AINews Portal

🕰️ History of AI

Seventy-five years from the Imitation Game to humanoid robots — every milestone that mattered, and why, in one line each.

  1. 1950

    Turing proposes the Imitation Game

    Frames "can machines think?" as a testable question.

  2. 1956

    Dartmouth workshop coins “artificial intelligence”

    The field gets its name and founding agenda.

  3. 1958

    Rosenblatt builds the Perceptron

    First trainable neural network — and first AI hype cycle.

  4. 1969

    Minsky & Papert publish “Perceptrons”

    Exposes single-layer limits; helps trigger the first AI winter.

  5. 1986

    Backpropagation popularized (Rumelhart, Hinton, Williams)

    Makes training multi-layer networks practical — the engine of everything since.

  6. 1997

    Deep Blue defeats Kasparov

    A machine beats the world chess champion; search + compute wins.

  7. 1998

    LeCun’s LeNet reads checks with CNNs

    Convolutional networks prove themselves in production.

  8. 2012

    AlexNet wins ImageNet

    GPUs + deep CNNs crush the field — the deep-learning era begins.

  9. 2014

    GANs invented (Goodfellow)

    Machines learn to generate; seeds today’s image generation.

  10. 2015

    OpenAI founded; DeepMind’s DQN masters Atari

    Reinforcement learning + the lab that would ship ChatGPT.

  11. 2016

    AlphaGo defeats Lee Sedol

    Intuition-like play in Go, a decade ahead of expectations.

  12. 2017

    “Attention Is All You Need” — the transformer

    The architecture behind every modern frontier model.

  13. 2018

    BERT and GPT-1

    Pretraining on raw text becomes the dominant paradigm.

  14. 2020

    GPT-3 (175B)

    Scale produces few-shot abilities nobody explicitly programmed.

  15. 2021

    GitHub Copilot; DALL·E

    AI starts writing code and making images for normal users.

  16. 2022

    Stable Diffusion open-sourced; ChatGPT launches

    Image gen for everyone, then the fastest-growing app in history.

  17. 2023

    GPT-4, Claude, LLaMA leaks/open weights

    Frontier reasoning arrives; the open-weights movement explodes.

  18. 2024

    Multimodal frontier models; o1 reasoning; Nobel Prizes for AI

    Models see/hear/speak, think step-by-step; Hinton & Hassabis honored.

  19. 2025

    Agents go mainstream — coding agents, MCP, computer use

    AI stops just answering and starts doing multi-step work.

  20. 2026

    Reasoning models commoditize; video generation matures; humanoids ship

    Frontier capability spreads to open models; AI enters the physical world.

For what happened this month instead of this century, see Model Releases and the Heat Board.

History of AI — timeline of every major milestone · AI News Portal