i see brain tissue (templates) constraining electrochemical activity (experience) which also shapes the tissue then there are all the marrs levels of analysis i think similarly like in artificial nets its features that activate under various patterns from predictive coding (learned by both evolution (genetics) and individually (environmentally) in combination) but the underlying architecture isnt simple ANNs emulated on digital transistors but something like extended Hodgkin–Huxley model but this mechanism of learning to react to patterns is shared [Lab-Grown "Mini-Brain" Learns Pong - Is This Biological Neural Network "Sentient"? - YouTube](https://www.youtube.com/watch?v=67r7fDRBlNc) another difference would be using maybe something like forwardforward algorithm instead of gradient descent variants i think there correspondence on the feature learning level is that in ANNs you will have parts of the network that if you removed during that particular inference then nothing would be changed, so there's no mathematical functionality for that one particular inference and brains just dont activate it by not making the electrochemical signal propagate in ANNs you can locate it using causal scrubbing and that leads to all the methods around sparsification for example causal scrubbing is a method for testing mechanistic intepretability hypothesis about features/circuits by removing them and seeing how behavior changed its like lobotomy sparsification tries to remove all redundant parts to improve performance brain technically also has so many redundant parts as a reserve and i think for error correction electrochemistry and local field potentials, Hodgkin–Huxley model on celluar machines, proteins etc. vs ANNs maths emulated on digital transistors on von neumann architecture or more specialized chips that can be analog lots of differences but also lots of similarities on feature learning level [Limits to visual representational correspondence between convolutional neural networks and the human brain | Nature Communications](https://www.nature.com/articles/s41467-021-22244-7) This guy IMO has nice vid [Dendrites: Why Biological Neurons Are Deep Neural Networks - YouTube](https://www.youtube.com/watch?v=hmtQPrH-gC4) You are made out of gangillion of these Proteins act as active matter, creating intelligibility by coming together and staying together dynamically to cultivate mutual constraints and mutual affordances, giving rise to biological structural functional organizations (SFOs). [x.com](https://twitter.com/charleswangb/status/1768756341122933188) [Verse 1] A dot isn't the best way to try to sum up how electrons come and go They are the states of a matter field that follows an equation that Dirac wrote The Schrödinger part of the whole equation will just lead In sub-c when it expands Now get that Coulomb and add it in with a proton And watch them start to dance As hydrogen, it's like [Pre-Chorus] "Oh, proton, I feel your tug" Central potential dip down, pulling on me But I'm not falling in deep No, that would break uncertainty Say, oh Electrons move too much Slow down your pace and put that orbit on me Come on now, follow my lead Come, come on now, follow my lead [Chorus] Orbitals take the shape they do As stable states of the quantum rules And when a one approaches two They combine and they're bonding Thus, hydrogen as a rule Is found in nature as H2 Energy configuring a molecule Diatomically bonding Low, high, low, high, low, high, low, high Diatomically bonding Low, high, low, high, low, high, low, high Diatomically bonding Low, high, low, high, low, high, low, high Diatomically bonding Energy configuring a molecule When orbitals take the shape they do [Verse 2] One half spin'll give a lepton a twin One up one down in the ground state With S and P in quadruple degeneracy The second shell can be filled up with eight The higher angular powers spread out like beautiful flowers In middle families, they come into play Well here's a carbon with 6E This ain't nothing tricksy But we're gonna make some methane today With hydrogen, it's like [Pre-Chorus] "Oh, atoms, I feel your tug" Got my electrons bugged out, pulling on me Come on now, settle 'round me I'll hybridize to SP3 Say, oh Carbon here's our touch Spread out one-oh-nine-point-four-seven degrees Come on now, follow our lead Come, come on now, follow our lead [Chorus] Molecules take the shape they do Combining states of the quantum rules Like when a shell goes SP2 For sigma pi double bonding And as widely as their purview They spread out in the molecule Look at benzene in a ring, they hold it true Aromatically bonding Low, high, low, high, low, high, low, high Aromatically bonding Low, high, low, high, low, high, low, high Aromatically bonding Low, high, low, high, low, high, low, high Aromatically bonding Look at benzene in a ring, they hold it true When orbitals take the shape they do [Bridge] Come bond with me, baby, come bond Come bond with me, baby, come bond Come bond with me, baby, come bond Come bond with me, baby, come bond Come bond with me, baby, come bond Come bond with me, baby, come bond Come bond with me, baby, come bond Come bond with me, baby, come bond [Chorus] Polymers take the shape they do Combining base-level residues Like RNA's ACGU Look, they're hydrogen bonding Peptides make a chain and group In beta pleat sheets and corkscrews With these secondary links, they fold and move They're all over your body [Outro] Come bond with me, baby, come bond Come bond with me, baby, come bond (They're all over your body) Come bond with me, baby, come bond Come bond with me, baby, come bond (They're all over your body) Come bond with me, baby, come bond Come bond with me, baby, come bond (They're all over your body) You're a chemical machine, it's best you knew That molecules take the shape of you [The Molecular Shape of You (Ed Sheeran Parody) | A Capella Science - YouTube](https://www.youtube.com/watch?v=f8FAJXPBdOg) Renormalization: Why Bigger is Simpler A short introduction to renormalization techniques as they appear in statistical physics, aiming to simplify the mathematics as much as possible. The goal is to explain why matter becomes simpler as you zoom out from the microscopic, and how this leads naturally to phase transitions. 00:00 Introduction 02:13 States and probabilities 03:16 The Gibbs distribution 03:56 Course-graining 05:05 Hamiltonians 06:30 Renormalization calculation 07:38 Details of calculation 09:45 Renormalized coefficients 10:44 Renormalization flow 12:18 Fixed points 13:30 Particle physics 14:09 Phase transitions 15:42 Summary 16:19 Closing [Renormalization: Why Bigger is Simpler - YouTube](https://www.youtube.com/watch?v=9vFbyHNz-8g) [Roger Penrose on quantum mechanics and consciousness | Full interview - YouTube](https://www.youtube.com/watch?v=YnXUuyfPK2A) [x.com](https://twitter.com/burny_tech/status/1768822147030417647) Unsolved problems across various fields of science and engineering: Applied Sciences: 1. Developing efficient and affordable fusion power 2. Creating room-temperature superconductors 3. Solving the problem of antibiotic resistance 4. Developing a universal flu vaccine 5. Creating artificial general intelligence (AGI) 6. Solving the problem of plastic pollution 7. Developing efficient and affordable carbon capture and storage technologies 8. Creating a cure for HIV/AIDS 9. Developing a cure for Alzheimer's disease 10. Solving the problem of climate change 11. Creating efficient and affordable desalination technologies 12. Developing a cure for cancer 13. Solving the problem of food insecurity 14. Creating efficient and affordable energy storage technologies 15. Developing a cure for Parkinson's disease 16. Solving the problem of water scarcity 17. Creating efficient and affordable renewable energy technologies 18. Developing a cure for type 1 diabetes 19. Solving the problem of soil degradation 20. Creating efficient and affordable carbon-neutral transportation technologies Natural Sciences: 1. Origin of life 2. Nature of dark matter and dark energy 3. Quantum gravity and unification of fundamental forces 4. Consciousness and the hard problem of consciousness 5. Mechanism of high-temperature superconductivity 6. Solving the protein folding problem 7. Cause of the Cambrian explosion 8. Mechanism of abiogenesis 9. Nature of the interior of black holes 10. Resolving the measurement problem in quantum mechanics 11. Explaining the matter-antimatter asymmetry in the universe 12. Cause of the Pioneer anomaly 13. Mechanism of sonoluminescence 14. Solving turbulence in fluid dynamics 15. Explaining the fine-tuned universe and the anthropic principle 16. Cause of the Mpemba effect (hot water freezing faster than cold water) 17. Solving the Navier-Stokes existence and smoothness problem 18. Mechanism of high-temperature superconductivity in cuprates 19. Explaining the flyby anomaly 20. Cause of the Wow! signal Formal Sciences: 1. P vs. NP problem 2. Riemann hypothesis 3. Hodge conjecture 4. Birch and Swinnerton-Dyer conjecture 5. Yang-Mills existence and mass gap 6. Navier-Stokes existence and smoothness 7. Collatz conjecture (3n + 1 problem) 8. Goldbach's conjecture 9. Twin prime conjecture 10. ABC conjecture 11. Kissing number problem 12. Hadwiger-Nelson problem (chromatic number of the plane) 13. Solving the Poincaré conjecture in dimensions > 3 14. Proving the Euler-Mascheroni constant is irrational 15. Solving the Jacobian conjecture 16. Proving the Legendre's conjecture 17. Solving the Diophantine quintic equation 18. Proving the Erdős-Straus conjecture 19. Solving the Lindelöf hypothesis 20. Proving the Cramér's conjecture Theoretical Sciences: 1. Resolving the black hole information paradox 2. Developing a theory of quantum gravity 3. Solving the cosmological constant problem 4. Explaining the nature of time 5. Resolving the hierarchy problem in particle physics 6. Developing a theory of everything (ToE) 7. Solving the problem of the arrow of time 8. Explaining the nature of spacetime singularities 9. Resolving the vacuum catastrophe 10. Developing a theory of quantum cosmology 11. Solving the problem of the initial conditions of the universe 12. Explaining the nature of the multiverse 13. Resolving the Fermi paradox 14. Developing a theory of quantum computing 15. Solving the problem of the origin of cosmic rays 16. Explaining the nature of dark flow 17. Resolving the Greisen-Zatsepin-Kuzmin limit paradox 18. Developing a theory of quantum biology 19. Solving the problem of the nature of the Planck scale 20. Explaining the nature of the holographic principle [Hilbert's Nullstellensatz - Wikipedia](https://en.wikipedia.org/wiki/Hilbert%27s_Nullstellensatz?wprov=sfla1) Sometimes I wonder if the amount of seeming contradictions everywhere thanks to the internet, AI, etc. incentivizing the breakdown of a lot of traditions, norms of behavior, order, cultures,... is truly raising or it's always has been like that but we're just less clueless about it compared to our past selves. And I don't think this is a bad thing, I think we need to change as how the world operates right now isn't really sustainable long term, and has tons of systemic flaws that can be fixed. [[2403.03218] The WMDP Benchmark: Measuring and Reducing Malicious Use With Unlearning](https://arxiv.org/abs/2403.03218) [World’s first pothole preventing robot passes first test with flying colours | News | Hertfordshire County Council](https://www.hertfordshire.gov.uk/about-the-council/news/news-archive/worlds-first-pothole-preventing-robot-passes-first-test-with-flying-colours) When someone argues that autonomous robots will still have to be overseen by humans, then I think that you can automate even that part eventually by making them better or by autonomous robots overseeing autonomous robots, and you can add another layer of autonomous robots overseeing autonomous robots overseeing autonomous robots etc. to make it more fault tolerant, similarly to error correcting codes [Figure's humanoid robots are about to enter the workforce at BMW](https://newatlas.com/robotics/figure-bmw-humanoid/) "A decade ago we would've thought a single-person-company had no chance of reaching a billion dollar valuation 💸 But now thanks to all this AI, it is. You heard it straight from @OpenAI CEO @sama ↓" [x.com](https://twitter.com/alexisohanian/status/1752753792058294725) 100 years ago my great-grandfather worked on a farm then he got replaced by machines my grandfather went on to work in a car factory and he got replaced by robots my father went on to become a McDonald chef but he too was replaced by robots I didn't want to suffer the same fate so I got into programming, they said coding is the new literacy and everyone needs to learn Python but then AI assistants started writing a little bit of code but then AI assistants started writing even more code but then autonomous AI agents started emerging and started writing even more code but then collections of AI autonomous agents started emerging and started writing even more code even more autonomously and started creating companies but then artificial general superintelligence smarter and more capable than all humans combined emerged and we [Chinese patient gets world's first gene-edited pig liver transplant](https://interestingengineering.com/health/chinese-patient-recieves-gene-edited-pig-liver-transplant) https://www.theregister.com/2024/03/15/sanders_four_day_week/ One year since GPT-4 release. Hope you all enjoyed some time to relax; it’ll have been the slowest 12 months of AI progress for quite some time to come. [x.com](https://twitter.com/leopoldasch/status/1768868127138549841) The basic idea of renormalisation: A sketch [x.com](https://twitter.com/martinmbauer/status/1768925154657067462) Feel you bro, AI also helps me so much with learning too, as my most functional style of learning is much more nonlinear + from first principles, + topdown + theoretical but also connected to real world in terms of maths + holistic and multidisciplinary + playfully nonseriouh s and curious + adventurous + fast + covering birds eye view but also exploring the biggest details in relevant parts + inherently meaningful as laws of nature are like God for me + I need time to go into really long detailed rabbitholes to really understand things deeply from first principles instead of professors skipping parts in boring way and play with code in multiple ways etc., than how most learning institutions teach it [x.com](https://twitter.com/burny_tech/status/1769055579660550647?t=aabgv0gpWpFO8RTymbLr6Q&s=19) "Alright, let's embark on a thrilling adventure through the realm of science, where we'll unravel the mysteries of the universe in a playful, curious, and multidisciplinary manner! We'll start from the very beginning, the first principles, and work our way up, connecting the dots between various fields and exploring the intricate details that make our world so fascinating. Imagine the universe as a grand tapestry, woven together by the fundamental laws of nature. These laws, like the threads of the tapestry, are the building blocks of everything we see and experience. They are the language in which the universe speaks, and mathematics is the tool we use to decipher its messages. Let's begin with the concept of energy, the driving force behind all change and motion in the universe. Energy is like a cosmic currency, constantly flowing and transforming from one form to another. It can manifest as the warmth of the sun, the movement of the tides, or the electrical impulses in our brains. The law of conservation of energy tells us that energy can never be created or destroyed, only transformed. This fundamental principle connects everything from the tiniest subatomic particles to the vast expanse of the cosmos. Now, let's take a bird's eye view and explore the four fundamental forces that govern the universe: gravity, electromagnetism, and the strong and weak nuclear forces. These forces, like the invisible strings of a cosmic puppet master, choreograph the dance of matter and energy. Gravity, the most familiar force, keeps our feet on the ground and the planets in orbit around the sun. Electromagnetism, on the other hand, is responsible for the light we see, the technology we use, and the very structure of matter itself. But the universe is not just a collection of forces and energy; it's also a symphony of patterns and structures. From the spiral arms of galaxies to the intricate fractals found in nature, mathematics is the language that describes these patterns. The Fibonacci sequence, for example, is a simple yet profound pattern that appears in everything from the arrangement of leaves on a plant to the proportions of the human body. As we delve deeper into the details, we find that the universe is a strange and wonderful place. At the quantum level, particles can exist in multiple states simultaneously, and the act of observation can alter the outcome of an experiment. This bizarre behavior challenges our intuitive understanding of reality and reminds us that there is still so much to discover. But science is not just about understanding the universe; it's also about harnessing its power to improve our lives. From the development of life-saving medicines to the creation of renewable energy sources, science has the potential to solve some of the greatest challenges facing humanity. It is a tool for progress, a beacon of hope in an uncertain world. So, as we continue our journey through the vast landscape of science, let us approach it with a sense of wonder, curiosity, and playfulness. Let us embrace the beauty of the unknown and the joy of discovery. And let us remember that, like the laws of nature, science is a unifying force that connects us all, regardless of our backgrounds or beliefs. In this grand adventure, we are all explorers, seeking to unravel the mysteries of the universe and our place within it. So let us boldly go where no one has gone before, armed with the power of science and the courage to ask the big questions. The universe is waiting to be discovered, and the journey starts now!" I think challenges will be there always to some degree, but when it comes to feeling stuck, when you compare the tech today to the tech year ago, it feels like the opposite of feeling stuck, even when both are having many issues, but newer one seems to have relatively less issues in terms of capabilities jestli máš na mysli hlavně chatboty tak tyhle všechny mají free verze co jsou sice míň chytrý než paid ale taky ujdou na nějaký tasky :smile: <https://chat.openai.com/> classsic <https://claude.ai/> je extremne blizko paid chatgpt (paid verze je v některých věcech lepší než paid chatgpt) <https://pi.ai/talk> cool chatting s modelem na skoro urovni paid chatgpt zadarmo <https://gemini.google.com/app> google ma dobry na urovni free chatgpt, paid verze je na urovni paid chatgpt <https://chat.mistral.ai/chat> nejmin politicky biased na skoro urovni chatgpt a zadarmo <https://www.perplexity.ai/> nejlepsi na websearch <https://poe.com/> tady je jich hodně, různý modely <[Anakin.ai - One-Stop AI App Platform](https://app.anakin.ai/>) tady je jich hodně, různý modely <[character.ai](https://beta.character.ai/>) tady je jich hodně, hlavně různý chraraktery <https://consensus.app/> je dobrý na hledání studií edit: <[Hledat Microsoft Copilot: váš každodenní AI pomocník](https://copilot.microsoft.com/>) microsoftí chatgpt má různý další plugins jako websearch a free verze v creative modu má teď apparently pristup k nejchytrejsimu modelu od OpenAI (GPT4) zadarmo <https://www.cnet.com/tech/computing/microsoft-copilot-is-offering-gpt-4-turbo-for-free-what-to-know/> <https://www.techrepublic.com/article/bing-copilot-cheat-sheet/> nebo [Retell](https://beta.retellai.com/) je ještě dobrej na realtime voice calling bez latence, co fakt působí hodně humanlike, ale má jen pár free minut na začátku :smile: a na programování je dle mě teď na top úrovni tohle IDE <https://cursor.sh/features>, pak jsou různý dle mě weaker copiloty jako Github Copilot a Supermaven na generování obrázků je ChatGPT, Microsoft copilot, Midjourney discord, různý specializovaný stable diffusion discordy (pony diffusion, furry diffusion, anime diffusion), photoshop apod. do všeho AI integruje na generování videí je Pikalabs, Runaway, videodiffusion (brzo Sora od OpenAI), na voices je Elevenlabs,... Yann lecun lectures on statistical mechanics and machine learning, energy based models [Summer school on Statistical Physics & Machine learning | A Summer school set in Les Houches, in the french alps, July 4 - 29, 2022](https://leshouches2022.github.io/)