History of Artificial Intelligence

Warren McCulloch and Walter Pitts publish "A Logical Calculus of Ideas Immanent in Nervous Activity," introducing the first mathematical model of a neural network.

Alan Turing publishes "Computing Machinery and Intelligence," proposing the imitation game (later called the Turing Test) as a measure of machine intelligence.

The Dartmouth Conference, organised by John McCarthy, Marvin Minsky, Nathaniel Rochester, and Claude Shannon, coins the term "Artificial Intelligence" and launches it as a formal discipline.

Frank Rosenblatt develops the Perceptron — the first trainable artificial neural network — demonstrating that machines can learn from data.

Joseph Weizenbaum at MIT creates ELIZA, the first conversational chatbot, demonstrating that computers could simulate human dialogue in a limited domain.

Minsky and Papert's "Perceptrons" highlights the limitations of single-layer networks, triggering reduced AI funding and the first "AI Winter."

Alain Colmerauer develops Prolog, enabling logic-based AI programming and expert systems development.

XCON, an expert system developed by DEC, saves the company $40M per year by configuring computer orders — triggering a commercial AI boom.

Rumelhart, Hinton, and Williams popularise backpropagation, enabling training of multi-layer neural networks and reviving neural network research.

The expert systems market collapses and the Lisp Machine market fails, triggering the second AI Winter as investment dries up.

IBM's Deep Blue defeats world chess champion Garry Kasparov — the first time a computer beats a reigning world chess champion under standard tournament conditions.

Yann LeCun introduces LeNet and the MNIST handwritten digit dataset, laying the foundation for modern convolutional neural networks.

Geoffrey Hinton and colleagues publish a landmark paper on deep belief networks, launching the modern deep learning era with efficient training algorithms for deep neural networks.

Fei-Fei Li launches ImageNet, a database of 14 million labelled images across 20,000 categories that becomes the benchmark for visual AI research.

IBM Watson defeats Jeopardy! champions Ken Jennings and Brad Rutter, demonstrating natural language understanding and information retrieval at superhuman levels.

AlexNet wins the ImageNet competition by a massive margin, reducing the error rate from 26% to 15.3% and proving the commercial potential of deep convolutional neural networks.

Ian Goodfellow introduces Generative Adversarial Networks (GANs), enabling AI to generate realistic synthetic data and images for the first time.

OpenAI is founded by Elon Musk, Sam Altman, Greg Brockman, Ilya Sutskever, and others with a $1B commitment to develop safe, beneficial AGI.

DeepMind's AlphaGo defeats world Go champion Lee Sedol 4-1, mastering a game considered far too complex for computers — a landmark in reinforcement learning.

Google researchers publish the Transformer architecture, introducing multi-head self-attention and enabling efficient parallel processing of sequences — the foundation of every modern LLM.

Google releases BERT and OpenAI releases GPT-1, pioneering the era of large pre-trained language models that can be fine-tuned for many downstream tasks.

OpenAI initially withholds GPT-2 citing "safety concerns," sparking the first major public debate about AI safety and misuse potential.

OpenAI releases GPT-3 with 175 billion parameters — the largest language model ever trained at the time — demonstrating emergent few-shot learning capabilities that stunned the research community.

GitHub Copilot launches as the first AI pair programmer, and OpenAI's DALL-E demonstrates text-to-image generation — marking AI's entry into creative work and software development.

OpenAI launches ChatGPT (reaching 100M users in 2 months — the fastest consumer product in history), while Stable Diffusion democratises AI image generation. The generative AI era begins in earnest.

GPT-4 demonstrates multimodal reasoning; Meta releases Llama 2 as open-source; Google launches Bard and Gemini; Anthropic releases Claude. AI becomes the fastest-growing technology sector in history.

The UK hosts the first AI Safety Summit at Bletchley Park; the US issues an Executive Order on AI safety; the EU passes the AI Act — marking the start of formal AI regulation globally.

OpenAI's o1 introduces chain-of-thought reasoning; AI agents begin handling multi-step tasks autonomously; Claude 3.5, Gemini 1.5, and Llama 3 push multimodal capabilities further.

Humanoid robots from Figure AI, Agility Robotics, and 1X Technologies deploy in warehouses; Boston Dynamics' Atlas goes electric; autonomous vehicle deployments expand in major cities.

OpenAI, Google DeepMind, and Anthropic each claim progress toward AGI-capable systems; reasoning and planning benchmarks approach human performance across most professional domains.