1940s
Turing Machine
Theoretical foundation of computation and algorithmic processes.
Feedback Control Systems
Automated goal-directed behavior using error-correction mechanisms.
Three Laws of Robotics
Science fiction foundational concepts influencing AI safety paradigms.
McCulloch-Pitts Neuron
The first mathematical model of a neural network.
Game Theory
Mathematical modeling of strategic interaction among rational decision-makers.
Von Neumann Architecture
The foundational design for stored-program computers.
ENIAC Programming
Programming the first general-purpose electronic digital computer.
Information Theory
Mathematical framework for quantifying, storing, and communicating information.
Cybernetics
The study of control and communication in animals and machines.
Hebbian Learning
A mechanism for neural plasticity: 'cells that fire together wire together.'
1950s
The Turing Test
A benchmark for machine intelligence based on conversational indistinguishability from humans.
Heuristic Search
Solving complex problems efficiently by using rules of thumb to guide the search.
Georgetown-IBM Translation Experiment
The first highly publicized demonstration of machine translation.
Symbolic AI Paradigm
The approach of representing knowledge through explicit, human-readable symbols and rules.
Dartmouth Conference
The founding event of Artificial Intelligence as an academic field.
Logic Theorist
The first AI program, designed to prove mathematical theorems.
General Problem Solver (GPS)
An early AI program intended to work as a universal problem-solving machine.
The Perceptron
The first trainable artificial neural network algorithm.
LISP Programming Language
The primary programming language for artificial intelligence research for decades.
Samuel's Checkers Program
The first self-learning software program.
1960s
Bayesian Methods in AI
Applying probability theory to handle uncertainty in reasoning.
Dendral
The first expert system, used for chemical analysis.
Unification Algorithm
Algorithm for solving equations between symbolic expressions, key to logic programming.
Fuzzy Logic
Reasoning with degrees of truth rather than absolute true/false boolean logic.
ELIZA
The first chatbot, demonstrating early natural language interaction.
Semantic Networks
A graph structure for representing knowledge through interconnected concepts.
Shakey the Robot
The first mobile robot to reason about its actions.
SHRDLU
An early natural language understanding computer program in a 'blocks world'.
Perceptron Limitations (Minsky & Papert)
Mathematical proof showing single-layer neural networks cannot solve non-linear problems like XOR.
STRIPS Planning
Stanford Research Institute Problem Solver—an automated action planner.
1970s
Expert Systems
Computer systems emulating the decision-making ability of a human expert.
Backpropagation (Theory)
The mathematical foundation for computing gradients in neural networks.
Automated Theorem Proving
The proving of mathematical theorems by a computer program.
Constraint Satisfaction Problems
Mathematical questions defined as a set of objects whose state must satisfy constraints.
Non-Monotonic Reasoning
A formal logic where conclusions can be retracted in light of new evidence.
MYCIN
A pioneering expert system for medical diagnosis.
PROLOG
Programming in Logic—a language based on formal logic.
The First AI Winter
A period of reduced funding and interest in AI research.
Knowledge Representation Frames
A data-structure for representing stereotyped situations in human memory.
HARPY Speech System
The first system capable of transcribing continuous speech using a large vocabulary.
1980s
Expert System Boom
Widespread corporate adoption of rule-based AI systems.
Hopfield Networks
A recurrent neural network serving as a content-addressable memory system.
Recurrent Neural Networks
Neural networks with feedback loops, allowing memory over time.
Boltzmann Machines
A type of stochastic recurrent neural network.
Genetic Algorithms
Search heuristics inspired by the theory of natural evolution.
Backpropagation Revival
Practical implementation of the backpropagation algorithm for multi-layer neural networks.
Parallel Distributed Processing
The manifesto of connectionism, modeling cognition as network activity.
The Second AI Winter
The collapse of the expert system market and specialized hardware.
Convolutional Neural Networks
Neural network architecture optimized for image and grid-like data.
Q-Learning
A foundational model-free reinforcement learning algorithm.
1990s
Statistical Machine Translation
Translating text based on statistical models derived from bilingual text corpora.
Bayesian Networks
Probabilistic graphical models representing causality and uncertainty.
Support Vector Machines (SVM)
A powerful supervised learning model for classification and regression.
Random Forests
An ensemble learning method using multiple decision trees.
Data Mining & KDD
The process of discovering patterns in large data sets.
Intelligent Agents
Autonomous software entities directing their activity towards achieving goals.
Long Short-Term Memory (LSTM)
An RNN architecture capable of learning long-term dependencies.
Deep Blue Chess Victory
The first computer to defeat a reigning world chess champion under tournament conditions.
Reinforcement Learning Formalization
The definitive mathematical framework for autonomous decision-making.
LeNet-5
A pioneering convolutional neural network for reading zip codes.
2000s
Word Embeddings (Pre-cursors)
Representing words as continuous vectors in a dense mathematical space.
MapReduce / Big Data
A programming model for processing massive datasets across distributed clusters.
DARPA Grand Challenges
A prize competition for American autonomous vehicles tracking desert and urban courses.
Sentiment Analysis
The automated identification of subjective information or emotional states from text.
Transfer Learning
Reusing a model trained on one task as the starting point for a second related task.
GPU Computing for ML
Utilizing graphics processing units for massive parallel mathematical computation.
Netflix Prize
An open competition for the best collaborative filtering algorithm to predict user ratings.
Deep Belief Networks
Generative graphical models trained greedily, one layer at a time.
Watson Jeopardy! Architecture
An open-domain question answering system.
ImageNet
A massive visual database designed for use in visual object recognition software research.
2010s
Self-Supervised Learning
Training methods where the data itself provides the supervision, eliminating manual labeling.
Explainable AI (XAI)
Methods and techniques to interpret the hidden decisions mathematically of deep neural networks.
AlexNet
The deep CNN that sparked the deep learning revolution by crushing the ImageNet benchmark.
Word2Vec
Highly scalable models to produce dense vector representations of language.
Variational Autoencoders (VAE)
A generative probabilistic model that compresses data into a latent space and reconstructs it.
Generative Adversarial Networks
An architecture where two neural networks contest with each other to generate entirely new, realistic data.
Attention Mechanism
A mechanism allowing models to automatically focus on relevant parts of the input sequentially.
Neural Machine Translation (NMT)
End-to-end translation models replacing complex pipelined statistical translation methods.
Batch Normalization
A mechanism to accelerate the training of deep neural networks by normalizing layer inputs.
Residual Networks (ResNets)
An architecture using skip connections to enable the training of hundred-layer deep networks.
Neural Style Transfer
An algorithm that takes the artistic style of one image and applies it to the content of another.
Real-time Object Detection (YOLO)
You Only Look Once (YOLO) algorithms framing object detection as a single regression problem.
Neural Architecture Search (NAS)
Using algorithms (like RL or evolutionary methods) to automatically design optimal neural networks.
Federated Learning
Training a centralized AI model using decentralized data stored on millions of user devices.
AlphaGo / AlphaZero
The AI system that conquered the ancient, mathematically intractable game of Go.
Transformer Architecture
The defining neural network architecture built entirely heavily reliant on 'self-attention'.
Reinforcement Learning from Human Feedback (RLHF)
A safety and fine-tuning mechanism for aligning AI behavior with human intent.
Capsule Networks
A novel neural network design to better capture spatial hierarchies in images.
BERT (Bidirectional Encoder Representations)
A landmark pre-trained open-source transformer model for understanding language context.
GPT Series (GPT-1, GPT-2)
The first generation of the Generative Pre-trained Transformer family.
2020s
GPT-3
A 175-billion parameter LLM that proved language models possessed 'emergent' zero-shot capabilities.
Diffusion Models
Thermodynamics-inspired networks that generate data by reversing a noise-addition process.
Retrieval-Augmented Generation (RAG)
Combining LLMs with external knowledge bases to prevent hallucinations.
Vision Transformers (ViT)
Applying the Transformer NLP architecture directly to image sequences.
AI Safety & Alignment
The formalized research field ensuring superintelligent systems follow human intent without disastrous side effects.
DALL-E Series
The definitive text-to-image synthesis models that brought generative AI to the mainstream.
Low-Rank Adaptation (LoRA)
A wildly efficient parameter-tuning method allowing individuals to fine-tune massive models on consumer hardware.
ChatGPT
The conversational AI interface that reached 100 million users in two months.
Constitutional AI
A mechanism for training helpful and harmless AI via a set of written principles.
Chain-of-Thought Prompting
Forcing standard language models to elicit step-by-step logical reasoning before answering.
Mixture of Experts (MoE)
A sparse network architecture selectively activating only specialized sub-networks per token.
GPT-4 / GPT-4o
The apex multimodal foundational model, parsing text, vision, and audio natively.
AI Agents & Tool Use
Equipping LLMs with the ability to dynamically call external APIs, calculators, and search browsers.
Synthetic Data Generation
Using frontier AI models to perfectly generate complex training data for smaller models.
State Space Models (Mamba)
A linear-time sequence architecture positioned as a highly efficient alternative to Transformers.
Multimodal Foundation Models
Models trained jointly on text, images, video, and audio data streams.
Reinforcement Learning from AI Feedback (RLAIF)
Replacing human trainers with AI trainers to scale model alignment securely.
AI Regulation (EU AI Act)
The world's first comprehensive horizontal legal framework strictly governing AI deployment.
Agentic Coding
AI systems that autonomously write, debug, test, deploy, and review complete codebases.
Test-Time Compute Scaling
Scaling inference compute dynamically to vastly improve model reasoning.
ImageNet
View on WikipediaA massive visual database designed for use in visual object recognition software research.
Why It Was Important
Fei-Fei Li recognized that advanced algorithms were useless without massive, labeled datasets to train them. Utilizing Amazon Mechanical Turk, her team labeled over 14 million images across 20,000 categories. The subsequent ImageNet Challenge provided the benchmark that sparked the 2012 deep learning revolution.
Who Invented It
Fei-Fei Li
Stanford professor and pioneer of data-centric AI.
Applications
- Computer Vision Benchmarking
- Deep Learning Datasets
- Transfer Learning bases