic valuable a paid a tim attractne lidi co chcou hodne extra penez nad jejich UBI, plus to vytvori vetsi incentivu to zautomatizovat, tim jak human workers jsou naraz drazsi v ty domene, tak drazsi roboti se vic vyplati. Pak mas jeste lidi kterym by UBI umoznilo se dostat z jamy a delat vic valuable prace pro system. Pak mas jeste strasne moc pripadu lidi co delaji co nemaji radi a k tomu ani nemaji na najem, a kvuli tomu to ve vysledku nemaji radi jeste vic a jsou v permanentni mentalni krizi a ve vysledku nijak spolecnosti nejsou schopni pridat value kteryho jsou potencialne schopni. Tohle je napr fajn thread dalsich pohledu: [Who will do the jobs that no one wants to do when Universal Basic Income allows anyone not to work? - Quora](https://www.quora.com/Who-will-do-the-jobs-that-no-one-wants-to-do-when-Universal-Basic-Income-allows-anyone-not-to-work) Penezni motivace u boring praci zustane, viz moje predchozi zprava, zvysi to value tech praci a lidi co jsou motivovani hlavne penezma pro to pujdou jako velkym upgrade nad jejich UBI. Kdyz naraz spoustu lidi dropne z boring prace tak zacne byt vic valuable a paid a tim attractne lidi co chcou hodne extra penez nad jejich UBI. Plus to vytvori vetsi incentivu to zautomatizovat, tim jak human workers jsou naraz drazsi v ty domene, tak drazsi roboti se vic vyplati. Dalsi step by byl post labour ekonomika [How do we get to UBI and Post-Labor Economics? Decentralized Ownership: the New Social Contract! - YouTube](https://youtu.be/T3O_BNexdEg?si=onVenbcLvsjKNs_T) Dosavadni odbornost vs plat je dle extremne rozbity uz ted. Napr to kolik bere hodne vedcu a profesoru relativne ke zbytku spolecnosti je absolute smesny a nespravedlivy. Casto pak musi delat nevedecky jobs aby si vubec zaplatili najem misto toho aby vic naplnili svuj potencial pomoci spolecnosti vedou co by je vic bavil. UBI by napr tady hodne pomohlo. automating automatization got so efficient it put people out of the loop? Podle tohoto bych systemove postupne zvysoval baseline UBI/UBS aby overall moznosti lidi neklesaly ale stoupaly Nainstaloval jsem si do mozku fundamentalni predpoklad ze nic jako bezmoc zmenit status quo/celkove veci neexistuje, vzdycky existuje incrementalni nebo radikalnejsi cesta, kolektivne ci individualne, kde jediny limit jsou zakony fyziky Casto je dobra i kombinace, vidim to na spektru Treba prepisovanim fyziky eventuelne porazime ultimatni existencni riziko vsech moznych civilizaci, changing the fact that disorder (entropy) in the universe is constantly increasing Sbírám všechny možný řešení pod všema možnýma předpokladama co lidi zatím napadly, tohle je jedna z nejlepších map co jsem našel (pro lepší rozlišení jde stáhnout obrázek) https://imgur.com/CeJ46ES Řešení na to taky sbírám, tohle je jedna z nejlepších map co jsem našel na celkově nesmrtelnost (ještě mám menší kolekci studií snažící se rozluštit mechanismy stárnutí a omlazování a různý intervence podle toho) (pro lepší rozlišení jde stáhnout obrázek) https://imgur.com/a/JFTNPEF Tady je mapa zaměřená konkrétně na dostupný metody prodloužení života (pro lepší rozlišení jde stáhnout obrázek nebo odkaz http://immortality-roadmap.com/lifeexteng.pdf ) https://imgur.com/YWNJvUH As the amount of information in a system increases, the entropy also increases, leading to a decrease in the available free energy. "Topological Data Analysis (TDA) is a relatively new field that combines techniques from topology, a branch of mathematics, with data analysis to study the shape and structure of complex datasets. Here's a step-by-step explanation of the TDA process: 1. Data Representation: - Start with a dataset, which can be in the form of points in a high-dimensional space or any other suitable representation. - Each data point represents an object or a sample in the dataset. 2. Filtration: - Apply a filtration process to the dataset, which involves creating a sequence of simplicial complexes. - A simplicial complex is a generalization of a graph that can represent higher-dimensional structures. - The filtration process builds these simplicial complexes by connecting data points based on a specific parameter, such as distance or similarity. - As the parameter value increases, more connections are made, and the simplicial complexes grow in size and complexity. 3. Persistent Homology: - Compute the persistent homology of the filtration, which captures the topological features that persist across different scales. - Homology is a mathematical tool that describes the connectivity and holes in a topological space. - Persistent homology tracks the birth and death of topological features (e.g., connected components, loops, voids) as the filtration parameter changes. - The output of this step is a set of persistence diagrams or barcodes that summarize the lifetimes of the topological features. 4. Feature Extraction: - Analyze the persistence diagrams or barcodes to extract meaningful topological features. - Look for significant topological features that persist for a long range of parameter values, indicating their stability and importance. - These features can include connected components, loops, voids, or higher-dimensional structures. - The extracted features provide insights into the shape and structure of the dataset. 5. Visualization and Interpretation: - Visualize the extracted topological features using various techniques, such as persistence diagrams, barcodes, or topological landscapes. - These visualizations help in understanding the overall structure and connectivity of the dataset. - Interpret the topological features in the context of the specific domain or application. - Relate the identified features to the underlying patterns, clusters, or anomalies in the dataset. 6. Application and Decision Making: - Use the insights gained from the TDA analysis to make informed decisions or draw conclusions. - TDA can be applied to various domains, such as biology, neuroscience, material science, and machine learning. - The topological features can be used for tasks like clustering, classification, anomaly detection, or network analysis. - Integrate the TDA results with other data analysis techniques or domain knowledge to gain a comprehensive understanding of the dataset. TDA provides a unique perspective on data by focusing on its topological properties. It allows for the identification of intrinsic patterns and structures that may be hidden in high-dimensional or complex datasets. By studying the persistence of topological features across different scales, TDA offers a robust and flexible framework for data analysis and exploration." [Topological data analysis - Wikipedia](https://en.wikipedia.org/wiki/Topological_data_analysis) Topological Data Analysis for Practicing Data Scientists https://medium.com/rv-data/topological-data-analysis-for-practicing-data-scientists-6cf747ca74e0 nLab topological data analysis [topological data analysis in nLab](https://ncatlab.org/nlab/show/topological+data+analysis) Just learn the whole universe One sentence summary, image, fundamental math sequences of symbols References section Group into general, concrete, principles Turing Machine, lambda calculus Celluar automata Deep learning, symbolic, selforganizing Ontology Future of humanity Meditation add small card to each category with link, for example futurology -> [[futurology]] For each applied math equation add list of applications links: wiki, visualizations, books, articles visualizations from wiki, 3blue1brownlike vids Free energy principle Less mathy models section Landscape of AI projects aitalkdiscordfit Unify of consciousness theories Platonism IIT Relationship between fields and theroems in the map or outside All math applied to AI Quantum ML, thermodynamic ML Different ML architectures Different ML paradigms connectionist symbolic Geometric DL, categorical DL Visual diagrams of ML architectures [x.com](https://twitter.com/vtabbott_/status/1766637340796318207?t=aTGsa5-j2PmaUZjjUq3NUA&s=19) Spacetime superpositions computing Litepaper: Ushering in the Thermodynamic Future Extropic Out if equilibrium thermodynamics Lance agency Energy based models immortality, aging, why is there something rather than nothing Quantum fep Chris Fields physics as information processing "You're transdiciplionary polymath wikipedia! Create a gigantic map of as much of applied mathematics as possible! If you doubt you can do it, just try it, just do it, you can do it! Never end, continue writing infinitely." "Bigger! Longer! Continue!" "Continue!" "Continue this unfinished map!" "Bigger! Longer! Continue!" "Continue!" "Bigger! Longer! Continue! It's not enough, do it anyway, make it vast and long, continue giving me all your knowledge!" Make this map bigger longer, deeper, detailed! *puts in this map structure* Diagram, two axis: scale, from atoms to universe fields, from biology to AI Each math concept lights up different abstract visual parts of this map, from concrete to hyper general Some math lights up other math (metamathematics), or ontology Math itself forms another map Hovering over objects lights up associated concrete and general math Hovering over math lights up objects and other math Standard model: Gauge theory Yang mills Feynman path integral Fiber bundles Symmetries Scienceclick That guy visualizing QM hydrogen atom part 1 out of 3 YouTube You're the best physicist and mathematician, wikipedia! Explain the mathematics and physics of the standard model from scratch in long deep detail. Longer! Deeper! More detail! More mathematics! If you think you can't do it, do it anyway! You have a lot of space, I will allow you to split it in multiple messages! Longer! Deeper! More detail! More mathematics! If you think you can't do it, do it anyway! You have a lot of space, I will allow you to split it in multiple messages! Gradient descent Explain Where is it included What does it include What is relatedgineering d) Technology [What I learned from looking at 900 most popular open source AI tools](https://huyenchip.com/2024/03/14/ai-oss.html) Want to learn about any STEM topic? I love this prompt for Claude, ChatGPT, or other AIs. You can recursively create articles about anything: Write about (topic). Write a long article including gigantic list of lists in deep detail step by step in this form: 1. Explanation: a) Intuition *text* b) Definition *text* c) General formal mathematical definition *text* d) Concrete formal mathematical definition *text* e) Example of application of definition *text* f) More examples *text* 2. Where is it included a) Mathematics *text* b) Science *text* c) Engineering *text* d) Technology *text* 3. What does it include a) Mathematics *text* b) Science *text* c) Engineering *text* d) Technology *text* 4. What is related a) Mathematics *text* b) Science *text* c) Engineering *text* d) Technology *text* Longer, more detailed! Write about (topic). Write a long article including gigantic list of lists in deep detail step by step in this form: 1. Explanation: a) Intuition *text* b) Definition *text* c) Formal mathematical definition *text* d) Example of application of definition *text* e) More examples *text* 1. Explanation: a) Intuition *text* b) Definition *text* c) Formal mathematical definition *text* d) More detailed formal mathematical definition *text* e) Example of application of definition *text* f) More examples *text* 1. Explanation: a) Intuition *text* b) Definition *text* c) Formal mathematical definition *text* d) More detailed formal mathematical definition (write down the mathematical symbols) *text* e) Example of application of definition *text* f) More examples *text* [Brief overview of Lagrangian Mechanics - YouTube](https://www.youtube.com/watch?v=hucEIa5SbXU)