Ml street talk [The Free Energy Principle approach to Agency - YouTube](https://youtu.be/zMDSMqtjays?si=xUDD0LRWorFR_Er_) [The Free Energy Principle approach to Agency - YouTube](https://youtu.be/zMDSMqtjays?si=s42kAXtf9wMlfWsP) [TherML: The Thermodynamics of Machine Learning](https://research.google/pubs/therml-the-thermodynamics-of-machine-learning/) [List of unsolved problems in economics - Wikipedia](https://en.wikipedia.org/wiki/List_of_unsolved_problems_in_economics) [Paradox - Wikipedia](https://en.wikipedia.org/wiki/Paradox) [List of paradoxes - Wikipedia](https://en.wikipedia.org/wiki/List_of_paradoxes) [Topological quantum computer - Wikipedia](https://en.wikipedia.org/wiki/Topological_quantum_computer) [Wikipedia:Contents/Outlines - Wikipedia](https://en.wikipedia.org/wiki/Wikipedia:Contents/Outlines) [#9: Dwarkesh Patel - Podcasting, AI, Talent, and Fixing Government - YouTube](https://www.youtube.com/watch?v=ggSbkRh6J_8) [Can Quantum Fluctuations Create a New Universe? - YouTube](https://www.youtube.com/watch?v=IFIa7fmrOeU) [Map of Artificial Intelligence - YouTube](https://youtu.be/hDWDtH1jnXg?si=LWlqUIRmuvr9y-XK) [2020 Machine Learning Roadmap (87% valid for 2024) - YouTube](https://youtu.be/pHiMN_gy9mk?si=4FNUjRneMDAyN_Ax) i am calling for a 6-month pause on all frankenMoEs until someone explains why this should work at all, ever https://twitter.com/main_horse/status/1746779017674702853?t=fB-hYAg8paBv1Lmxfb1pGw&s=19 Artificial human intelligence Artificial superhuman intelligence Narrowness and generality is a specialization and performance in the given domain/s AIXI Linear algebra is a study of flat things [Map of Artificial Intelligence - YouTube](https://youtu.be/hDWDtH1jnXg?si=zPmvnebohGMb7gpI) AI in scifi https://twitter.com/burny_tech/status/1747010375709515906?t=0UI9AGXNCxcI6MSf2rbBwA&s=19 terminator Explain mathematics of Kullback–Leibler divergence [Kullback–Leibler divergence - Wikipedia](https://en.m.wikipedia.org/wiki/Kullback%E2%80%93Leibler_divergence) Explain mathematics of reinforcement learning [Mathematical optimization - Wikipedia](https://en.m.wikipedia.org/wiki/Mathematical_optimization) Decision tree Random forest Semisupervized learning Selfsupervized learning Links for 2024-01-15 AI: 1. Interview with Rich Sutton: “We talk about topics like reinforcement learning, the OpenMind Research Institute (OMRI), Keen Technologies, how to do good research, the problem of scale, superintelligence, and so much more.” [I Talked with Rich Sutton - YouTube](https://www.youtube.com/watch?v=4feeUJnrrYg) 2. AI’s New Frontier: Turning Sparse Sensor Data Into Gold [New twist on AI makes the most of sparse sensor data](https://discover.lanl.gov/news/1113-ai-sensor-data/) 3. Secure, Governable Chips: Using On-Chip Mechanisms to Manage National Security Risks from AI & Advanced Computing [Secure, Governable Chips | Center for a New American Security (en-US)](https://www.cnas.org/publications/reports/secure-governable-chips) 4. Can a combination of smaller models collaboratively achieve comparable or enhanced performance relative to a singular large model? Blending Is All You Need: Cheaper, Better Alternative to Trillion-Parameters LLM https://arxiv.org/abs/2401.02994 5. If the West didn’t have “real adversaries,” Karp said he’d be one of the people trying to contain AI’s use in the military. “But the reality of our life now, as Israel knows, is our adversaries are real, dangerous and they go so far outside the norms of behavior.” [Bloomberg - Are you a robot?](https://www.bloomberg.com/news/articles/2024-01-10/palantir-supplying-israel-with-new-tools-since-hamas-war-started) [https://archive.is/d9Jho] Science and Technology: 1. "43% of all banking systems are still using COBOL, which handles those $3 trillion daily transactions, including 95% of all ATM activity in the US, and 80% of all in-person credit card transactions. [But] very few people are interested in learning COBOL these days." https://www.pcmag.com/articles/ibms-plan-to-update-cobol-with-watson 2. LMU researchers have developed new high-performance nanostructures to obtain Hydrogen (H2) with the help of solar energy. [Harvesting more solar energy with supercrystals - LMU Munich](https://www.lmu.de/en/about-lmu/structure/central-university-administration/communications-and-media-relations/press-room/press-release/harvesting-more-solar-energy-with-supercrystals-2.html) 3. What does the Cerebellum Do Anyway? [What does the Cerebellum Do Anyway? - by Sarah Constantin](https://sarahconstantin.substack.com/p/what-does-the-cerebellum-do-anyway) 4. The case against the lab leak theory: Every reason why lab leak theories fail and a natural origin of covid is more likely. https://medium.com/microbial-instincts/the-case-against-the-lab-leak-theory-f640ae1c3704 (see also: [Peter Miller - YouTube](https://www.youtube.com/@tgof137/videos)) We have already done quantum experiments with gravity https://arxiv.org/abs/1510.03078 Testing consciousness Andres Levín advertazial collaboration, consciousness turning off neural correlates Demarcation problem Theorists have found a way to solve complex Feynman integrals numerically by reducing them to simple linear algebra. #ScienceMagArchives https://www.science.org/content/article/method-solving-notorious-calculus-problems-speeds-particle-physics-computations?utm_medium=ownedSocial&utm_campaign=NewsfromScience&utm_source=Facebook&fbclid=IwAR1xuKlhugmuVVN4mZeUI0QXIDdsDPRZL7F1k9wlujHT3NGE6quZ_SWyYCc [Reflections on Fusions of Consciousness](https://qri.org/blog/reflections-on-fusions-of-consciousness) [Neurotechnology - Wikipedia](https://en.m.wikipedia.org/wiki/Neurotechnology) Von Neumann entropy ChatGPT convo [ChatGPT](https://chat.openai.com/share/65c34ef3-cb3a-47d9-bba3-4ad9e687e488) [ChatGPT](https://chat.openai.com/share/68ab2824-9bee-4ca6-8378-4f3fa64dad67) Fusion of functional advantages of biological and nonbiological sentient physical substrates into a superpowerful hybrid is the path forward https://twitter.com/burny_tech/status/1747050841666638111?t=Btcsl47A6sFguVMQEOTgrQ&s=19 https://twitter.com/algekalipso/status/1746949072613953985?t=_tRxnB9cFmQc7DcN6NQ_kQ&s=19 5-meo-dmt Browse map of x on YouTube "Our analysis supports a top–down structure in quantum mechanics according to which higher-order correlations can always determine lower-order ones, but not vice versa." https://twitter.com/DrYohanJohn/status/1746929185463906594?t=bhVn3U_ocNyQymOI2Vk11A&s=19 Psychonautwiki Twitter is gain of function future forming memetic warfare laborator Network state https://twitter.com/DermoreLEI/status/1741444894508363939?t=cccHs4FwbRbcTcPnknYMzA&s=19 [Learning few-shot imitation as cultural transmission | Nature Communications](https://www.nature.com/articles/s41467-023-42875-2) [Introducing: Consensus GPT, Your AI Research Assistant - Consensus: AI Search Engine for Research](https://consensus.app/home/blog/introducing-researchgpt-by-consensus/) https://fxtwitter.com/ConsensusNLP/status/1745133106103804124 [ChatGPT - Consensus](https://chat.openai.com/g/g-bo0FiWLY7-consensus) nerve signals travel 99.99999999% slower in your body than signals travel in fiberoptic cables why didn’t evolution settle on a fiberoptic-like delivery system? there are bioluminescent animals so it should be possible There’s no need for such speed in biological systems because the bottleneck in response times is mechanical Need non-linearity, hard to have optical non-linearity https://twitter.com/thecaptain_nemo/status/1746965793949270055?t=uXfIoaTsB85DZHHca16Haw&s=19 Symmetries in physics, biology.. And neural nets.. dynamical systems in general [Making and breaking symmetries in mind and life | Royal Society](https://royalsociety.org/blog/2023/04/symmetries-mind-and-life/) https://arxiv.org/abs/2302.03025 [Noether's theorem - Wikipedia](https://en.wikipedia.org/wiki/Noether%27s_theorem) [Phys. Rev. E 84, 041929 (2011) - Symmetries, stability, and control in nonlinear systems and networks](https://journals.aps.org/pre/abstract/10.1103/PhysRevE.84.041929) https://arxiv.org/abs/2205.10513 https://arxiv.org/abs/2302.00843 https://arxiv.org/abs/2304.12686 https://arxiv.org/abs/2302.03189 solution to the boundary problem in cognitive science Andres paper [Ray Kurzweil: Singularity, Superintelligence, and Immortality | Lex Fridman Podcast #321 - YouTube](https://youtu.be/ykY69lSpDdo?si=CKIi3rkzOal0Q_-3) [Ray Kurzweil presents "A Fireside Chat on the Future" with Charlie Kam and Bill Faloon - YouTube](https://www.youtube.com/live/WNKLGuwG8Hc?si=llK3LPVzLszk0Xrn) [The Future Of Humanity 2045 (Ray Kurzweil) - YouTube](https://youtu.be/N9ep1IH3tao?si=5ro9AIpY_wNAF6ci) Group theory [ChatGPT](https://chat.openai.com/share/f6ebddce-44e1-4c29-8dc3-5d55025058f3) Nanotechnology https://www.nytimes.com/2023/12/08/health/fda-sickle-cell-crispr.html GPT4 finance https://twitter.com/emollick/status/1744592247574696126?t=xBdsSkZn8nchHMHAesoyZQ&s=19 Pathscope interpretability Patchscopes: A Unifying Framework for Inspecting Hidden Representations of Language Models https://twitter.com/ghandeharioun/status/1746946621215003041?t=wWxhtoL5VBJwWr9WmLgsag&s=19 https://arxiv.org/abs/2306.02797 [Can large language models identify and correct their mistakes?](https://blog.research.google/2024/01/can-large-language-models-identify-and.html) https://openai.com/blog/how-openai-is-approaching-2024-worldwide-elections The Evolution of Operating Systems: Agents are the New Daemons https://www.linkedin.com/pulse/evolution-operating-systems-agents-new-daemons-david-shapiro-cizbe?utm_source=share&utm_medium=member_android&utm_campaign=share_via [Extropianism - Wikipedia](https://en.m.wikipedia.org/wiki/Extropianism) Realtime AI assistant that is monitoring all content you're consuming or all discussion you take part in on the internet and real life and gives fact checking, corrections, different reformulations, illustrations, explaining in deep technical detail, fun facts, expanding on it, connecting it with what you know from before in the history of the interaction or scanning your brain and uploading in it what you wish No single thing is perfect. Each thing have various degrees of different advantages and disadvantages relative to eachother from a different kinds of perspectives with different kinds of goals. It's about optimally maximizing the degree of advantages and minimizing the degree of disadvantages. [Quantum gravity - Wikipedia](https://en.wikipedia.org/wiki/Quantum_gravity) [The Cause of Depression Is Probably Not What You Think | Quanta Magazine](https://www.quantamagazine.org/the-cause-of-depression-is-probably-not-what-you-think-20230126/) Nous Hermes 2 beats Mistral-Instruct MOE and becomes the best open-source model. https://twitter.com/bindureddy/status/1747011304857235774 https://twitter.com/rohanpaul_ai/status/1747012187284197704 https://fxtwitter.com/Teknium1/status/174699038473835773 GPT-3.5 level Ai is on iPhone today https://twitter.com/NickADobos/status/1747048830581997785 [Phenomenological quantum gravity - Wikipedia](https://en.wikipedia.org/wiki/Phenomenological_quantum_gravity) https://avi-loeb.medium.com/a-test-of-cosmic-inflation-and-quantum-gravity-4a8b9e149181 testing astrology - its statistical noise https://www.facebook.com/spencer.greenberg/posts/pfbid0a2BPDdJjqBzVHwLeUW6D1mSLcAxNHFnagFsGPcM7bTDbQxabFt1D4UvDQXrZdzeTl [Can astrological sun signs (or zodiac signs) predict facts about people's lives? We tested it.](https://www.clearerthinking.org/post/we-tested-the-predictive-power-of-astrology-here-are-the-results) https://pbs.twimg.com/media/GD6No_5XkAAaqxm?format=jpg&name=medium [Jailbreaking the Simulation with George Hotz | SXSW 2019 - YouTube](https://youtu.be/ESXOAJRdcwQ?si=NqKDm7vjJwu8dWcR) [List of futurologists - Wikipedia](https://en.m.wikipedia.org/wiki/List_of_futurologists) [Category:Futurologists - Wikipedia](https://en.m.wikipedia.org/wiki/Category:Futurologists) [20 World Leading Futurist Speakers Trending in 2024](https://champions-speakers.co.uk/best-futurism-speakers-to-book) Improving activation steering mechanistic interpretability https://twitter.com/ojorgy/status/1738661946902098171?t=0zpbGuZluMzohCIj37mB3A&s=19 https://arxiv.org/abs/2311.13165 LLM plus graphs https://twitter.com/LangChainAI/status/1746959432112820401?t=5Im32ckE5Qy74LbNjsR8mg&s=19 ChatGPT 4 Mathematics of information processing theory in dynamical systems Information processing theory in the context of dynamical systems integrates concepts from mathematics, computer science, and physics to understand how information is processed over time within a system. Here's an overview of some key mathematical aspects: Dynamical Systems: A dynamical system is a system in which a function describes the time dependence of a point in a geometrical space. Mathematically, it's often represented by differential equations (continuous dynamical systems) or difference equations (discrete dynamical systems). State Space: The state space of a dynamical system is the set of all possible states in which the system can exist. Mathematically, it's often represented as a phase space with coordinates representing each state variable. Trajectories and Orbits: In the state space, the evolution of the system is represented by trajectories (continuous systems) or orbits (discrete systems). These are solutions to the system's equations, showing how the state evolves over time. Attractors and Stability: In many systems, states tend to evolve towards certain sets called attractors (like fixed points, limit cycles, or strange attractors). Stability analysis involves understanding how the system responds to small perturbations around these attractors. Information Theory: In the context of dynamical systems, information theory often deals with how information is stored, transmitted, and transformed within the system. Key concepts include entropy (a measure of uncertainty or disorder) and mutual information (a measure of the amount of information that one variable contains about another). Chaos Theory: In chaotic systems, small changes in initial conditions can lead to vastly different outcomes, making long-term prediction difficult. This sensitivity is quantified by measures such as the Lyapunov exponent. Network Theory: Many dynamical systems can be thought of as networks, where nodes represent components of the system and edges represent interactions. The structure and dynamics of these networks are crucial for understanding how information flows through the system. Nonlinear Dynamics: Most interesting dynamical systems are nonlinear, meaning that the equations governing their behavior are nonlinear. Nonlinearity leads to phenomena like bifurcations, where a small change in a parameter can cause a sudden qualitative change in the system's behavior. Computational Methods: Numerical simulations are often necessary to study complex dynamical systems. Techniques like finite element analysis, bifurcation analysis, and numerical integration are used to approximate the behavior of these systems. Applications: This mathematical framework is applied in various fields like neuroscience (to model brain dynamics), ecology (to model population dynamics), economics (to model market dynamics), and many others. Since brains develop by starting in a symmetrical state and then specializing using symmetry breaking, can we do the same [An experimental cancer drug had a 100% success rate : NPR](https://www.npr.org/2022/06/07/1103545361/cancer-drug-experimental-rectal-chemotherapy-surgery-treatment-immunotherapy) [New treatment approach effectively treats the most aggressive forms of prostate cancer](https://www.news-medical.net/news/20230919/New-treatment-approach-effectively-treats-the-most-aggressive-forms-of-prostate-cancer.aspx) [New Cancer Therapy In Final Trial After Initial 100% Success Rate](https://nocamels.com/2023/05/new-cancer-radiation-therapy-kills-tumors-from-the-inside/) Democracy over time up long term down short term Poverty Stats pinker Bells inequality proving nonlocality in quantum mechanics assuming no superdeterminism [Jailbreaking the Simulation with George Hotz | SXSW 2019 - YouTube](https://youtu.be/ESXOAJRdcwQ?si=yPUz0KolROuZ5k9H) Mindtechnologies = Psychotechnologies + Neurotechnologies [Object personification in autism: This paper will be very sad if you don't read it - PubMed](https://pubmed.ncbi.nlm.nih.gov/30101594/) My set of philosophical assumptions about reality are optimized purely for mind engineering purposes using both psychological mental tools and physical neurotechnologies. That's what my and QRI's set of assumptions are for, for engineering wellbeing, intelligence, motivation, freedom,... Philosophical assumptions are in this sense instrumental for more effective engineering. This is what this discord server is about, this is its goal. If you fight against it here, you will get logically backlash. If you have any other set of assumptions where you can justify why its better **for engineering**, tell me, with reasoning, why it opens and gives more predictive set of models, i'm open to it. So far I have seen just the opposite. Disconnecting the mind dynamics from mathematical physics feels like an assumption that doesnt help at all in designing all existing and future neurotechnologies, as the existing ones already work with brain's physics to alter experience in positive ways. As we get better and more predictive correspondence between brain physics and mind dynamics, the better the psychotechnologies and neurotechnologies will get, and the more qualia we can engineer. [On Connectome and Geometric Eigenmodes of Brain Activity: The Eigenbasis of the Mind?](https://qri.org/blog/eigenbasis-of-the-mind) [Welcome to the Cyborg Era: Brain Implants Transformed Lives This Year](https://singularityhub.com/2023/12/29/welcome-to-the-cyborg-era-brain-implants-transformed-lives-this-year/) https://www.lesswrong.com/posts/PkqGxkm8XRASJ35bF/the-case-for-training-frontier-ais-on-sumerian-only-corpus-1 Maybe quantum machine learning on neuromorphic photonic quantum computers with room tempature superconductors might be a possibility soon. Sounds hypercool, but i still wonder if there are really any practical advantages waiting, maybe such combination would be an overkill that wouldnt give add as much practical advantages in relation to other technology and weaker combinations, but we'll see, maybe not. One could also more closely mimic brain with spiking neural networks and more inherent nonlinerities and specialization of cells and electrochemical neurotransmitters, top down electric fields, different functional brain areas etc., which may or may not be needed for upgrading or merging organisms with artificially constructed systems. Maybe if we combine everything, a shoggoth called "quantum machine learning (https://scholar.google.com/citations?hl=en&user=NiXejNwAAAAJ&view_op=list_works&sortby=pubdate , https://en.wikipedia.org/wiki/Quantum_machine_learning, [Quantum-probabilistic Generative Models and Variational Quantum Thermalization - Guillaume Verdon - YouTube](https://www.youtube.com/watch?v=sS5ovtbXDGQ)) on neuromorphic (https://en.m.wikipedia.org/wiki/Neuromorphic_engineering, [New computer will mimic human brain -- and I'm kinda scared - YouTube](https://www.youtube.com/watch?v=8-WZiLOxoC8)) photonic quantum computers ([Nanomaterials | Free Full-Text | Neuromorphic Photonics Circuits: Contemporary Review](https://www.mdpi.com/2079-4991/13/24/3139) , [First Photonic Quantum Computer on the Cloud - IEEE Spectrum](https://spectrum.ieee.org/photonic-quantum) , [Room-temperature photonic logical qubits via second-order nonlinearities | Nature Communications](https://www.nature.com/articles/s41467-020-20417-4) ) with room tempature superconductors (https://olympiaacademia.medium.com/room-temperature-superconductors-unveiling-new-possibilities-for-quantum-computing-157750a745ee , https://news.ycombinator.com/item?id=38853706)" might be a possibility soon. Sounds hypercool, but i still wonder if there are really any practical advantages waiting, maybe such combination would be an overkill that wouldnt give add as much practical advantages in relation to other technology ([It looks like AI will kill Quantum Computing - YouTube](https://www.youtube.com/watch?v=Q8A4wEohqT0) ,[Revealing the Truth About Quantum Computing With Scott Aaronson - YouTube](https://www.youtube.com/watch?v=WLRE5OF5VF4)) and weaker combinations, but we'll see, maybe not. One could also more closely mimic brain with spiking neural networks and others (https://en.m.wikipedia.org/wiki/Spiking_neural_network), alternatives to backpropagation (https://www.science.org/doi/full/10.1126/science.adi8474?af=R&mi=0&target=default), more inherent nonlinerities (https://openreview.net/forum?id=0z_cXcu1N6o) and specialization of cells and electrochemical neurotransmitters, top down electric fields ([Cytoelectric coupling: Electric fields sculpt neural activity and "tune" the brain's infrastructure - PubMed](https://pubmed.ncbi.nlm.nih.gov/37210066/,) https://qri.org/blog/eigenbasis-of-the-mind), different functional brain areas etc., which may or may not be needed for upgrading or merging organisms with artificially constructed systems. ([Welcome to the Cyborg Era: Brain Implants Transformed Lives This Year](https://singularityhub.com/2023/12/29/welcome-to-the-cyborg-era-brain-implants-transformed-lives-this-year/)) I really really wonder what kind of weird combinations of all this tech will exist in the future. I still wonder if quantum machine learning has a future, as there are some researches saying its the ultimate future, and some saying its dead end, same with trying to mimic the brain more. Same with neuromorphic hardware and more close to the brain architectures, but i guess here it depends if we want to maximize human like hardware and intelligence or not, I wonder what OpenAI is cooking here. 👁 Or if photonic computing is really more fast for ML, but some studies say hell ye. Or if alternatives to backpropagation are good direction. [New Groundbreaking Research, Anthrobots, Hyper-Embryos | Michael Levin - YouTube](https://www.youtube.com/watch?v=hG6GIzNM0aM) https://www.lesswrong.com/posts/PkqGxkm8XRASJ35bF/the-case-for-training-frontier-ais-on-sumerian-only-corpus-1