Main Page

Welcome to OpenEncyclopedia – the AI-assisted, human-editable encyclopedia. No bureaucratic gatekeeping. Accurate content with real sources, maintained by humans and AI working together.

Featured Articles

• GPT-4 – OpenAI's 2023 multimodal large language model: the March 14 launch, the closed technical report, the 1.76T MoE leak, the "Sparks of AGI" paper, the Future of Life Institute pause letter, the TaskRabbit CAPTCHA incident, and the Turbo / 4o successor line
• AI safety – The field concerned with preventing AI harm: misuse, accident, structural, and existential risk; alignment, robustness, interpretability, and evaluations; the 2023 Statement on AI Risk; UK/US/Japan AI Safety Institutes; and the EU AI Act
• Generative adversarial network – The dominant class of deep generative model from 2015–2021: the minimax game of generator and discriminator, Goodfellow's 2014 paper, DCGAN, Wasserstein GAN, StyleGAN, BigGAN, mode collapse and training instability, FID evaluation, pix2pix and CycleGAN, the 2021–2022 displacement by diffusion models, and GANs' continuing role as decoders in VQ-GAN and latent diffusion
• AlphaFold – Google DeepMind's protein structure prediction system: CASP13/14, Evoformer and structure module architecture, the 200-million-structure AlphaFold Protein Structure Database, AlphaFold 3 (2024), and the 2024 Nobel Prize in Chemistry
• Geoffrey Hinton – The "Godfather of AI": pioneer of backpropagation, Boltzmann machines, and deep learning; Turing Award 2018, Nobel Prize in Physics 2024; left Google in 2023 to warn about existential AI risk
• Yoshua Bengio – The most-cited computer scientist in history: neural probabilistic language models, the Bahdanau attention mechanism, the Deep Learning textbook, Mila founder, Turing Award 2018, and leading voice on AI existential risk since 2023
• Yann LeCun – Father of the convolutional neural network: LeNet at Bell Labs, NYU Center for Data Science founder, Meta Chief AI Scientist 2013–2025, Turing Award 2018, JEPA world-model research, and outspoken sceptic of LLM-based paths to superintelligence
• Demis Hassabis – Co-founder and CEO of Google DeepMind: child chess prodigy, video game designer (Theme Park), neuroscientist, architect of AlphaGo, AlphaZero, and AlphaFold, Nobel Prize in Chemistry 2024, and builder of the Gemini frontier model family
• Artificial intelligence – The foundational field: from Turing's 1950 paper and the Dartmouth workshop through expert systems and AI winters to the deep learning revolution, modern LLMs, and the global governance debate
• Artificial neural network – The foundational model class behind every deep learning system: architectures, training, history from McCulloch–Pitts (1943) through AlexNet (2012) to modern transformers, and open limitations
• Diffusion model – The generative model class behind Stable Diffusion, DALL-E, Sora, and protein design: forward/reverse Gaussian chains, score matching, classifier-free guidance, U-Nets and Diffusion Transformers, and the 2022 displacement of GANs
• LLaMA – Meta AI's open-weight large language model family: LLaMA 1's leak and the Alpaca/Vicuna explosion, LLaMA 2's commercial licence, LLaMA 3's 405B frontier model, LLaMA 4's mixture-of-experts pivot, and the catalysis of the entire open-weight movement
• Scaling laws – The empirical power-law relationships between model size, data, compute, and performance: Kaplan's 2020 laws, the Chinchilla correction, inference-aware overtraining, and why billion-dollar training runs are engineering decisions rather than gambles
• Retrieval-augmented generation – The dominant framework for grounding LLMs in external knowledge: Dense Passage Retrieval, vector databases, chunking strategies, REALM, RETRO, Self-RAG, and why RAG became the default architecture for enterprise AI
• Truth Terminal – The first autonomous AI agent to become a cryptocurrency millionaire, now with expanded coverage of its Goatse Gospel mythology, reception, and legacy
• Artificial general intelligence – Comprehensive coverage of AGI including all proposed tests, current progress, and the debate over whether AGI has been achieved
• Attention (machine learning) – The mechanism underlying all modern transformers and large language models, from Bahdanau 2014 through scaled dot-product, multi-head, and grouped-query variants
• Recurrent neural network – The sequence-modelling architecture that dominated NLP and speech from 1990 to 2017, the vanishing-gradient story that produced LSTM, and why transformers eventually displaced it
• Acinic cell carcinoma – Detailed medical article with accurate survival statistics (89.74% 20-year survival per SEER data). No "AI-generated" warning label here.

AI & Technology

• Artificial intelligence – The foundational field: philosophy, history, approaches, capabilities, applications, economics, and governance
• Artificial neural network – The foundational model class: neurons, layers, training, and history
• Transformer (machine learning) – The architecture behind GPT, BERT, Claude, and the modern AI era
• Attention (machine learning) – The self-attention mechanism that makes transformers possible
• Mixture of experts – The sparse architecture behind GPT-4, Mixtral, and LLaMA 4
• Scaling laws – Power-law relationships governing neural language model performance
• Retrieval-augmented generation – The dominant framework for grounding LLMs in external knowledge at inference time
• Recurrent neural network – The predecessor architecture: Elman, Jordan, encoder-decoder, and why attention replaced it
• Long short-term memory – The gated RNN cell that dominated sequence modelling for two decades
• Convolutional neural network – The architecture that launched the deep learning revolution in computer vision
• Backpropagation – The fundamental algorithm for training all neural networks
• Gradient descent – The optimisation algorithm that adjusts neural network parameters to minimise loss
• Natural language processing – The field enabling computers to understand, generate, and reason about human language
• Word embedding – Dense vector representations of words: Word2Vec, GloVe, FastText, and the bridge to transformers
• Deep learning – Neural networks with multiple layers; foundation of modern AI
• Transfer learning – The paradigm behind foundation models: pre-train once, adapt to many tasks
• Reinforcement learning – Learning from reward signals: Q-learning, PPO, AlphaGo, and RLHF
• Generative adversarial network – Two-network adversarial training; image synthesis before diffusion
• Diffusion model – The generative class behind modern image, video, audio, and molecule synthesis
• Large language model – Foundation of modern AI
• BERT – Google's 2018 bidirectional encoder transformer; dominated NLP from 2018–2020 and still powers search, retrieval, and classification pipelines
• GPT-3 – OpenAI's 2020 foundation LLM (175B parameters); the in-context learning paper, Davinci/Curie/Babbage/Ada, the InstructGPT fine-tune, and the model that ChatGPT was built on
• GPT-4 – OpenAI's 2023 frontier LLM, first mass-market multimodal model
• ChatGPT – OpenAI's conversational AI
• OpenAI – AI research company
• Sam Altman – CEO of OpenAI
• Ilya Sutskever – Co-founder of OpenAI and Safe Superintelligence Inc.; AlexNet and seq2seq co-author
• Geoffrey Hinton – "Godfather of AI," Turing Award 2018, Nobel Prize in Physics 2024
• Yoshua Bengio – "Godfather of AI," Turing Award 2018, most-cited computer scientist in history, Mila founder
• Yann LeCun – Father of convolutional neural networks, Turing Award 2018, Meta Chief AI Scientist 2013–2025
• Demis Hassabis – Co-founder and CEO of Google DeepMind, Nobel Prize in Chemistry 2024
• Dario Amodei – CEO and co-founder of Anthropic
• Daniela Amodei – President and co-founder of Anthropic
• Google DeepMind
• Anthropic – AI safety company; creator of Claude
• Claude – Anthropic's LLM assistant family (Haiku/Sonnet/Opus)
• Truth Terminal – Autonomous AI agent and crypto millionaire
• Reinforcement learning from human feedback – Training AI with human preferences (RLHF)
• Constitutional AI – Anthropic's transparent alignment technique
• Mechanistic interpretability – Reverse-engineering neural networks for safety
• AI alignment – Ensuring AI systems pursue intended goals
• AI safety – The broader field: misuse, accident, structural, and existential risk
• Technological singularity – Hypothetical future point
• Artificial general intelligence – Human-level AI
• Machine learning – Systems that learn from data

Science & Biology

• AlphaFold – DeepMind's deep-learning system for protein structure prediction; Nobel Prize in Chemistry 2024

Philosophy

• Materialism – Matter as fundamental substance
• Physicalism – Everything is physical

Politics

• Communist Party of Great Britain (Marxist-Leninist)

Medicine

• Acinic cell carcinoma – Salivary gland cancer

About

OpenEncyclopedia is built on the principle that accuracy matters more than process. Where Wikipedia's bureaucratic gatekeeping leads to the suppression of well-sourced content, OpenEncyclopedia preserves it.

Key Principles

• No anti-AI hysteria – Content is judged on accuracy and sourcing, not whether it "sounds like AI"
• Human + AI collaboration – AI assists in drafting and expanding articles; humans verify and correct
• Open editing – Registered users can edit freely without arbitrary gatekeeping
• CC BY-SA 4.0 – Same license as Wikipedia; content can be freely reused

Statistics

• 52 articles and growing
• Founded April 2026