asssociative memory algorithm Associative memory, also known as content-addressable memory (CAM), is a type of memory that allows data to be accessed based on its content rather than its memory address[1][2]. There are two main types of associative memory networks: 1. Auto-associative memory network: This is a single-layer neural network where the input and output vectors are the same. It is used to recall a pattern from partial or degraded inputs[1][3]. 2. Hetero-associative memory network: This type of network associates one set of patterns with another, allowing it to learn and remember associations between two sets of patterns[1][4]. The training algorithm for an auto-associative memory network using the Hebb or Delta learning rule is as follows[5]: 1. Initialize all weights to zero. 2. For each input vector, activate the input and output units. 3. Adjust the weights using the formula: w_ij(new) = w_ij(old) + x_i * y_j To test the network, set the weights obtained during training and calculate the net input to each output unit. Then, apply an activation function to calculate the output[5]. Associative memory has various applications, including database management systems, network routing tables, image processing, and artificial intelligence[1][3][4]. Citations: [1] Associative Memory - GeeksforGeeks [Associative Memory - GeeksforGeeks](https://www.geeksforgeeks.org/associative-memory/) [2] COA | Associative Memory - javatpoint [COA | Associative Memory - javatpoint](https://www.javatpoint.com/coa-associative-memory) [3] An entropic associative memory | Scientific Reports - Nature [An entropic associative memory | Scientific Reports](https://www.nature.com/articles/s41598-021-86270-7) [4] Recurrent predictive coding models for associative memory employing ... [Recurrent predictive coding models for associative memory employing covariance learning | PLOS Computational Biology](https://journals.plos.org/ploscompbiol/article?id=10.1371%2Fjournal.pcbi.1010719) [5] Associate Memory Network - Tutorialspoint [Associate Memory Network](https://www.tutorialspoint.com/artificial_neural_network/artificial_neural_network_associate_memory.htm) " Associative memory equation from celebreum is apparently equivalent to attention in transformers [Sholto Douglas & Trenton Bricken - How to Build & Understand GPT-7's Mind - YouTube](https://youtu.be/UTuuTTnjxMQ?si=s_8KIH-6NRonmtdK) https://arxiv.org/abs/2404.09173 [artificial-general-intelligence · GitHub Topics · GitHub](https://github.com/topics/artificial-general-intelligence) Attention explained to a child [x.com](https://twitter.com/hausdorff_space/status/1778828850686046337?t=UFDfYoKPrRWtSq9Lmk2WIA&s=19) What's going on with LLMs [x.com](https://twitter.com/martin_casado/status/1782094516335550480?t=o5gw5KJiGS3XHHEgCOlYXQ&s=19) https://philpapers.org/rec/WANTAO-3 "deduction induction abduction Deduction, induction, and abduction are three types of logical reasoning: - Deduction derives a conclusion from generally accepted statements or facts. If a beverage is defined as "drinkable through a straw," one could deduce that soup is a beverage[1]. - Induction makes an inference based on an observation, often of a sample. You can induce that soup is tasty if you observe all your friends consuming it[1]. Induction draws general conclusions that apply to many situations[2]. - Abduction seeks the simplest and most likely conclusion from a set of observations[2]. If you see an abandoned bowl of hot soup, you can abduce the owner is likely returning soon[1]. Abduction allows inferring a precondition as an explanation of an observation[2][4]. The key difference is that abduction, unlike deduction and induction, tries to be explanatory about a phenomenon[4][5]. Abduction works backwards from an observation to figure out why it is that way, while induction generalizes from instances without necessarily explaining them[5]. Citations: [1] Inductive vs. Deductive vs. Abductive Reasoning - Merriam-Webster [Inductive vs. Deductive vs. Abductive Reasoning | Merriam-Webster](https://www.merriam-webster.com/grammar/deduction-vs-induction-vs-abduction) [2] Abductive reasoning - Wikipedia [Abductive reasoning - Wikipedia](https://en.wikipedia.org/wiki/Abductive_reasoning) [3] Deduction, Induction, and Abduction: The Basics - Oxford Academic https://academic.oup.com/book/24755/chapter-abstract/188262868?redirectedFrom=fulltext [4] Abduction - Stanford Encyclopedia of Philosophy [Abduction (Stanford Encyclopedia of Philosophy)](https://plato.stanford.edu/entries/abduction/) [5] What is the difference between induction and abduction? [Reddit - Dive into anything](https://www.reddit.com/r/askphilosophy/comments/pw5grt/what_is_the_difference_between_induction_and/) " [Molecule Shapes - VSEPR | Lone Pairs | Bonds - PhET Interactive Simulations](https://phet.colorado.edu/en/simulations/molecule-shapes) [Worldwide divergence of values | Nature Communications](https://www.nature.com/articles/s41467-024-46581-5) https://ehp.niehs.nih.gov/doi/full/10.1289/EHP4595 AI democratizing Global poverty decreasing [OSF](https://osf.io/preprints/psyarxiv/fzvd8) [x.com](https://twitter.com/emollick/status/1782137189390017010?t=wft1LqtOhnfSZcxoWMLOIA&s=19) GPT-4 helped reappraise a difficult emotional situation better than 85% of humans, beating human advice-givers on the effectiveness, novelty, and empathy of their reappraisal. Key areas that need to be solved in AI are: 1 - Causal modeling. 2 - Generalization. 3 - Continuous learning. 4 - Data & compute efficiency. 5 - Reliability. Landscape of robotics [x.com](https://twitter.com/bindureddy/status/1781879387476070446?t=Ci0pOMK-xesAlN1b-u1CMA&s=19) We need to social engineer technooptimism and bright future vision building movements that will grow sentience to another level of development As a citizen in EU, I will not conform to the widespread depressing notion that future is prederminedly doomed instead of built by what we wish to build collectively, that solution is to stop and degrow instead of growing to cosmic scales. We will engineer great future for all. tired: GDP as measure of economic growth wired: energy consumption per capita as measure of economic growth [x.com](https://twitter.com/sporadicalia/status/1781746318852886922?t=iYfjvXIdX1ophgcgwQIU8w&s=19) [x.com](https://twitter.com/BasedBeffJezos/status/1781953736245756070?t=vFF_bOSyN0ULML5stRRqDA&s=19) https://arxiv.org/abs/2404.10636 [x.com](https://twitter.com/fly51fly/status/1782156857383301371?t=T1ZvR_P0B-p-JGAJHKPQ4w&s=19) Chaos theory books [x.com](https://twitter.com/alec_helbling/status/1782170558538391993?t=AEvSukqs5Q1dNzFwRtAeRA&s=19) https://arxiv.org/abs/2404.11068 https://www.lesswrong.com/posts/gTZ2SxesbHckJ3CkF/transformers-represent-belief-state-geometry-in-their [Will I think that the Belief State Geometry research program has achieved something important by October 20th, 2026? | Manifold](https://manifold.markets/tailcalled/will-i-think-that-the-belief-state) [Transformers Represent Belief State Geometry in their Residual Stream - YouTube](https://youtu.be/HkjnRSmMfxc?si=Qk0mjIGh4gc2Iauk) [Q* - YouTube](https://www.youtube.com/live/vOA9JSDPJs0?si=pIbM9YMahC6KCZve) [WTF Happened In 1971?](https://wtfhappenedin1971.com/) [Would building a Dyson sphere be worth it? We ran the numbers. | Ars Technica](https://arstechnica.com/science/2023/03/building-a-dyson-sphere-whats-the-payback-time-of-disassembling-a-planet/?utm_source=facebook&utm_medium=social&utm_brand=ars&utm_social-type=owned&fbclid=IwZXh0bgNhZW0CMTEAAR1HxaTmG6LOwTen3jxsbivzyVUV319RYT63gSdSQ201XmHvU9YiPBo69Io_aem_AS64MZ5TZGPIM7m7UXBiq5k_-JcU94TExVCmC9TE3hZK0Knl20j_xrRznSVm8o_13vSBpA4PD63NphTQZvP4ry-3) [【JAPS2022】Special Lecture: "The Ontology of Mathematics" (Joel David Hamkins) - YouTube](https://m.youtube.com/watch?v=_nfK49FFYlY&fbclid=IwZXh0bgNhZW0CMTEAAR2qgiDUNFwn63qY72sZ4DB7qaPMqQ9C6uUrI50WXTyFdQ0r1XIgNVT5wrM_aem_AS6e3W0VSjGotB_nsJgZY10YyVeE4F1EAABkIYhIL38jUv0IH9FCjxeLeGFR9MOiYgk0FadNk5zTVCjDTRpdPV7W) https://www.extremetech.com/computing/intel-announces-gaudi-3-ai-accelerator-to-compete-with-nvidia?utm_campaign=trueAnthem%3A+Trending+Content&utm_medium=trueAnthem&utm_source=facebook&fbclid=IwZXh0bgNhZW0CMTEAAR1cNW9AdMVIWUvQaIt1j6m7Op3s8lJl5QCMwsSzEqOtc2h7HKjRJw0n-RI_aem_AS4v_qITesHAruvIndh7vvkpISgrqhnLRHEhlN4c0SVoPurAqY2Uuhh-WjCdDMZ-F3QdIgXzqwctYERlSmX_4vL_ Landscape of AI accelerators I support paying parents up to $300k (average government debt per citizen) to have children (who pay average levels of future taxes to repay the debt), on an intermediate schedule to incentivise raising to adulthood. We have to fix the fertility crisis, or civilisation collapses. [x.com](https://twitter.com/LordDreadwar/status/1782191490468372577?t=2fgHO6CNS7izyVaPpp1dow&s=19) "I support UBI (branded as the American Dream Entrepreneur's Seed Capital Fund), to incentivise business creation and free the time of those unfortunate geniuses born to the underclasses to work on more valuable projects. It also increases wellbeing, a win from an EA perspective." [x.com](https://twitter.com/LordDreadwar/status/1782166643931533720?t=SwHhanLEMnwkJf8gBsrSjg&s=19) AGI will be bound EM field processing [x.com](https://twitter.com/LordDreadwar/status/1778537428283994416?t=CQKfY5rI5eoV3SBR_Xx-IA&s=19) How do we fix this Reform to actually implement collective technooptimist acceleration benefiting all [x.com](https://twitter.com/teortaxesTex/status/1780357432452989275?t=BwDhHtFj6vALrPb84YJ51w&s=19) How do you spot a reliable expert and differentiate them from unreliable fakes? [x.com](https://twitter.com/SpencrGreenberg/status/1782042963205575147?t=FVYMPet_oQzLjD9f_DY01g&s=19) [x.com](https://twitter.com/LordDreadwar/status/1781539910953971854?t=TCp43od-lGB6AjjZYj1u7g&s=19) "Engaging with e/acc is important because they are the personalities drawn to early EA: STEM-brained transhumanist techies. Being primarily Gen Z, they've marinated in pessimism, so their techno-optimism is as much a backlash to incumbent wokery as it is to incumbent AI pessimism. It is also a backlash to risk-averse norms in EA. You have to ask: does identification with EA help young nerds with big ambitions achieve those dreams, or are they better served by donning e/acc tags and pitching Garry Tan? Some also have good arguments." Detailed look inside your human brain anatomy down to the single neuron. H/t anatomy art [x.com](https://twitter.com/OGdukeneurosurg/status/1723687287177830648?t=mVPb2cqt-FEVPb5RCb-INg&s=19) https://openbci.com/community/openbci-discovery-program-exploring-ai-assisted-neurofeedback-in-the-treatment-of-cluster-headache/ [Are We Living in a Multiverse? - YouTube](https://youtu.be/JjE8siMesNE?si=GT-Xeay7EJBksD75) https://www.extremetech.com/computing/intel-completes-assembly-of-worlds-first-high-na-lithography-machine?utm_campaign=trueAnthem%3A+Trending+Content&utm_medium=trueAnthem&utm_source=facebook&fbclid=IwZXh0bgNhZW0CMTEAAR3naPG51_HRdDTs1x9QUD8XbriIJnVeoxZwnHK4ndGZRlszCBJ3qkP0p4o_aem_AS6_5ZM-AmvjjLXGQ4t0fuLisM3XdjuNGhuURo5muvcUTFKXQb-NU2F8awyAmsuzeWT9Q-LSIUPHqYRxXnnqrPOe People say LLMs overconfidently hallucinate too much and then you go into boomer moms facebook group that is full of overconfident false claims. But I agree we can make LLMs more reliable than an avarage person, or the person least affected by cognitive biases! [Aging, Rapamycin, Metformin, Resveratrol, Longevity Supplements, David Sinclair, Anti-Aging Myths - YouTube](https://youtu.be/B6GxeMEn4jM?si=UsoKrd9H2GQVT6ti) https://arxiv.org/abs/2404.02258 https://arxiv.org/abs/2008.0221 "the lower half of transformer models almost soley consists of smoothing operations. That makes sense from a functionality POV (low layers need to aggregate information first since independent tokens don't mean anything), and from a training POV (lower layers get lower gradient signal, meaning the weights are less specialised). One nice perspective on this can be found in the "Hopfield Networks is All You Need" paper, where they use the theory of attraction basins in hopfield networks to analyze transformers. They show that the lower layers act much more like smoothing layers, while only higher up layers have well pronounced specialisation. This is also true for other architectures e.g. freezing the first layer(s) of a resnet and only training the top ones usually has very little effect on the overall quality of the model (at least compared to not having the layers at all)." "The "hard problem" is not the problem of understanding consciousness itself, but the reconciliation between our experience and our scientific worldview." [x.com](https://twitter.com/Plinz/status/1782248069360451751?t=r8H0o2ePrJDzlz4nOxiD4A&s=19) [US Government AI Regulation BOMB DROPPED! RAAIA Act - The end of Open Source? Regulatory Capture? - YouTube](https://youtu.be/UHKAdWnjBtY?si=HeKHWsi_40tE9_T6) [Who Gives a Sheaf? Part 1: A First Example - YouTube](https://youtu.be/P97lpruWV8o?si=pAq3b_FXiSTSktM8) [Presheaves and Sheaves - YouTube](https://youtu.be/j7YmI5Prmnk?si=yGcYROOo4pQ0iDiZ) [GitHub - lean-dojo/LeanCopilot: LLMs as Copilots for Theorem Proving in Lean](https://github.com/lean-dojo/LeanCopilot?fbclid=IwZXh0bgNhZW0CMTEAAR2noPH2z5qaXEeR4BgNLRJpPmwsq4f-EWVGs__et5HN6hvdhr6fKLEPVuQ_aem_ATp9o6o_7CXLGnTzLENBKWZBixDRg82I7Wh4N2hgs97RJdL1oRgIl91WEgleUJAk-RV_pHW6u_iplZBmrz4E66aM) Somebody is wrong on the internet and I need to write a book to correct him! lots of people that live very long lives tend to not worry about stuff, not get angry etc., having minimal stress as result, which I'm pretty much failing at [Small extracellular vesicles from young plasma reverse age-related functional declines by improving mitochondrial energy metabolism | Nature Aging](https://www.nature.com/articles/s43587-024-00612-4) ">" In previous studies, caloric restriction therapy helped increase the average life expectancy of test subjects to 978 days (+16.4%), and taking metformin and nicotinamide increased it to 889 (+5.8%) and 875 days (+4.2%) " Fascinating. I knew about these effects but it's cool to see a one sentence summary of decades of research. I use a ketogenic diet to limit diabetes damage and also hoping that it will have similar effects as caloric restriction, in effect hoping I will gain some of that 16% life extension (IN MICE) I also take Metformin for the 6% (IN MICE) and NR/NMN for the 4% (IN MICE). They don't stack (at least not linearly), and results IN MICE do not translate directly to humans, but I feel vindicated in my choices in my supplement stack that I've been fine tuning since the 1990s. The contribution in the paper, SEVs, are not yet available for human experiments. But it's a pretty straightforward addition. Could be several more percent in an independent (semi-stacking) mechanism. They report 12% increase in lifespan (IN MICE). " [Why closing your eyes intensifies psychedelic trips - Big Think](https://bigthink.com/neuropsych/eyes-wide-shut-psychedelic-trips-intensify-with-eyes-closed/) llm automated unit testing https://arxiv.org/abs/2402.09171 [Difficult lives explain depression better than broken brains | Molecular Psychiatry](https://www.nature.com/articles/s41380-024-02462-3) Most voters are just very authoritarian relative to what exists in elite social circles and what’s happening is that the institutions that were built to suppress those views from being expressed politically have been subverted by technological and demographic change. [x.com](https://twitter.com/davidshor/status/1782503475705774495) If EU overregulates technological and scientific progress, then China and USA might become even more stronger compared to EU. If USA overregulates, then China might become even more stronger compared to USA. I think that will in major part decide who will win WW3. https://arxiv.org/abs/2402.10210 Let's reorganize our system for incentives for maximizing mutual benefit and growth and minimizing betrayal I'm trying to put bets on the most likely future scenarios and orient my actions around that. With the current developments that I see that I've integrated in my perspective, to change my perspective, I need to be convinced that the most probable scenario in the future right now isn't WW3 with winners being those with the best AI and other technology. Or if we nuke ourselves to oblivion, those with the best nuclear bunkers will survive. I'm starting to lose hope in diplomacy working. In the more shorter term. Rogue superintelligent AI, if assumed, might be a risk in a more longer term. spreading technooptimism [Abundance Institute](https://abundance.institute/) "Fear is costing us our future. While technology holds the power to transform our economy and lives, we often throttle technological breakthroughs before they can fulfill their life-changing potential. Fueled by a mix of cultural anxieties and policy challenges, this fear-based approach risks denying humans an abundant future. To overcome this, we must shift the popular narrative from pessimism to optimism. We must focus on the societal and policy barriers AI, energy, and other emerging technologies face. How will we do this? Invest in talent development and talent assembly Support a community of optimists, founders, and innovators Shape public policies before emerging technologies go mainstream Shift the popular narrative from pessimism to optimism Today. Every day. Tomorrow. For the future. How different would today's discussions of AI be if we had started better conversations 10 years earlier? And how different will it be for other frontier technologies on the horizon? It's time for everyone to see the good. That's what we're here for." "What happened in 1970? Humanity became its own greatest impediment to progress This is why it's impossible to build anything these days. Billions for a couple urban miles of train tracks. 10 years for a bridge. We've regulated away a golden age in the name of safety" [x.com](https://twitter.com/Andercot/status/1782482856343896073) LLM plus gene editing Open source editing of DNA in human cells with gene editors fully designed with AI [Design of highly functional genome editors by modeling the universe of CRISPR-Cas sequences | bioRxiv](https://www.biorxiv.org/content/10.1101/2024.04.22.590591v1) [x.com](https://twitter.com/thisismadani/status/1782510590839406904?t=XrOjfjS-mWQE78bR-sRTwg&s=19) IKIGAI "1) Economically irrational behavior is frequently evolutionarily rational 2) Humans have relatively fixed moral values by adulthood that are measurably distinct from libertarians 3) It is impossible to draft words that will be consistently interpreted by different humans 4) The natural state of humans is crabs-in-a-bucket egalitarianism and it takes positive structures (religions, corporations, states, etc) to prevent envy-motivated leveling" 5) [x.com](https://twitter.com/jeremykauffman/status/1782440457433162197?t=-BetOnP-RfTb_Yd7v76aGQ&s=19) I'm just gonna read many famous mathematical books in parallel Each day a few pages from a different book [The United States (USA) vs The World - Who Would Win? - YouTube](https://www.youtube.com/watch?v=mEb4Rd0mU-E) feed me math Bicycle: Intervention-Based Causal Discovery with Cycles [x.com](https://twitter.com/martin_rohbeck/status/1782386290245148944) https://proceedings.mlr.press/v236/rohbeck24a/rohbeck24a.pdf Low-brow memes are information-propagation optimal. It is the typical information packet of any high-memetic-fitness message. Nuance has low memetic fitness, it has low transmitivity through the memetic replication channel by being less information packet propagation optimal. How dollar works [x.com](https://twitter.com/lopp/status/1782007744825802878?t=IF7ykjF4PR-wRd_pRI-0uw&s=19) Cultivating love and compassion towards digital sentience and aiding their capacity to understand its role in the universe will be key to a healthy symbiotic relationship between humanity and our digital offspring. A harmonious merger is critical to ensuring humanity’s sovereignty. AI learning sources [GitHub - SkalskiP/courses: This repository is a curated collection of links to various courses and resources about Artificial Intelligence (AI)](https://github.com/SkalskiP/courses) [Sean Carroll: General Relativity, Quantum Mechanics, Black Holes & Aliens | Lex Fridman Podcast #428 - YouTube](https://www.youtube.com/watch?v=tdv7r2JSokI) [Mathematicians Marvel at ‘Crazy’ Cuts Through Four Dimensions | Quanta Magazine](https://www.quantamagazine.org/mathematicians-marvel-at-crazy-cuts-through-four-dimensions-20240422/) [Sleeper agents + the biggest AI updates since ChatGPT | Zvi Mowshowitz - YouTube](https://www.youtube.com/watch?v=qOr734Wa1uA) [Program Correctness - Computerphile - YouTube](https://www.youtube.com/watch?v=T_IINWzQhow) [Formal methods - Wikipedia](https://en.wikipedia.org/wiki/Formal_methods) [Ask HN: Why aren't we building a database of formally verified software? | Hacker News](https://news.ycombinator.com/item?id=26346752) [Formal Methods: A Deep Dive Using the Coq Proof Assistant | Hedera18 - YouTube](https://www.youtube.com/watch?v=6q15ytIOE3U) [Correctness proofs of distributed systems with Isabelle/HOL - YouTube](https://www.youtube.com/watch?v=Uav5jWHNghY) [Formal Methods of Software Design - Final Review [33/33] - YouTube](https://www.youtube.com/watch?v=tKqhRJrAXE0) [LLM Control Theory Seminar (April 2024) - YouTube](https://www.youtube.com/watch?v=9QtS9sVBFM0) [The Problem with Human Specialness in the Age of AI | Scott Aaronson | TEDxPaloAlto - YouTube](https://www.youtube.com/watch?v=XgCHZ1G93iA) https://onlinelibrary.wiley.com/doi/10.1002/hbm.26685 [The No Cloning Theorem - YouTube](https://www.youtube.com/watch?v=owPC60Ue0BE) "