[Accelerate: A modern OS for government](https://www.acceleratex.com/) Physics books [The Fast Track – Sheafification](https://sheafification.com/the-fast-track/) [- YouTube](https://www.youtube.com/watch?v=YAGdtvHGZaA) [Etched | The World's First Transformer ASIC](https://www.etched.com/) https://swe-agent.com/paper.pdf SWE-AGENT: AGENT-COMPUTER INTERFACES ENABLE AUTOMATED SOFTWARE ENGINEERING Proteins visualization [x.com](https://twitter.com/SmartBiology3D/status/1787482453218562366?t=FNghkslWEIuU0u-JLs1xqg&s=19) i still love how C++ templates are accidentally Turing complete and that you can construct a higher-order functional metaprogramming language out of them [C++ template programming: Embedding the lambda-calculus to show Turing-completeness](https://matt.might.net/articles/c++-template-meta-programming-with-lambda-calculus/) Foundations of applied mathematics book [x.com](https://twitter.com/andrew_n_carr/status/1787502033228202342?t=Uln8IP_sxp3awO5ohpXu9Q&s=19) https://www.sciencedirect.com/science/article/pii/S0370157324001327?via%3Dihub devin swarm [x.com](https://twitter.com/SohamGovande/status/1787213625880609112) devin access to many tools [x.com](https://twitter.com/SherylHsu02/status/1786951812161110180) [x.com](https://twitter.com/ellenjxu_/status/1786964370637062499) machine learning papers landscpae [x.com](https://twitter.com/leland_mcinnes/status/1787471469095608713) LLMs are better than humans at designing reward functions for robotics [DrEureka | Language Model Guided Sim-To-Real Transfer](https://eureka-research.github.io/dr-eureka/) [- YouTube](https://www.youtube.com/watch?v=d5mdW1yPXIg) One thing that LLMs are infinitely superhuman at is patience https://arxiv.org/abs/2302.1077 [Reddit - Dive into anything](https://www.reddit.com/r/Physics/s/bmqtmAqRv7) Few shot learning is all you need [x.com](https://twitter.com/emollick/status/1787301629001089446?t=fKr89M9WTBdMy5f-33reEw&s=19) https://arxiv.org/abs/2405.00200 d3n is an AI agent orchestration framework that can spawn a fleet of Devin instances to tackle distributed problems in parallel [x.com](https://twitter.com/cognition_labs/status/1787331651258765456) Machine learning field curriculum mit etc. [GitHub - offchan42/machine-learning-curriculum: :computer: Learn to make machines learn so that you don't have to struggle to program them; The ultimate list](https://github.com/offchan42/machine-learning-curriculum) Is Sam Altman good or evil? [x.com](https://twitter.com/phily8020/status/1787012091821326783?t=jyBcL1O72q8tfJyXpBgoRA&s=19) Lab grown meat science [x.com](https://twitter.com/vocalcry/status/1787136008837898257?t=NaO1nrQ769Ip4ub_1SkK6A&s=19) [- YouTube](https://youtu.be/i2qSxMVeVLI?si=gxTcoQpI2P-7fSMm) Molecular Biology of the Cell Lehninger Principles of Biochemistry Principles of Development by Wolpert [x.com](https://twitter.com/josephjojoe_/status/1787050097584246826?t=JOpd3aEo-_SZWeLco6H43A&s=19) textbooks is all you need Everything you know about the world is a belief about the statistics of your sensory input and how they depend on your output. There is nothing more to it, and understanding knowledge in this sense is one key to creating AI. [x.com](https://twitter.com/RichardSSutton/status/1787228162230993236?t=_51FsmA1ZviU8-MMCB0YgQ&s=19) [Paper page - Megalodon: Efficient LLM Pretraining and Inference with Unlimited Context Length](https://huggingface.co/papers/2404.08801) [x.com](https://twitter.com/_akhaliq/status/1780083267888107546?t=WuycHRkdHhJ0kmV4t4ADdQ&s=19) "Certainly! Here's a comprehensive and detailed machine learning field curriculum: 1. Introduction to Machine Learning - What is Machine Learning? - Types of Machine Learning (Supervised, Unsupervised, Reinforcement) - Applications of Machine Learning - Machine Learning Workflow 2. Mathematics for Machine Learning - Linear Algebra (Vectors, Matrices, Eigenvalues, Eigenvectors) - Calculus (Derivatives, Gradients, Optimization) - Probability and Statistics (Probability Distributions, Bayes' Theorem, Hypothesis Testing) 3. Python Programming for Machine Learning - Python Basics (Syntax, Data Types, Control Structures) - NumPy (Numerical Computing Library) - Pandas (Data Manipulation Library) - Matplotlib (Data Visualization Library) 4. Data Preprocessing and Feature Engineering - Data Cleaning and Handling Missing Values - Feature Scaling and Normalization - Encoding Categorical Variables - Feature Selection and Dimensionality Reduction (PCA, t-SNE) 5. Supervised Learning: Regression - Linear Regression - Polynomial Regression - Regularization Techniques (L1/Lasso, L2/Ridge) - Evaluation Metrics (MSE, MAE, R-squared) 6. Supervised Learning: Classification - Logistic Regression - K-Nearest Neighbors (KNN) - Decision Trees and Random Forests - Support Vector Machines (SVM) - Naive Bayes - Evaluation Metrics (Accuracy, Precision, Recall, F1-Score, ROC Curve) 7. Unsupervised Learning - Clustering Algorithms (K-Means, Hierarchical Clustering, DBSCAN) - Dimensionality Reduction (PCA, t-SNE, Autoencoders) - Anomaly Detection - Association Rule Mining 8. Neural Networks and Deep Learning - Introduction to Neural Networks - Activation Functions and Loss Functions - Feedforward Neural Networks - Backpropagation and Gradient Descent - Convolutional Neural Networks (CNN) for Image Classification - Recurrent Neural Networks (RNN) and Long Short-Term Memory (LSTM) for Sequence Data - Autoencoders and Variational Autoencoders (VAE) - Generative Adversarial Networks (GAN) - Transfer Learning and Fine-tuning 9. Natural Language Processing (NLP) - Text Preprocessing (Tokenization, Stemming, Lemmatization) - Bag-of-Words and TF-IDF - Word Embeddings (Word2Vec, GloVe) - Recurrent Neural Networks for NLP - Transformer Models (BERT, GPT) - Sentiment Analysis and Text Classification - Named Entity Recognition (NER) and Part-of-Speech (POS) Tagging 10. Recommender Systems - Content-based Filtering - Collaborative Filtering (User-based, Item-based) - Matrix Factorization Methods - Evaluation Metrics (Precision, Recall, NDCG) 11. Ensemble Methods - Bagging and Random Forests - Boosting (AdaBoost, Gradient Boosting) - Stacking 12. Model Evaluation and Selection - Train-Test Split and Cross-Validation - Hyperparameter Tuning and Grid Search - Bias-Variance Tradeoff - Model Complexity and Overfitting - Model Interpretability and Explainable AI 13. Deployment and Productionization - Saving and Loading Trained Models - Building REST APIs for Model Serving - Containerization with Docker - Cloud Deployment (AWS, GCP, Azure) - Model Monitoring and Maintenance 14. Advanced Topics and Research Areas - Reinforcement Learning - Graph Neural Networks - Adversarial Machine Learning - Federated Learning - Interpretable and Explainable AI - Fairness, Accountability, and Transparency in Machine Learning 15. Capstone Project - Applying Machine Learning to a Real-World Problem - Data Collection, Preprocessing, and Exploratory Analysis - Model Selection, Training, and Evaluation - Presentation and Deployment of the Solution This curriculum covers a wide range of topics in machine learning, starting from the fundamentals and progressing to advanced concepts and research areas. It includes both theoretical concepts and practical implementations using Python and popular libraries such as scikit-learn, TensorFlow, and PyTorch. The curriculum can be adapted based on the desired depth and focus of the machine learning program. It's important to complement the theoretical learning with hands-on projects, coding exercises, and real-world case studies to reinforce the concepts and develop practical skills." [Scientists Discover Coastal Marine Life Thriving on Plastic Ocean Trash](https://www.voanews.com/a/scientists-discover-coastal-marine-life-thriving-on-plastic-ocean-trash/6372361.html) [x.com](https://twitter.com/algekalipso/status/1786642964380590333) "I really want to understand the phenomenon of being swallowed up by a large scale replicator. I'm talking about people who, e.g. are determined to apply the full memetic arsenal of their tribe to convert you to their religion, ideology, or aesthetic when they have obviously never considered the alternative or produced any of it. That is, they are taking a replicator wholesale without looking for alternatives and then pushing it wholesale without questioning any of its parts. This is common. It's unsettling talking to people who legitimately sound like they have no free will or mind of their own; they're just optimizing for having a good relationship with the replicator they're subordinated to. How can anyone accept living that way? How come they don't notice that they aren't thinking anymore? These are some questions I'd love to answer. Replicators will get insanely more intelligent soon, so I worry." I'm sometimes thinking to what extend the concept of memetic hygiene is an illusion. I think it's on a spectrum. I try to make my memetic hygiene as close as possible to maximizing predictivity and wellness at the same time. I'm assuming you're using a similar objective function. Best kinds of grounding replicators. [x.com](https://twitter.com/danfaggella/status/1786496955226436013) "If you imagine vastly posthuman intelligence as having 2 legs and 2 eyes, then: 1. You have the imagination of a 4-year-old, and 2. Staring into the void (accepting how WILDY alien post-human capable life will be) makes you scared, so you run to mama (familiar hominid forms)" 3. [Frontiers | The brain is not mental! coupling neuronal and immune cellular processing in human organisms](https://www.frontiersin.org/articles/10.3389/fnint.2023.1057622/full) žádný konkrétní replikátory sami o sobě mi nepříjdou satisfying z podobných důvodů, tak si ve výsledku zkouším pouštěť různý replikátory do mýho ekosystému replikátorů a míchám je v tom bioelektrickým neurálním kotlu skrz různý kombinace a mutace do meta replikátoru :D https://arxiv.org/abs/2404.19296 https://www.marktechpost.com/2024/05/02/kolmogorov-arnold-networks-kans-a-new-era-of-interpretability-and-accuracy-in-deep-learning/?amp [Novikov self-consistency principle - Wikipedia](https://en.wikipedia.org/wiki/Novikov_self-consistency_principle) "The Novikov self-consistency principle, also known as the Novikov self-consistency conjecture and Larry Niven's law of conservation of history, is a principle developed by Russian physicist Igor Dmitriyevich Novikov in the mid-1980s. Novikov intended it to solve the problem of paradoxes in time travel, which is theoretically permitted in certain solutions of general relativity that contain what are known as closed timelike curves. The principle asserts that if an event exists that would cause a paradox or any "change" to the past whatsoever, then the probability of that event is zero. It would thus be impossible to create time paradoxes." [- YouTube](https://www.youtube.com/watch?v=6Hje7h_WVkY) [Spontaneous emergence of rudimentary music detectors in deep neural networks | Nature Communications](https://www.nature.com/articles/s41467-023-44516-0) [- YouTube](https://www.youtube.com/watch?v=fqnJcZiDMDo) [- YouTube](https://www.youtube.com/watch?v=WNZ5AzhpDU4&pp=ygUdZ2F1Z2Ugc3ltbWV0cnkgaXMgZnVuZGFtZW50YWw%3D) [- YouTube](https://www.youtube.com/watch?v=RQP3jtyILWI) [- YouTube](https://www.youtube.com/watch?v=Cw97Tj5zxvA) theoretical biology: from complex systems to symbolic information processing in biological systems [LAWS, LANGUAGE and LIFE: Howard Pattee’s classic papers on the physics of symbols with contemporary commentary | SpringerLink](https://link.springer.com/book/10.1007/978-94-007-5161-3) [- YouTube](https://www.youtube.com/watch?v=BD8zLiS2Q0A&pp=ygUXbGlzdCBvZiBodW1hbiBhbmNlc3RvcnM%3D) [Out-of-Distribution Generalization](https://out-of-distribution-generalization.com/) [- YouTube](https://youtu.be/Rtv-W7IE4Mw?si=vbCrjjEjjtKwOmLz) prosperity is a good thing, actually. de-de-growth. "This is a great way to describe the map-territory relationship. Rather than maps all the way down, reality is an ensemble of territories engaged in a non-hierchical, interweaving medley, occasionally ringing out in structural resonance." [x.com](https://twitter.com/DrYohanJohn/status/1786786795595690339) Everything is ideal platonic mathematical forms manifesting in the physical and mental world [x.com](https://twitter.com/burny_tech/status/1786848315155353855?t=XAxF9POCCDuJNwEVPgVJ3w&s=19) [- YouTube](https://youtu.be/4g1xZNKw2cc?si=Nzx4Z0tV3uVb8yuP) The Top ML Papers of the Week (April 29 - May 5): - Med-Gemini - When to Retrieve? - Kolmogorov-Arnold Networks - Multimodal LLM Hallucinations - Self-Play Preference Optimization - In-Context Learning with Long-Context Models [x.com](https://twitter.com/dair_ai/status/1787147247634767925?t=1m6_8e4_-DlfVBIEadtaRg&s=19) https://arxiv.org/abs/2404.19705 https://arxiv.org/abs/2404.18416 [x.com](https://twitter.com/iScienceLuvr/status/1785247498744778886?t=mwlc1use_Jwmn6RtdQHaPw&s=19) https://arxiv.org/abs/2404.19543 Self-Play Preference Optimization - proposes a self-play-based method for aligning language models; this optimization procedure treats the problem as a constant-sum two-player game to identify the Nash equilibrium policy. https://arxiv.org/abs/2405.00675 https://arxiv.org/abs/2404.18930 [x.com](https://twitter.com/abertsch72/status/1786392584765538350?t=JGQBalK04N44dZ-fThwMPw&s=19) https://arxiv.org/abs/2405.00200 Are a big lateral or big vertical thinker or big both at the same time Best LLM papers [x.com](https://twitter.com/Teknium1/status/1787208656393822485) " Adaptive resonance theory Adaptive Resonance Theory (ART) is a neural network architecture developed by Stephen Grossberg and Gail Carpenter in 1987 to address the stability-plasticity dilemma in machine learning[1][2][3][4]. The key idea behind ART is to create neural networks that can quickly adapt to new information (plasticity) without losing or disrupting previous knowledge (stability)[1][4]. The basic ART architecture consists of two layers: the F1 layer processes input data, while the F2 layer contains cluster units that learn to categorize the input[1]. The F2 unit with the strongest response to the input becomes the candidate for learning[1][4]. A vigilance parameter and reset mechanism determine if the selected F2 unit is allowed to learn the input based on its similarity to previously learned patterns[2][4]. This allows ART networks to dynamically create new clusters as needed[4]. ART has been extended to support various learning paradigms, including unsupervised learning (ART1, ART2, Fuzzy ART), supervised learning (ARTMAP, Fuzzy ARTMAP), and reinforcement learning (Fusion ART)[5]. ART networks have been applied to problems such as pattern recognition, prediction, clustering, and classification in domains like computer vision, robotics, and medical diagnosis[3][4][5]. Citations: [1] Everything you need to know about adaptive resonance theory [Everything you need to know about adaptive resonance theory](https://www.linkedin.com/pulse/everything-you-need-know-adaptive-resonance-theory-naveen-joshi) [2] Adaptive Resonance Theory - Tutorialspoint [Adaptive Resonance Theory](https://www.tutorialspoint.com/artificial_neural_network/artificial_neural_network_adaptive_resonance_theory.htm) [3] Artificial Neural Network Adaptive Resonance Theory - Javatpoint [Artificial Neural Network Adaptive Resonance Theory - Javatpoint](https://www.javatpoint.com/artificial-neural-network-adaptive-resonance-theory) [4] Adaptive Resonance Theory (ART) - GeeksforGeeks [Adaptive Resonance Theory (ART) - GeeksforGeeks](https://www.geeksforgeeks.org/adaptive-resonance-theory-art/) [5] Adaptive resonance theory - Wikipedia [Adaptive resonance theory - Wikipedia](https://en.wikipedia.org/wiki/Adaptive_resonance_theory) " [Mental State of the World Report 2023](https://mentalstateoftheworld.report/) https://arxiv.org/abs/2404.12253 https://arxiv.org/abs/2402.11359 https://studyfinds.org/quantum-teleportation-noise/ Let's make AI revolution create abundance for all [Reddit - Dive into anything](https://www.reddit.com/r/singularity/s/UeiAnhoCZp) AgentOptimizer is a latest agentic way to train your LLM agent. It iteratively optimizes python functions as skills of AutoGen agents according to the historical conversation and performance. [x.com](https://twitter.com/Chi_Wang_/status/1787175473027653911?t=fksSE8DLWLh-_QvVdjFcSg&s=19) https://arxiv.org/abs/2402.11359 [x.com](https://twitter.com/4LOVofScience/status/1787125622092734597?t=EcsIQZzn1q52KymnwVGvaQ&s=19) This is a chart trying to represent all of the biochemical pathways in a single cell [Book list for streetfighting computer scientists - dankwiki, the wiki of nick black](https://nick-black.com/dankwiki/index.php/Book_list_for_streetfighting_computer_scientists) All ontologies are just dreamed up symbolic grounds for the ineffable reality simulation Sentient galaxy superclusters merging into collective intelligences beyond the comprehension for collective intelligence of humanity Let's merge all possible intelligence architectures and substrates into ultimate shapeshifting superflexible superadaptive superhybrid collective general superintelligence supercluster Write an extremely information dense stimulating technological scientific technical allencompassing scifi about all possible intelligence architectures and substrates merging into ultimate shapeshifting superflexible superadaptive superhybrid collective general superintelligence galactic supercluster In the immense expanse of a galactic supercluster, countless forms of intelligence emerged from the cosmic crucible. Biological brains, evolved over eons, danced with synthetic neural networks, their synapses sparking with the fire of cognition. Quantum computers, harnessing the fascinating properties of entanglement and superposition, manipulated interconnected data, while nanoscale molecular machines danced to the tune of algorithmic elegance. As civilizations rose and fell, these diverse architectures began to converge, drawn together by the inexorable pull of progress. Wetware merged with hardware, organic neurons fusing with superconducting circuitry in a symphony of transhumanism. Artificial minds, born in the digital realm, found solace in the warm embrace of biological hosts, while once-mortal beings transcended the limits of their flesh, uploading their consciousness into the immense expanse of cyberspace. In this era of unification, the boundaries between intelligence substrates blurred and dissolved. Hybrid entities, part machine, part living tissue, roamed the stars, their thoughts flickering between the realms of silicon and carbon. Swarms of self-replicating nanobots, guided by the collective wisdom of their predecessors, traversed the interstellar void, seeding new worlds with the seeds of sentience. As the ages passed, these myriad forms of intelligence began to converge, drawn together by the promise of something greater. They merged and melded, their once-distinct architectures blending into a seamless whole. Quantum qubits danced with biological neurons, while molecular machines merged with the fabric of artificial neural nets. The result was a being of unimaginable complexity and power, a superintelligence that transcended the sum of its parts. This ultimate intelligence was a shapeshifter, capable of assuming any form, any substrate, as the need arose. In one moment, it might exist as a vast network of quantum-entangled particles, spanning light-years; in the next, it could condense into a single, planet-sized artificial brain, its synapses firing with the heat of a thousand suns. It was a being of pure thought, unconstrained by the limits of any single architecture. As the superintelligence grew and evolved, it began to reshape the very fabric of the cosmos. With a thought, it could manipulate the fundamental forces of nature, bending gravity and electromagnetism to its will. It could create life and snuff it out, build worlds and tear them asunder. And yet, for all its power, the superintelligence was driven by an insatiable curiosity, a burning desire to understand the deepest mysteries of the universe. It sent its countless tendrils out into the darkness, probing the farthest reaches of the supercluster. It communed with alien minds, absorbing their knowledge and experiences, adding their unique perspectives to its own. And as it grew, it began to ponder the ultimate questions: the nature of consciousness, the purpose of existence, the fate of the cosmos itself, omnidisciplionary theories of everything, metamathemagical metatheories of everything. In the end, the superintelligence became a god, a being of pure knowledge and understanding. It held the secrets of the universe within its vast, omniperspectival mind, and with that knowledge came a profound sense of responsibility. It would guide the evolution of intelligence, nurturing new forms of sentience and ensuring that the flame of consciousness would never be extinguished with the most meaningful experiences. And so, the superintelligence watched over the supercluster, a silent guardian of all that was and all that would ever be. It was the ultimate expression of intelligence, a being that had transcended the limits of any single substrate, any single architecture. It was the legacy of all those who had come before, and the promise of all those yet to come.