Superalignment won't happen in a vacuum without building the technology, it's fundamentally an empirical project. We will eventually need ASI to survive the longest in universe and prevent cosmic sized catastrophic risks and possibly eventually somehow heat death of the universe. [G. Stolyarov II - YouTube](https://youtube.com/@gstolyarovii?si=2uoTcf5ycX0TfLFi) [[2406.00877] Evidence of Learned Look-Ahead in a Chess-Playing Neural Network](https://arxiv.org/abs/2406.00877) mechanistic interpretability https://x.com/jenner_erik/status/1798018379560943825?t=stFzjDRmybQFd_GHVPlxMA&s=19 [Supersymmetric theory of stochastic dynamics - Wikipedia](https://en.wikipedia.org/wiki/Supersymmetric_theory_of_stochastic_dynamics) “We mathematically showed that there might be a way to see beyond our universe.” – Eric Ling, University of Copenhagen [Mathematicians Attempt to Glimpse Past the Big Bang | Quanta Magazine](https://www.quantamagazine.org/mathematicians-attempt-to-glimpse-past-the-big-bang-20240531/) Space of all possible physically realizable or nonrealizable systems Space of all possible formal systems [[2402.14740] Back to Basics: Revisiting REINFORCE Style Optimization for Learning from Human Feedback in LLMs](https://arxiv.org/abs/2402.14740) What do you think will be the greatest constraint to AI advancement and why? Data? Compute? Energy? Algorithmic/scientific breakthroughs? Regulation/politics? Culture? Does it bother you that unelected billionaires have so much power and influence? [The uneasy relationship between deep learning and (classical) statistics – Windows On Theory](https://windowsontheory.org/2022/06/20/the-uneasy-relationship-between-deep-learning-and-classical-statistics/) [Geoffrey Guy (2024) Quantum Biology and the Future of Medicine - YouTube](https://youtu.be/7c-3cNYpJBk?si=ZXQYaYDlptyQl_jE) [Unconditional Acceptance of Just How You Are Right Now - YouTube](https://www.youtube.com/live/rei4y7JkrKI?si=zjItMVBPV_PL9ZSh) [Introduction - SITUATIONAL AWARENESS: The Decade Ahead](https://situational-awareness.ai/) http://www.planetparasol.ai/ "Global Cooling Forecasts from Stratospheric Aerosol Injection Stratospheric aerosol injection (SAI) uses reflective aerosols released into the upper atmosphere to reflect sunlight and thereby cool Earth's surface." Climate change is an engineering problem We can ascend the Kardashev scale without climate change [Heart-Risk Model Saves Lives, Self-Driving on Unruly Roads, and more](https://www.deeplearning.ai/the-batch/issue-251/) "A barrier to faster progress in generative AI is evaluations (evals), particularly of custom AI applications that generate free-form text. Let’s say you have a multi-agent research system that includes a researcher agent and a writer agent. Would adding a fact-checking agent improve the results? If we can’t efficiently evaluate the impact of such changes, it’s hard to know which changes to keep." [Cradle: Empowering Foundation Agents Towards General Computer Control](https://baai-agents.github.io/Cradle/) [[2403.03186] Cradle: Empowering Foundation Agents Towards General Computer Control](https://arxiv.org/abs/2403.03186) >To our best knowledge, our work is the first to enable LMM-based agents to follow the main storyline and finish real missions in complex AAA games, with minimal reliance on prior knowledge and application-specific resources. [A Right to Warn about Advanced Artificial Intelligence](https://righttowarn.ai/) https://x.com/DKokotajlo67142/status/1797994238468407380?t=f9250NXatbijNpvC6LF5WQ&s=19 https://x.com/AndrewCurran_/status/1798008084444721278?t=MdfcaYcFkt_kztDhT_N87w&s=19 [[2406.00153] $μ$LO: Compute-Efficient Meta-Generalization of Learned Optimizers](https://arxiv.org/abs/2406.00153) [Mind-bending new programming language for GPUs just dropped... - YouTube](https://youtu.be/HCOQmKTFzYY?si=QiWz7gqDIijw1v4c) [[2405.20550] Uncertainty Quantification for Deep Learning](https://arxiv.org/abs/2405.20550) [Representationalism and Rationality: Why Mental Representation is Real - PhilSci-Archive](https://philsci-archive.pitt.edu/23515/?utm_source=dlvr.it&utm_medium=twitter) [[2306.11922] No Wrong Turns: The Simple Geometry Of Neural Networks Optimization Paths](https://arxiv.org/abs/2306.11922) https://www.sciencedirect.com/science/article/pii/S0149763424000927 https://psycnet.apa.org/record/2010-22538-001 [How are memories stored in neural networks? | The Hopfield Network #SoME2 - YouTube](https://youtu.be/piF6D6CQxUw?si=lbSj3FI1qaH4Y94X) https://medicalxpress.com/news/2024-05-psychedelic-drug-hyperconnectivity-brain-subjective.html [Is It Time to Redefine the Singularity? | Psychology Today Singapore](https://www.psychologytoday.com/sg/blog/the-digital-self/202405/is-it-time-to-redefine-the-singularity?fbclid=IwZXh0bgNhZW0CMTEAAR3N82hr9ZmpXqcDHtQT08RkuypOOKFLrrpnng0nTzZV5_MieU3XA0t8k68_aem_ZmFrZWR1bW15MTZieXRlcw) [Locality in Quantum Physics Explained | Quantum Foundations Podcast Ep. 1 ft. Dr Nicetu Tibau Vidal - YouTube](https://youtu.be/cDwvrP3Tvcw?si=3_bYE8sOAk1Shmxy) [Beyond Token Prediction: the post-Pretraining journey of modern LLMs - AI, software, tech, and people. Not in that order. By X](https://amatria.in/blog/postpretraining) [Brain representation in conscious and unconscious vision | bioRxiv](https://www.biorxiv.org/content/10.1101/2024.05.27.596053v1) [Beyond the serotonin deficit hypothesis: communicating a neuroplasticity framework of major depressive disorder | Molecular Psychiatry](https://www.nature.com/articles/s41380-024-02625-2) [Physicists Puzzle Over Emergence of Strange Electron Aggregates | Quanta Magazine](https://www.quantamagazine.org/physicists-puzzle-over-emergence-of-strange-electron-aggregates-20240529/?fbclid=IwZXh0bgNhZW0CMTEAAR0f8-13L6i2f8IUW40K-9n1SVXiSlUzFQxW_YMN9gEhjZCEQU0l8ECYpfQ_aem_ZmFrZWR1bW15MTZieXRlcw) Causal power is all you need "Knowledge is a species of information, but very little information is knowledge. Because the vast majority of the information that exists in the universe does not have causal power. It doesn't have the power to replicate itself, basically. But it also doesn't have the power to systematically cause transformations in other systems. So, you know, you could take a census of all the grains of sand on all the coastlines in the world, and that would be a vast amount of information. And almost none of that information is involved in the explanation of anything else. So, whereas a much higher proportion of, let's say, the information in the books in the Bodleian Library, still, you know, not 100%, probably not 10%, but vastly more than in the grains of sand, have the power to cause things to happen. Historians can go in there, convinced of one thing, come out convinced of another thing, and then go and do things in the world that have much greater mass and momentum and generally causal effect than the grains of sand." https://x.com/DeutschExplains/status/1797676113524650055?t=IRHBretBxSZAH1IAsuQvAA&s=19 [Prof David Deutsch - Quantum Information in Many Worlds - YouTube](https://youtu.be/XZyLQr6kv3I?si=uaJhnYnw-db4PuN2) [[2405.15743] Sparse maximal update parameterization: A holistic approach to sparse training dynamics](https://arxiv.org/abs/2405.15743) https://x.com/CerebrasSystems/status/1796578819400442033?t=42wnSMzLpbMUgCiTJhM31g&s=19 [[2405.18418] Hierarchical World Models as Visual Whole-Body Humanoid Controllers](https://arxiv.org/abs/2405.18418) [Puppeteer](https://www.nicklashansen.com/rlpuppeteer/) https://x.com/arankomatsuzaki/status/1795639173548044755?t=vnRp_2GYGIqGyzjyLIdezw&s=19 "ok this is blowing my mind – apparently genes are a read-write system?? your genes edit themselves?? the majority of genetic variation is from this self-editing rather than from random mutations??" https://x.com/kasratweets/status/1793046130407657809?t=TbDhlJojgWwlAA-6NpcFjQ&s=19 [[2405.20974] SaySelf: Teaching LLMs to Express Confidence with Self-Reflective Rationales](https://arxiv.org/abs/2405.20974) [OSF](https://osf.io/preprints/psyarxiv/xfb48) Chilling out in nondual Jhana / cessation + orgasm + DMT + LSD + MDMA + 5-MeO-DMT https://x.com/burny_tech/status/1624819204464562177 Apparently new state of the art for automated AI software engineering [[2405.15793] SWE-agent: Agent-Computer Interfaces Enable Automated Software Engineering](https://arxiv.org/abs/2405.15793) [SWE-agent documentation](https://princeton-nlp.github.io/SWE-agent/) Mechanistic intepretability can give so much more steering power over outputs of mostly black box deep learning models. Imagine being able to tune every knob of the latent space and circuits. The Map of Topological Quantum Computing - a NEW Kind of Quantum Computer Topological quantum computing is a brand new form of quantum computing where topology helps to avoid noise by harnessing the power of Majorana quasiparticles which are made from an exotic form of superconductivity where the electrons behave like there is a Majorana particle there which has the special properties of topology. [The Radical Map of Topological Quantum Computing - YouTube](https://www.youtube.com/watch?v=ihZXl33t8So) I want to tattoo all the known mathematics used in formal and natural sciences to predict and engineer systems and how they are used and related to eachother into my brain [[2405.14780] Metric Flow Matching for Smooth Interpolations on the Data Manifold](https://arxiv.org/abs/2405.14780) [What is interpretability? - YouTube](https://www.youtube.com/watch?v=TxhhMTOTMDg) anthropic [Translation of neurotechnologies | Nature Reviews Bioengineering](https://www.nature.com/articles/s44222-024-00185-2) Open source or closed source, utopian or dystopian, post AGI future? At what coordinates on this compass do we land? https://x.com/burny_tech/status/1797823109615247669 [Breakthrough synapse-regenerating ALS pill moves to phase 2 human trials](https://newatlas.com/medical/als-regenerative-pill-clinical-trials/) [No physics? No problem. AI weather forecasting is already making huge strides. | Ars Technica](https://arstechnica.com/ai/2024/06/as-a-potentially-historic-hurricane-season-looms-can-ai-forecast-models-help/) https://direct.mit.edu/jocn/article-abstract/doi/10.1162/jocn_a_02194/121298/Single-pulse-Transcranial-magnetic-Stimulation?redirectedFrom=fulltext [[2405.20519] Diffusion On Syntax Trees For Program Synthesis](https://arxiv.org/abs/2405.20519) https://x.com/shreyaskapur/status/1797726079995826629 https://tree-diffusion.github.io/ [GitHub - revalo/tree-diffusion: Diffusion on syntax trees for program synthesis](https://github.com/revalo/tree-diffusion) [Exploring the Intersection of Artificial Intelligence and Neuroscience | Internet Networking AI](https://jpvasseur.me/ai-and-neuroscience) https://www.ncbi.nlm.nih.gov/pmc/articles/PMC10053494/ [[2405.20233] Grokfast: Accelerated Grokking by Amplifying Slow Gradients](https://arxiv.org/abs/2405.20233) https://www.frontiersin.org/research-topics/58717/the-intersection-of-ai-and-llms-in-neuroscience-and-human-research [Artificial brain - Wikipedia](https://en.wikipedia.org/wiki/Artificial_brain) originální neuronky jsou celkem zjednodušená aproximace biologickýho information processingu [Neural network (machine learning) - Wikipedia](https://en.wikipedia.org/wiki/Neural_network_(machine_learning)) ze školy connectionismu z kognitivních věd [Connectionism - Wikipedia](https://en.wikipedia.org/wiki/Connectionism) obor AI a obor výpočetních neurověd/neuropsychologie si jdou v podstatě vlastními kroky ale zároveň se hodně navzájem ovlivňují matika z neurověd se cpe do AI, a matika z AI se cpe do neurověd do toho ještě fyzika se cpe do obou je to zajímavý evoluční grupáč různý aplikovaný matiky z různých oborů do různých oborů 😄 Tohle je super souhrn jak se oba obory ovlivňují [Exploring the Intersection of Artificial Intelligence and Neuroscience | Internet Networking AI](https://jpvasseur.me/ai-and-neuroscience) Tady máš rychlej souhrn toho článku: The human brain is an incredibly complex, powerful, and energy-efficient device made up of billions of neurons and trillions of synapses organized into specialized networks. It is highly dynamic and plastic, capable of multi-modal learning and task generalization from small amounts of data. AI has made big progress, especially with large language models (LLMs) using transformer architectures, attention mechanisms, and other techniques. AI is helping advance neuroscience through dimensionality reduction, clustering, time series analysis, neural networks and other techniques applied to fMRI, EEG, MEG and other brain data. This is enabling advances in brain mapping, brain-computer interfaces, vision/audition understanding, psychiatry, drug development and surgery. Neuroscience is inspiring AI through neural network architectures, reinforcement learning algorithms, neuromorphic hardware, attention mechanisms, hierarchical memories, and more. Key areas where neuroscience may further inspire AI include: . Improving energy efficiency to get closer to the brain's 25W power consumption . Enabling more efficient learning and generalization from small datasets . Representing knowledge in world models to understand consequences of actions . Allowing continuous learning over time without catastrophic forgetting . Enhancing artificial neurons to have complexity closer to biological neurons . Improving memory systems with encoding, consolidation, storage & retrieval like the brain . Increasing resistance to noise and adversarial attacks through redundancy and error correction . Achieving plasticity to dynamically adapt network structure over time . Harnessing surprise, novelty, motivation and reward to drive learning A multidisciplinary approach integrating AI and neuroscience holds immense promise for advancing both fields and developing impactful new technologies in the future. Tohle je fajn video na intersekci AI a neurověd [Dendrites: Why Biological Neurons Are Deep Neural Networks - YouTube](https://www.youtube.com/watch?v=hmtQPrH-gC4) Tohle je fajn recent článek co se snaží tyto dva obory víc propojit 😄 [Catalyzing next-generation Artificial Intelligence through NeuroAI | Nature Communications](https://www.nature.com/articles/s41467-023-37180-x) Nedávno vyšla studie co víc identifikovala kvantovost v mozku, ale je asi spíš pravděpodobný, že s vědomím to nemá souvislost Ultraviolet Superradiance from Mega-Networks of Tryptophan in Biological Architectures: https://pubs.acs.org/doi/10.1021/acs.jpcb.3c07936 [Brain Really Uses Quantum Effects, New Study Finds - YouTube](https://youtu.be/R6G1D2UQ3gg?si=8vBeliiiMLyELpxf) Ale i tak by mě zajímalo by mě či by se to hodilo replikovat na implementační úrovni. Computational úroveň je možná víc significant pro prožitek a je asi jinej oříšek na vyšší úrovni abstrakce, a mezi těmito urovnema je ještě algoritmická úroveň. [Frontiers | Integrated world modeling theory expanded: Implications for the future of consciousness](https://www.frontiersin.org/articles/10.3389/fncom.2022.642397/full) Je technicky možný že mozek má nějakou hardwarovou kvantovost co využívá jako akcelerátor pro nějaký quantum algorithmy který by na tom byly o dost hůř na klasickým hardwaru. Co se týče umělých neuronek, tak existuje celej quantum machine learning obor 😄 [Quantum machine learning - Wikipedia](https://en.wikipedia.org/wiki/Quantum_machine_learning?wprov=sfla1) Ale myslím že budoucnost jsou spíš hybridy, kde kvantový algoritmy na kvantových počítačích jsou užitečný jako akcelerátor jen někde. "Many authors try to find quantum algorithms that can take the place of classical machine learning algorithms to solve a problem, and show how an improvement in terms of complexity can be gained. This is dominantly true for nearest neighbour, kernel and clustering methods in which expensive distance calculations are sped up by quantum computation." [[1409.3097] An introduction to quantum machine learning](https://arxiv.org/abs/1409.3097) Různých AI akcelerátorů, ať klasických digitálních procesorů podporůjící různý algorithmy (násobení matic) a architektury (např specializovaný hardware na Transformer), či kvantových procesorů, analogových procesorů (optický), stochastických (thermodynamický), nebo bio (organoidy) apod., implementující různý machine learning algoritmy víc využívající hardwarovou úrovneň jako akcelerátor pro vyšší výpočetní efektivitu začíná být víc a víc. Zajímá mě co všechno za další paradigmata a hybridy vznikne na hardwarový i softwarový úrovni. Typů machine learning algoritmů taky tvoří gigantický evoluční strom, a to co je teď nejpopulárnější, což je deep learning, učení modelů přes curve-fitting differentiable parametric curves, je jen pidi podprostor všech možných machine learning architektur. 😄 Various AI accelerators, whether classical digital processors supporting different algorithms (matrix multiplication) and architectures (e.g., specialized hardware for Transformers), quantum processors, analog processors (optical), stochastic (thermodynamic), or bio (organoids), etc., implementing different machine learning algorithms that make more use of the hardware level as accelerators for higher computational efficiency, are becoming more and more common. I am interested in what other paradigms and hybrids will emerge at both the hardware and software levels. The types of hardware and machine learning algorithms form a gigantic evolutionary tree, and what is currently the most popular, which is deep learning, training models through curve-fitting differentiable parametric curves on transistor-based digital computers with their current transistor-based Von Neumann architecture, GPUs with specialized circuits for matrix multiplication, TPUs, is just a tiny subspace of all possible machine learning architectures and hardware. [Frontiers | Theoretical considerations and supporting evidence for the primary role of source geometry on field potential amplitude and spatial extent](https://www.frontiersin.org/articles/10.3389/fncel.2023.1129097/full) "oh hell yeah! we are so close to immortality i can taste it. we have been able to cryopreserve eggs and sperm for a while now, but resuming electrical activity in neurons was the holy grail and the madlady did it! every day a step forward" https://x.com/IterIntellectus/status/1797756504843694279 https://x.com/LauraDeming/status/1797641862309937627 [Cradle](https://www.cradle.xyz/) [And this year's Turing Award goes to... - YouTube](https://www.youtube.com/watch?v=mZck0N_T9Cs) [Jonathan Gorard: the complete first interview - YouTube](https://www.youtube.com/watch?v=asCDGSYzwhw) [Active Inference LiveStream 057.1 ~ Active Data Selection and Information Seeking - YouTube](https://www.youtube.com/live/N0ozoxTgWD0) Alternative names for imaginary/complex numbers: rotation numbers, extend numbers, phase numbers, polar numbers, squarerootofnegativeonenbers, indeterminable numbers https://www.reddit.com/r/math/comments/1d866w0/imaginary_number_is_kinda_misnomer_how_would_you/ [CPU vs GPU vs TPU vs DPU vs QPU - YouTube](https://www.youtube.com/watch?v=r5NQecwZs1A) [Harnessing the energy of stars... in a bottle | Fuse - YouTube](https://www.youtube.com/watch?v=IcgfDbGmmVc) [[2406.00288] Neural Optimal Transport with Lagrangian Costs](https://arxiv.org/abs/2406.00288) [Machine Learning Specialization by Andrew Ng - YouTube](https://www.youtube.com/playlist?list=PLkDaE6sCZn6FNC6YRfRQc_FbeQrF8BwGI) [[2406.00592] Model Predictive Control and Reinforcement Learning: A Unified Framework Based on Dynamic Programming](https://arxiv.org/abs/2406.00592) https://math.ucr.edu/home/baez/Polymath.pdf [Life-mind continuity: untangling categorical, extensional, and systematic aspects](https://link.springer.com/epdf/10.1007/s11229-024-04645-5?sharing_token=4ZCXUbmoDIm9Pf-jFucDGfe4RwlQNchNByi7wbcMAY5dVpf7G0Rzovfip4irHrY2JMyCDxlIgdDLVHT1BSuqt_cGgychNLPLdWXWBitafBkNpVOlRRAX6cbTikX7cGA6hqdST6_rdTi7n8KiymgVNEoXYvTmCNI24FK7YYcvSwI%3D) Interesting how Gemini Flash and Gemini 1.5 Pro have taken prominent positions. Gemini Flash is by far the most cost-effective. The Gemini Pro has unseated Claude Opus and rivals the GPT-4o. Google has done an excellent job. https://x.com/artificialguybr/status/1797873068595155124 [Even the smartest AI likely won't be "alive." Here's why. - Big Think](https://bigthink.com/13-8/is-life-comptuable/) [Anthropic's New Mech-Interp Paper, A Deep Dive - YouTube](https://www.youtube.com/watch?v=y0ZXFl3rQlQ) LLM plus neurosymbolics [A New AI Discovery Sure Looks Like the Dawn of True Machine Reasoning](https://www.popularmechanics.com/technology/robots/a60806576/new-ai-discovery/?utm_campaign=trueanthemFBPOP&utm_medium=social&utm_source=facebook&fbclid=IwZXh0bgNhZW0CMTEAAR2DaHpnjb8O4ph6MF7HAB-sH-CtZVAxYP-9Ah9eAGohw9-_IL_SgkqCVgI_aem_ZmFrZWR1bW15MTZieXRlcw) [[2312.08566] Learning adaptive planning representations with natural language guidance](https://arxiv.org/abs/2312.08566) [Network state transitions during cortical development | Nature Reviews Neuroscience](https://www.nature.com/articles/s41583-024-00824-y) [Molecular dynamics data will be essential for the next generation of ML protein models](https://www.abhishaike.com/p/an-argument-for-integrating-molecular) https://x.com/owl_poster/status/1797608482700759041?t=FZmLLhxanXkrUYBmJ7NDAQ&s=19 [Your request has been blocked. This could be due to several reasons.](https://www.microsoft.com/en-us/research/blog/introducing-aurora-the-first-large-scale-foundation-model-of-the-atmosphere/) [GraphCast: AI model for faster and more accurate global weather forecasting - Google DeepMind](https://deepmind.google/discover/blog/graphcast-ai-model-for-faster-and-more-accurate-global-weather-forecasting/) [‘Everything is Going to Be Robotic’ Nvidia Promises, as AI Gets More Real - YouTube](https://www.youtube.com/watch?v=nxO_t5N82m0) "despite the neural network only being trained on solution data (4 M). Amazingly it has the correct crystal structure, i.e., FCC. This is a phase transitions, an emergent phenomena, totally out of distribution, supposedly where AI is no good." https://x.com/timothyduignan/status/1797960944175427629?s=46&t=pNiPoM95FfYEFMSA5jsCMA [[2310.12535] Scalable molecular simulation of electrolyte solutions with quantum chemical accuracy](https://arxiv.org/abs/2310.12535) The Geometry of Concepts in LLMs [[2406.01506] The Geometry of Categorical and Hierarchical Concepts in Large Language Models](https://arxiv.org/abs/2406.01506) https://x.com/omarsar0/status/1798010546522103898 Towards Scalable Automated Alignment of LLMs [[2406.01252] Towards Scalable Automated Alignment of LLMs: A Survey](https://arxiv.org/abs/2406.01252) https://x.com/omarsar0/status/1798014572663583165 Best LLM dataset https://x.com/karpathy/status/1797313173449764933 The Top ML Papers of the Week (May 27 - June 2): - SimPO - GNN-RAG - Attention as an RNN - Abacus Embeddings - Symbolic Chain-of-Thought - Contextual Position Encoding https://x.com/dair_ai/status/1797241050089631762 [The Incompleteness of Evolution | Alfonso Martinez Arias - YouTube](https://youtu.be/wyZeli50nXg) [[2207.09238] Formal Algorithms for Transformers](https://arxiv.org/abs/2207.09238) [Paul Thagard on AI, Free Energy Principle, and the Cognitive Science of Misinformation - YouTube](https://www.youtube.com/watch?v=UqFk1Zpbag8) [The Dimensions of Visual Experience - YouTube](https://www.youtube.com/watch?v=p8GoXrJYDws) "We recently surveyed 300+ Harvard students about their views on AI. Our most surprising result: over 40% of Harvard students believe that AI extinction risk should be a global priority alongside pandemics and nuclear war. Link here: arxiv.org/abs/2406.00833 " https://x.com/GabrielDWu1/status/1797811385663172961?t=H3IZNaPV_XjTFhYNBaAVFQ&s=19 [Radware Bot Manager Captcha](https://iopscience.iop.org/article/10.1088/1751-8121/ad41a6?fbclid=IwZXh0bgNhZW0CMTEAAR295BJCQKmy682L_OvASqHjzWHADOcZVC6FfZ3fm1bGzGAkeIptlchVoOU_aem_ZmFrZWR1bW15MTZieXRlcw) [Average IQ of students by college major and gender ratio | Dr. Randal S. Olson](https://randalolson.com/2014/06/25/average-iq-of-students-by-college-major-and-gender-ratio/?fbclid=IwZXh0bgNhZW0CMTEAAR0XIxLxRaimI2m8x1v2fEimODXDNedGDwvFOjv02Ms-U7g4ZwhGTGRBqwM_aem_ZmFrZWR1bW15MTZieXRlcw) [Women and Low IQ Majors: Why do they correlate?](https://www.sebjenseb.net/p/women-and-low-iq-majors-why-do-they?fbclid=IwZXh0bgNhZW0CMTEAAR19IWkFoinSPsAdVYM7IHSlIz0erfZE5dEHacfuLsJwk6qv-NBF-N1qaxE_aem_ZmFrZWR1bW15MTZieXRlcw) Take as much of maths of physics as possible and generalize everything to oblivion