AI History & Timeline

Key milestones from the birth of the field to today’s frontier models.

The Dawn of AI

1950 — Turing Test

Alan Turing proposes the ‘Imitation Game’ — a test for machine intelligence. His question: ‘Can machines think?’

1956 — AI Born

The Dartmouth Workshop coins the term ‘Artificial Intelligence’. John McCarthy, Marvin Minsky and others launch the field.

1966–1974 — First AI Winter

Early AI programs showed promise on toy problems but failed to scale. Funding dried up and progress stalled.

1986 — Backpropagation

Rumelhart, Hinton and Williams show backprop can train multi-layer neural networks — the foundation of modern deep learning.

1997 — Deep Blue

IBM’s Deep Blue defeats world chess champion Garry Kasparov — a landmark moment for AI capabilities.

2006 — Deep Learning Renaissance

Hinton and Salakhutdinov publish ‘Reducing the Dimensionality of Data with Neural Networks’ — sparking the deep learning revolution.

2012 — AlexNet

AlexNet’s GPU-trained CNN slashes image classification error rates — proving deep learning’s potential at scale.

2014 — GANs Invented

Ian Goodfellow invents Generative Adversarial Networks — the architecture behind modern image generation.

2017 — Attention Is All You Need

Google researchers introduce the Transformer architecture — the foundation of GPT, BERT, Claude and every major LLM today.

2018 — BERT & GPT-1

Google releases BERT; OpenAI releases GPT-1. The era of large pretrained language models begins.

2020 — GPT-3

OpenAI releases GPT-3 (175B parameters) — its few-shot capabilities shock the research community.

2021 — DALL-E & Codex

OpenAI releases DALL-E (text-to-image) and Codex (code generation) — two new modalities beyond text.

2022 — ChatGPT

OpenAI launches ChatGPT in November 2022. Reaches 100M users in 2 months — the fastest consumer product in history.

2023 — Frontier Model Race

GPT-4, Claude, Gemini. Frontier model race accelerates. Open models explode: Llama 2, Mistral 7B.

2024 — Multimodal + Reasoning

GPT-4o, Claude 3 Opus, Gemini 1.5 Pro. OpenAI o1 introduces deliberate ‘thinking’ (chain-of-thought at inference time).

2025 — Agentic AI + Open Surge

Claude Code, Cursor, Windsurf. DeepSeek R1 matches o1 at open weights. AI agents begin replacing multi-step human workflows.

2026 — Where We Are Now

Models with 1M+ context windows, real-time multimodal reasoning, and agentic systems that can autonomously complete complex tasks.