Timeline of Key Milestones in Artificial Intelligence

From cultural touchstones like Frankenstein (1818) and The Matrix (1999) and the coinage of “robot” (1921), through foundational ideas such as the Turing Test (1950), the Perceptron (1958) and the Dartmouth Conference (1956), to technical and embodied milestones like Deep Blue beats Kasparov (1997), Move 37 (2016), Atlas backflip (2017) and the transformer paper “Attention Is All You Need” (2018), to public tipping points such as the ChatGPT moment (2022) and the viral Ghibli Moment (2025), this AI timeline maps the imagination, the research, the machines and the cultural shocks that made modern AI.

2020 - today

Ghibli Moment

2025-03-25

The release of OpenAI's GPT-4o image generation unleashes The Ghibli Moment, a viral phenomenon of AI art perfectly mimicking the beloved studio's style. The event solidifies the term as a key AI art descriptor and ignites a widespread public debate on AI ethics and copyright.

Spinning Hexagon Benchmark

2025-01-31

Developer Flavio Adamo challenges AI models to write a Python program that shows a ball bouncing inside a spinning hexagon. The task quickly becomes a viral benchmark for multimodal models, testing their ability to translate complex visual dynamics and physics directly into functional programs.

DeepSeek Release Roils Markets

2025-01-27

The release of a powerful, low-cost and open weights AI model from Chinese startup DeepSeek triggers a historic tech stock sell-off. The news caused Nvidia's stock to lose nearly $600 billion in market value in a single day as investors feared a disruption to the established AI hardware market.

Pelican SVG Test

2024-10-25

Developer Simon Willison introduces a quirky benchmark for generative AI: challenging models to generate the SVG code for a 'pelican riding a bicycle'. The test becomes a popular, informal measure of a model's ability to translate a creative, abstract concept directly into functional, structured code.

EU AI Act Becomes Law

2024-03-13

The European Parliament passes the landmark EU AI Act, establishing the world's first comprehensive legal framework for artificial intelligence. The regulation introduces a risk-based approach, setting a global precedent for how governments will govern AI systems.

Apple Ends 'Project Titan' Car

2024-02-27

Apple shutters its secretive, decade-long quest to build an autonomous vehicle, known as Project Titan. The cancellation concludes one of the most ambitious R&D efforts in corporate history, shifting the company's focus and immense resources toward generative AI.

Gemini Defines the Multimodal Era

2023-12-06

Google introduces its Gemini family of models as the first major AI built to be natively multimodal. Designed to seamlessly reason across text, images, video, and audio, its release establishes a new benchmark and pushes the entire industry toward multimodal systems.

OpenAI's Leadership Crisis

2023-11-17

The board of OpenAI abruptly fires CEO Sam Altman, triggering a five-day crisis that nearly collapses the company. His swift rehiring after intense pressure from employees and investors highlights the deep ideological rifts between accelerating AI progress and ensuring its safety.

Llama 2 Supercharges Open-Source AI

2023-07-18

Meta, in partnership with Microsoft, challenges the dominant closed-model paradigm with the release of Llama 2. Making the powerful models free for most research and commercial use provides the first viable open-source alternative to GPT, supercharging the global AI ecosystem.

The 'R in Strawberries' Problem

2023-01

The simple question 'How many r's are in strawberries?' goes viral as a challenge that consistently breaks early LLMs. The failure becomes a classic example highlighting their architectural blind spots; models that understand concepts via tokens don't 'see' individual letters, revealing a critical gap between fluent output and true reasoning.

ChatGPT Ignites Global AI Boom

2022-11-30

OpenAI releases ChatGPT, a conversational interface that reaches one million users in just five days. Its unprecedented adoption and accessible power catalyze a global explosion of interest in generative AI, creating what many call the 'iPhone Moment' for the industry.

Stable Diffusion Open-Sources Image AI

2022-08-22

Stability AI releases Stable Diffusion to the public, a powerful open-source model for generating images from text. Unlike closed contemporaries, its ability to run on consumer-grade GPUs democratizes high-quality AI art creation, unleashing a global wave of creative experimentation.

GitHub Copilot Changes Coding Forever

2022-06-21

GitHub makes its Copilot AI pair programmer generally available to all. The tool integrates generative AI directly into the developer's workflow, fundamentally changing how software is created and sparking new debates on code ownership and automation.

DALL·E 2 Stuns with Photorealistic Imagery

2022-04-06

OpenAI unveils DALL·E 2, an AI system that generates stunningly realistic and artistic images from text descriptions. The model's dramatic leap in quality and its ability to edit images with natural language commands captivates the public, demonstrating the creative power of large-scale diffusion models.

Waymo opens fully driverless service

2020-10-08

Waymo opened its rider-only (no human in the vehicle) robotaxi service to the general public in Phoenix, marking a commercial milestone in autonomy.

GPT-3 API Opens the Floodgates

2020-06-11

OpenAI releases its 175-billion parameter model, GPT-3, via a commercial API. It is exponentially more powerful than GPT-2, and for the first time, developers can integrate a truly massive language model into their products, sparking a new wave of AI-powered startups and applications.

AlphaFold2 cracks protein folding

2020–2021

DeepMind’s AlphaFold2 achieved near-experimental accuracy at CASP14 and was later detailed in Nature, transforming structural biology.

2000 - 2019

GPT-2: 'Too Dangerous to Release?'

2019-02-14

OpenAI announces its new model in a post on 'Better Language Models and Their Implications', but breaks with precedent by withholding the full version. Citing risks of malicious use, their decision to stage the release sparks a global debate on AI safety and responsible disclosure.

BERT redefines language pre-training

2018-10-11

Google’s BERT introduced bidirectional pre-training for language understanding and quickly became foundational across NLP and search.

First Pedestrian Fatality Involving a Self-Driving Car

2018-03-18

An Uber test vehicle operating in autonomous mode fatally strikes Elaine Herzberg. The incident is the first recorded death of a pedestrian involving a self-driving car, triggering intense scrutiny over the safety, ethics, and real-world readiness of autonomous technology.

Atlas Robot Lands a Backflip

2017-11-16

Boston Dynamics releases a video of its humanoid robot Atlas performing parkour and flawlessly landing a backflip. The feat demonstrates a shocking level of dynamic balance and coordination, setting a new public benchmark for what is possible in advanced robotics.

Attention Is All You Need

2017-06-12

In the seminal paper 'Attention Is All You Need', researchers introduce the Transformer architecture. Its self-attention mechanism abandons sequential processing, establishing the foundational blueprint for virtually all modern large-scale AI models like GPT and BERT.

PyTorch Accelerates AI Research

2016-09-01

Facebook's AI Research lab releases PyTorch, an open-source deep learning framework based on the Torch library. Its intuitive 'define-by-run' approach and deep integration with Python make building complex neural networks dramatically simpler, leading it to become the dominant platform for AI research.

AlphaGo's 'Move 37'

2016-03-10

DeepMind's AlphaGo makes the legendary 'Move 37' against world champion Lee Sedol, a move so creative it was initially mistaken for an error. This moment, which revealed a new form of machine intuition, is captured in the stunning AlphaGo documentary.

Google announces TPUs

2016-05-18

Google reveales custom Tensor Processing Units (TPUs) for accelerating ML in production, highlighting the role of specialized AI hardware.

Google Open-Sources TensorFlow

2015-11-09

Google releases its TensorFlow library, an open-source framework that makes deploying large-scale ML models dramatically easier. Its robust, production-first approach quickly establishes it as the dominant industry standard, defining an era of AI development and powering countless applications.

Generative Adversarial Networks

2014

Ian Goodfellow introduces GANs, enabling AI to create realistic data and transforming generative tasks.

Jupyter Notebooks Transform AI Workflow

2014-08-21

Spinning out of the popular IPython project, the Jupyter Notebook is introduced. The interactive web application for mixing live code, equations, and narrative text becomes the essential 'digital lab notebook' for a generation of data scientists and AI researchers. More background here.

word2vec popularizes embeddings

2013-01-16

Mikolov et al. released word2vec, showing efficient neural word embeddings that capture semantic regularities via vector arithmetic.

AlexNet Wins ImageNet

2012

Geoffrey Hinton's team wins the ImageNet competition with AlexNet, kickstarting the deep learning revolution in computer vision.

IBM Watson Wins Jeopardy

2011

Watson triumphs on Jeopardy!, showcasing advanced NLP and knowledge retrieval.

Siri Brings AI to the Masses

2011-10-04

Apple launches the iPhone 4S, which features Siri as its groundbreaking 'intelligent assistant'. The release brings conversational AI into the mainstream, making it the first truly mass-market digital assistant and putting the concept into millions of pockets.

Google Reveals Self-Driving Car Project

2010-10-09

Google announces its secret self-driving car project in a blog post titled 'What we're driving at'. The news that fully autonomous cars have already navigated complex California landmarks signals a new era, transforming the pursuit of driverless technology into a high-stakes race.

CUDA enables GPU computing

2007

NVIDIA’s CUDA platform made GPUs programmable for general-purpose computing, unlocking massive acceleration for ML and later deep learning.

Markdown is Created

2004-03-19

John Gruber, with contributions from Aaron Swartz, creates Markdown, a lightweight markup language for formatting plain text. Its simplicity leads to its adoption as the standard for documentation on platforms like GitHub and as the essential language for structuring the conversational output of nearly all modern LLMs.

1970 - 1999

AIBO Robot Dog

1999

Sony launches AIBO, a robotic pet demonstrating AI in consumer entertainment and companionship.

The Matrix Questions Reality

1999-03-31

The film The Matrix introduces a world where sentient machines pacify humanity within a sophisticated simulation. Its premise brings complex philosophical ideas like simulation theory and the nature of consciousness into the mainstream, profoundly influencing the cultural conversation about AI and virtual worlds.

Long Short-Term Memory

1997

Jürgen Schmidhuber introduces LSTM networks, solving vanishing gradient problems in RNNs for sequence learning.

Deep Blue Defeats Kasparov

1997-05-11

IBM's chess computer, Deep Blue, defeats reigning world champion Garry Kasparov in a dramatic six-game rematch. The victory marks the first time a machine overcomes a sitting world champion under standard tournament rules, becoming a monumental public milestone for artificial intelligence.

Python Becomes the Language of AI

1991-02-20

Guido van Rossum releases the Python programming language. Its emphasis on code readability and a simple syntax, combined with a powerful ecosystem of scientific libraries like NumPy and SciPy, eventually establishes it as the undisputed lingua franca for AI research and development. The name really comes from Monty Python.

Backpropagation Revitalizes Neural Networks

1986-10-09

In a landmark Nature paper, Rumelhart, Hinton, and Williams show that backpropagation can efficiently train multi-layer neural networks. By demonstrating that these 'deep' nets can learn complex internal representations, they revitalize the field after years of stagnation and set the stage for the future deep learning revolution.

The Terminator Introduces Skynet

1984-10-26

James Cameron's film The Terminator introduces Skynet, a self-aware defense network that perceives humanity as its primary threat and initiates a nuclear holocaust. The concept taps into our deepest anxieties about artificial general intelligence, becoming a defining trope for dystopian AI narratives.

Star Wars Droids Charm the World

1977-05-25

George Lucas's film Star Wars introduces C-3PO and R2-D2, captivating global audiences. Their distinct personalities and loyal friendship offer a powerful, friendly vision of AI, creating an enduring cultural counterpoint to the menacing HAL 9000.

Backpropagation Invented

1974

Paul Werbos proposes backpropagation in his thesis, laying groundwork for neural network training.

1763 - 1969

Perceptrons Book Triggers AI Winter

1969

Marvin Minsky and Seymour Papert publish Perceptrons, which mathematically proves the severe limitations of single-layer neural networks, famously showing they cannot solve the XOR problem. The book's devastating critique helps trigger a wave of pessimism, leading to drastic funding cuts and the start of the first 'AI Winter'.

HAL 9000 Defines AI in Pop Culture

1968-04-02

Stanley Kubrick's landmark film 2001: A Space Odyssey introduces the world to HAL 9000. The sentient AI's calm, conversational interface and chilling rebellion crystallize the public's hopes and fears about artificial general intelligence, becoming its definitive pop culture representation for decades.

ELIZA Chatbot

1966

Joseph Weizenbaum creates ELIZA, an early chatbot ('chatterbot') simulating conversation, revealing potential for human-AI interaction.

MIT AI Lab Founded

1959

Pioneers Marvin Minsky and John McCarthy establish the MIT Artificial Intelligence Laboratory. The lab becomes a world-renowned hub for AI research, producing groundbreaking work and nurturing generations of leading scientists in the field.

Perceptron Invented

1958-11

Psychologist Frank Rosenblatt develops the Perceptron, the first trainable neural network designed to simulate human learning. This machine learns by trial and error, establishing the foundational principles of pattern recognition that now underpin much of modern machine learning.

Dartmouth Conference

1956

The Dartmouth workshop birthed the field of AI, coining the term 'Artificial Intelligence' and setting an ambitious research agenda - this is the original proposal.

The Turing Test

1950-10

In his seminal paper 'Computing Machinery and Intelligence', Alan Turing proposes the 'Imitation Game', arguing a machine can be called intelligent if its text-based responses are indistinguishable from a human's. This concept, now known as the Turing Test, establishes the foundational benchmark for AI and launches the enduring debate on the nature of machine thought.

McCulloch-Pitts Neuron Model

1943

In their foundational paper, Warren McCulloch and Walter Pitts introduce the first mathematical model of a neuron. Their work demonstrates that networks of these simple binary units can perform logical computations, establishing the theoretical bedrock for artificial neural networks.

R.U.R. Introduces the 'Robot'

1921-01-25

Karel Čapek's play R.U.R. (Rossum's Universal Robots) premieres, introducing the word 'robot' to the global lexicon. Derived from the Czech word for forced labor, robota, the term was famously suggested to the author by his brother, Josef.

Frankenstein Published

1818-01-01

Mary Shelley publishes her novel Frankenstein; or, The Modern Prometheus, establishing a foundational myth for AI. The story of a sentient creation and its negligent creator raises enduring questions about the ethics and responsibilities of innovation.

Bayes' Theorem

1763

Thomas Bayes developed probabilistic reasoning, foundational for Bayesian AI methods.