Timeline of AGI breakthroughs
From Turing's 1950 paper to the agentic systems of 2026 — the seventeen moments that most shaped the AGI trajectory.
Progress in AI has not been smooth. Long winters punctuated by sudden capability jumps are the rule, not the exception. This timeline collapses seventy-six years of research into the seventeen moments that most clearly changed what experts believed was possible. Each is a stepping stone, not a milestone in isolation.
- 1950
Turing's Computing Machinery and Intelligence
Alan Turing proposes the Imitation Game as an operational test of machine intelligence.
- 1956
Dartmouth Workshop
The summer workshop at Dartmouth coins the term 'artificial intelligence' and launches the field as a formal research programme.
- 1969
Perceptrons
Minsky and Papert publish Perceptrons, exposing the limits of single-layer networks and ushering in the first AI winter.
- 1986
Backpropagation popularised
Rumelhart, Hinton, and Williams popularise backpropagation, reviving neural-network research.
- 1997
Deep Blue beats Kasparov
IBM's Deep Blue defeats world chess champion Garry Kasparov, the first reigning champion to lose a match to a machine.
- 2012
AlexNet
AlexNet wins ImageNet by a wide margin, triggering the modern deep-learning era.
- 2014
GANs and seq2seq
Generative adversarial networks and sequence-to-sequence learning are introduced, expanding what neural networks can model.
- 2016
AlphaGo beats Lee Sedol
DeepMind's AlphaGo defeats Lee Sedol 4–1, a decade earlier than most experts predicted.
- 2017
Transformer architecture
Vaswani and colleagues publish 'Attention Is All You Need', introducing the Transformer.
- 2018
BERT and GPT-1
Pre-training plus fine-tuning becomes the dominant NLP paradigm.
- 2020
GPT-3 and scaling laws
GPT-3 demonstrates broad few-shot capability and the Kaplan scaling laws formalise predictable returns to scale.
- 2021
AlphaFold 2
AlphaFold 2 solves the protein-folding problem to near-experimental accuracy.
- 2022
ChatGPT
ChatGPT brings conversational LLMs to a global audience and reaches 100 million users in two months.
- 2023
GPT-4, Gemini, Claude
The first generation of widely deployed multimodal frontier models ships across three labs in one year.
- 2024
Reasoning models
OpenAI's o1 launches explicit reasoning-time compute as a new scaling axis; DeepMind's AlphaProof and AlphaGeometry 2 reach IMO silver-medal performance.
- 2025
International AI Safety Report
The first International AI Safety Report, chaired by Yoshua Bengio, is published with backing from 30 countries.
- 2026
Agentic frontier
Frontier labs ship long-horizon agentic systems and the first generation of state-run pre-deployment evaluations becomes routine.
For deeper context on individual entries, see the landmark papers list or the AGI Timeline Explorer tool for forward-looking scenarios.