← Mind Maps ·

Artificial Intelligence

Symbolic through connectionism to deep learning to LLMs and agents — the pioneers, the paradigms, the mechanisms

A mind map of artificial intelligence as a series of paradigm shifts: the pre-1956 foundations in logic and cybernetics; the symbolic era and its winters; the connectionist revival and the statistical turn; the deep-learning breakthrough; the foundation-model and LLM era; and the contemporary work on alignment, evaluation, and agentic systems. Named pioneers, institutions, papers, and mechanisms with dates across six branches.

Foundations & Formalism, pre-1956Symbolic AI & Expert Systems, 1956–1985Connectionist Revival & Statistical ML, 1986–2011Deep Learning Breakthrough, 2012–2017Foundation Models & LLMs, 2017–2023Alignment, Agents & Evaluation, 2017–presentMathematical logicComputation theoryCybernetics & information theoryEarly machines and programsPhilosophical originsThe Dartmouth foundingEarly programs and languagesPole institutionsKnowledge and expert systemsPhilosophy & critiqueFirst winters and the Fifth GenerationBackpropagation renaissanceRecurrent networks and memoryEarly deep learning and visionKernel methods, Bayes, ensemblesGame-playing and roboticsWeb-scale ML and ImageNetThe pivot: AlexNetRepresentation learningArchitectural innovationsReinforcement learning at scaleIndustrial labsThe TransformerPretraining and transferScalingMultimodality and generationScientific applicationsChatGPT and the deployment eraAlignment foundationsSafety research programsEvaluation benchmarksPost-training methodsAgentic systemsReasoning modelsGovernance and policyGeorge Boole — Laws of Thought, 1854Gottlob Frege — Begriffsschrift, 1879Russell & Whitehead — Principia Mathematica, 1910–1913David Hilbert — Entscheidungsproblem, 1900Kurt Gödel — incompleteness theorems, 1931Alan Turing — On Computable Numbers, 1936Universal Turing machine — the theory of general-purpose computationJohn von Neumann — First Draft, EDVAC, 1945Stored-program architecture — instruction and data in one memoryChurch–Turing thesis — the limits of the effectively computableNorbert WienerClaude ShannonMcCulloch & PittsMinsky & Edmonds — SNARC neural machine, Harvard 1951Christopher Strachey — Ferranti Mark 1 checkers, Manchester 1951Arthur Samuel — self-learning checkers at IBM, 1959 (coins "machine learning")Grace Hopper — A-0 compiler, 1952 (programs that write programs)Alan Turing — Computing Machinery and Intelligence, Mind 1950The Imitation Game — operational definition of machine intelligenceKarel Čapek — R.U.R., 1921 (coins "robot")I. J. Good — Speculations Concerning the First Ultraintelligent Machine, 19651955 — McCarthy, Minsky, Rochester, Shannon submit proposal1956 — Dartmouth Summer Research Project names the fieldNewell, Simon, Selfridge among the ~10 attendeesNewell & Simon — Logic Theorist at CMU/RAND, 1956Frank Rosenblatt — Perceptron, Cornell Aeronautical Laboratory 1957John McCarthy — LISP at MIT, 1958Recursion, garbage collection, symbolic expressions — the AI language for 30 yearsJoseph Weizenbaum — ELIZA/DOCTOR, MIT 1964–1966James Slagle — SAINT symbolic integrator, MIT 1965Terry Winograd — SHRDLU, MIT 1970Stanford AI Lab (SAIL) — founded by McCarthy, 1963MIT AI Lab — founded by Minsky & Papert, 1964CMU AI program — Newell & Simon lineageEdinburgh AI department — Donald Michie, 1963George Devol / Unimation — Unimate industrial robot, GM 1961Shakey the Robot — SRI International, 1969 (STRIPS planning)Feigenbaum & Lederberg — DENDRAL at Stanford, 1965Colmerauer & Roussel — Prolog at Aix-Marseille, 1972Edward Shortliffe — MYCIN medical expert system, Stanford 1972DEC — XCON/R1 expert system, 1980 (~$25M/yr savings)Doug Lenat — Cyc at MCC, 1984 (common-sense knowledge)Hans Berliner — BKG backgammon at CMU, 1979Minsky & Papert — Perceptrons, 1969 (XOR limitation)John Searle — Chinese Room argument, 1980Hubert Dreyfus — What Computers Can't Do, 1972Rodney Brooks — Elephants Don't Play Chess, MIT 1990 (subsumption)ALPAC Report, 1966 — NLP funding cut (first partial winter)Perceptrons backlash, 1969 — connectionist winter beginsJapan MITI — Fifth Generation Computer Systems, 1981 ($400M, 10 yr)Lisp Machine collapse, 1987 — hardware specialization failsSecond AI winter, 1987–1993 — DARPA cuts general AI fundingRumelhart, Hinton & Williams — backprop paper, Nature 1986McClelland & Rumelhart — Parallel Distributed Processing books, MIT Press 1986Paul Werbos — backprop derived in 1974 Harvard thesis (underattributed)Alexey Ivakhnenko — Group Method of Data Handling, USSR 1965–1971 (first multi-layer)John Hopfield — Hopfield networks, Caltech 1982Sepp Hochreiter — vanishing gradient problem, TUM 1991Hochreiter & Schmidhuber — LSTM, Neural Computation 1997Gating mechanisms — input, forget, output — solve vanishing gradient for sequencesYann LeCun — CNN for ZIP codes, Bell Labs 1989LeNet-5 — gradient-based learning for documents, 1998Geoffrey Hinton & Ruslan Salakhutdinov — deep belief nets, Science 2006Yann LeCun coins "deep learning," ~2003Rajat Raina & Andrew Ng — GPU deep learning, ICML 2009Cortes & Vapnik — Support Vector Machines, Bell Labs 1995Judea Pearl — Bayesian networks, UCLA 1988Leo Breiman — bagging 1996, Random Forests 2001Koller & Friedman — Probabilistic Graphical Models, MIT Press 2009Hidden Markov Models — the workhorse of pre-2012 speech recognitionChristopher Watkins — Q-learning, Cambridge PhD 1989Gerald Tesauro — TD-Gammon, IBM 1992IBM Deep Blue defeats Kasparov, 1997DARPA Grand Challenge 2004 — no vehicle finishes; Stanley (Stanford) wins 2005Boss (CMU) wins DARPA Urban Challenge, 2007Rodney Brooks — behavior-based robotics, MIT (founds iRobot)Page & Brin — PageRank, Stanford 1998Phrase-based statistical machine translation — IBM model, 2001Fei-Fei Li — ImageNet proposal, CVPR 2007ILSVRC competition launched, 2009 — 1.2M images, 1000 classesNorvig & Russell — AI: A Modern Approach textbook standardKrizhevsky, Sutskever & Hinton — AlexNet, Toronto 201215.3% top-5 error vs. 26.2% runner-up on ILSVRCReLU activations, dropout regularization, two GTX 580 GPUsCUDA/cuDNN — NVIDIA's software stack as the enabling substrateMikolov et al. — Word2Vec, Google 2013Kingma & Welling — Variational Autoencoders, Amsterdam 2013Ian Goodfellow et al. — Generative Adversarial Networks, 2014GloVe — Pennington, Socher & Manning, Stanford 2014Sutskever, Vinyals & Le — seq2seq with LSTMs, Google 2014Bahdanau et al. — attention for translation, Montréal/Jacobs 2014He, Zhang, Ren & Sun — ResNet, Microsoft Research Asia 2015Skip connections — enabling 152+ layer trainingBatch normalization — Ioffe & Szegedy, Google 2015DeepMind DQN plays Atari, Nature 2015AlphaGo defeats Lee Sedol 4–1, Seoul 2016Monte Carlo Tree Search + policy/value nets trained on games + self-playProximal Policy Optimization (PPO) — Schulman et al., OpenAI 2017Google Brain founded — Ng, Dean, Corrado, 2011FAIR (Facebook AI Research) founded, 2013Google acquires DeepMind for ~$500M, 2014OpenAI founded — Musk, Altman, Sutskever, Brockman, 2015 ($1B)Microsoft Research, IBM Research, Baidu IDL — parallel programsVaswani et al. — Attention Is All You Need, Google 2017Self-attention replaces recurrence — O(n²) but fully parallelMulti-head attention — multiple subspaces of representationPositional encodings — rotary (RoPE), learned, sinusoidal variantsFlash Attention 2 — Dao, Stanford 2023 (IO-aware exact attention)Howard & Ruder — ULMFiT, 2018 (fine-tune pretrained LM)OpenAI — GPT-1, Radford et al., 2018 (117M params)Google — BERT, Devlin et al., 2018 (340M params, bidirectional)T5 — Raffel et al., Google 2020 (text-to-text unified framing)Self-supervised learning — next-token, masked LM, contrastiveOpenAI — GPT-2, 2019 (1.5B params; staged release)Kaplan et al. — scaling laws, OpenAI 2020 (power-law loss curves)OpenAI — GPT-3, 2020 (175B params; in-context learning)Google — Switch Transformer, 2021 (1.6T params, sparse MoE)Google — PaLM 540B, 2022 (chain-of-thought emerges at scale)DeepMind — Chinchilla, 2022 (compute-optimal data/params ratio)OpenAI — CLIP, 2021 (contrastive image-text pretraining)OpenAI — DALL-E, 2021 (text-to-image generation)Google DeepMind — Gemini, Dec 2023 (native multimodality, 3 sizes)Stability AI — Stable Diffusion 1.4 open release, Aug 2022Latent diffusion models — Rombach et al., 2022Whisper — open-weight speech recognition, OpenAI 2022DeepMind — AlphaFold 2 at CASP14, 2020–2021Near-experimental accuracy on protein 3D structure~200M protein structures released publiclyAlphaFold-Multimer, AlphaFold 3 — beyond single-chain predictionOuyang et al. — InstructGPT, OpenAI 2022 (RLHF at scale)OpenAI — ChatGPT launch, Nov 30, 2022 (1M users in 5 days)OpenAI — GPT-4, Mar 2023 (multimodal, 90th percentile bar exam)Meta — LLaMA 1 and 2, 2023 (open-weight release; ecosystem explosion)Anthropic — Claude 2, 2023 (100K context window)Mistral AI — open-weight European models, 2023Christiano et al. — Deep RL from Human Preferences, OpenAI 2017Russell et al. — Cooperative Inverse Reinforcement Learning, Berkeley 2017MIRI (agent foundations) vs. empirical alignment — the two schoolsStuart Russell — Human Compatible book, 2019 (beneficial AI problem)Concrete Problems in AI Safety — Amodei et al., 2016Unsolved Problems in ML Safety — Hendrycks, Carlini et al., 2021Red-teaming — adversarial evaluation as safety practiceMechanistic interpretability — Chris Olah, Anthropic (circuits, superposition)Anthropic Responsible Scaling Policy (RSP), 2023Hendrycks et al. — MMLU, 2020 (57-subject knowledge benchmark)GPQA Diamond — graduate-level science benchmark, 2023SWE-bench — software-engineering task benchmark, Princeton 2023LMSYS Chatbot Arena — human preference leaderboardHumanEval, MATH, BIG-Bench, HellaSwag — domain-specific evalsRLHF — reward-model + PPO pipeline (InstructGPT, 2022)Anthropic — Constitutional AI, 2022 (written principles + RLAIF)DPO — Direct Preference Optimization, Rafailov et al., Stanford 2023RLAIF — reinforcement learning from AI feedbackAnthropic Claude 3.7 — extended thinking (user-visible reasoning), 2025Yao et al. — ReAct (reason + act), Princeton/Google 2023AutoGPT & BabyAGI — open-source agent frameworks, 2023Toolformer — Schick et al., Meta 2023 (self-taught tool use)Claude computer use — Anthropic, 2024Model Context Protocol (MCP) — Anthropic open standard, 2024OpenAI o3 and o4-mini — tool-in-reasoning-loop, 2025OpenAI o1 — inference-time chain-of-thought, Sep 2024DeepMind AlphaProof / AlphaGeometry 2 — IMO silver-medal, 2024Claude 3.7 Sonnet extended thinking — token-budgeted deliberation, 2025Google Gemini 2.5 Pro — 1M-token context, multimodal reasoning, 2025Test-time compute as a new scaling axisEU AI Act passed — European Parliament, Jun 2023 (risk-based tiers)US Executive Order on AI — Biden, Oct 2023Bletchley Declaration — UK AI Safety Summit, Nov 2023Anthropic Model Spec / Character Overview, 2024OpenAI Model Spec, 2024US export controls on AI chips — BIS rules 2022, 2023, 2024Cybernetics: Or Control and Communication, 1948Feedback, goal-directed behavior, communication as unifying scienceA Mathematical Theory of Communication, Bell System Tech J., 1948Entropy, channel capacity, the bitA Logical Calculus of the Ideas Immanent in Nervous Activity, 1943The neuron as a threshold logic unit — founding document of connectionismArtificial IntelligenceBrian Tighe · Mind Maps
Orbital mind map. Scroll to zoom, drag to pan, or use the buttons above (+ / − / 0 keys also work). Hover a node to highlight its path to the center and the subtree beneath it.

How to read this

The center holds the topic. The six branches fan out bilaterally — three on each side — each in its own color. Sub-branches nest three levels deep under each top-level branch. Hover a leaf to trace the path back to the center; hover a branch to see everything it contains.

This is the shape the topic has when you try to hold the whole field in your head at once. It is not an argument; it is a scaffold. The essays argue against or within scaffolds like this one.

More in the series