History
A timeline of AI.
Not every announcement — the results that changed what was possible. 28 of them, from the artificial neuron in 1943 to the moment compute moved to inference. Each links to the concept behind it.
1943 – 1956
Before the field had a name
The idea that thinking might be mechanised arrived before the machines could do anything with it. This period sets the questions the field is still arguing about.
1943#
McCulloch and Pitts describe a mathematical model of a neuron — inputs, a threshold, an output. It is a drastic simplification of biology and it turns out to be enough. Every neural network since is built from descendants of this unit. Worth noting how early this is: the idea that thinking might be computation predates the computers to test it on.
1950#
Computing Machinery and Intelligence opens by asking whether machines can think and immediately abandons the question as too vague to answer. Turing proposes judging behaviour instead — if it converses indistinguishably from a person, that's the evidence we get. It's the most influential dodge in computer science, and it set the field's habit of measuring what systems do rather than settling what intelligence is.
1956#
The Dartmouth workshop proposal coins the term artificial intelligence and suggests that a summer's work by ten people might make significant progress on language, abstraction and self-improvement. It's worth reading now for the confidence. That gap — between what looked close and what proved hard — is the field's oldest and most repeated pattern.
1958 – 1980
The first wave, and the first winter
Early results were genuinely exciting, and the promises made about them were not. What followed is the field's most instructive failure — and it happened twice.
1958#
Rosenblatt builds a machine that learns to classify by adjusting weights from examples — the first working neural network, and the direct ancestor of everything modern. The press coverage promised machines that would walk, talk and reproduce. The actual device could separate simple patterns. Both facts matter: the technique was real, and the claims made about it were not.
1969#
Minsky and Papert prove that a single-layer perceptron cannot learn XOR — a limitation real and narrow, since layered networks can. But the result landed as a verdict on the whole approach, funding moved elsewhere, and neural network research went quiet for over a decade. A cautionary tale about how a specific technical finding becomes a general mood.
1973#
A review commissioned by the UK government concludes that AI had failed to deliver on its promises, and British funding is largely withdrawn. Similar cuts follow elsewhere. The first AI winter isn't caused by the technology failing — it's caused by the technology failing to match what had been claimed for it. That distinction is the whole lesson.
1986 – 2009
Quiet progress
The algorithms that power everything today were mostly invented in this stretch, and mostly ignored. What was missing wasn't ideas. It was data and hardware.
1986#
Rumelhart, Hinton and Williams show how to assign blame across the layers of a network — trace the error backwards, work out how much each weight contributed, adjust accordingly. Backpropagation makes deep networks trainable, and every model running today still uses it. The algorithm was in place decades before the data and hardware that would make it matter.
1989#
LeCun and colleagues apply convolutional networks to handwritten digits, and the system goes into production reading cheques. It's the first commercially significant deep learning system, and it works because the architecture encodes a real fact about images: nearby pixels are related, and a pattern means the same thing wherever it appears.
1997#
Chess falls to a machine, and the reaction establishes a pattern the field has repeated ever since. Before: chess was the benchmark of machine intelligence. After: it was 'just' search and evaluation, and the real test moved elsewhere. The goalposts didn't move dishonestly — people genuinely learned that chess needed less than they'd assumed.
1997#
Hochreiter and Schmidhuber design a recurrent unit that can hold information across long sequences, working around the vanishing gradient problem that made earlier recurrent networks forget. LSTMs dominate sequence modelling for nearly two decades — until attention makes recurrence itself optional.
2009#
Fei-Fei Li's team publishes a labelled image dataset of unprecedented scale. It isn't an algorithm, and it changes more than most algorithms did. The lesson the field took years to absorb: the bottleneck had been data, not ideas. Nearly everything in the next decade runs on this insight.
2012 – 2016
The deep learning era
Data and hardware arrived. The results that followed were fast enough that the field's own researchers were repeatedly surprised.
2012#
A convolutional network wins the ImageNet competition by a margin wide enough to end the argument, trained on two consumer gaming GPUs. This is the result that starts the deep learning era — not because the technique was new (it wasn't), but because the data and the hardware had finally arrived to make a thirty-year-old idea work.
2013#
Mikolov's team shows that words can be mapped to vectors where distance means similarity and directions carry meaning — the origin of king − man + woman ≈ queen, and of embeddings generally. Everything that searches by meaning rather than by keyword descends from this.
2014#
Goodfellow's setup pits a generator against a discriminator — one faking, one detecting, both improving by losing to each other. GANs make machine-generated images convincing for the first time and dominate for years, before diffusion largely displaces them. The training instability that plagued them is the reason.
2015#
He and colleagues add shortcut connections so gradients have a path back that isn't multiplied down through every layer. Suddenly networks can be hundreds of layers deep instead of dozens. It's a small architectural idea that removes the ceiling on everything else.
2016#
Go was supposed to be a decade away — too many positions to search, too much dependent on intuition. Reinforcement learning combined with search closes it, and move 37 of game two is still discussed because it was a move no human would play and it was correct. Then the goalposts moved again.
2017 – 2021
The transformer era
One architecture displaced most of the others, and then scale turned out to matter more than anyone had planned for.
2017#
Vaswani and colleagues drop recurrence entirely. Instead of reading a sentence word by word, the transformer weighs every word against every other word at once — and because that's parallel, it trains on hardware that recurrence couldn't exploit. Nearly every significant model since is built on this architecture. It's the single most consequential paper of the era.
2018#
BERT and the first GPT establish the recipe that still holds: train on an enormous general corpus, then adapt to your task afterwards. Transfer learning stops being a technique and becomes the assumption. This is why small teams can build things — you inherit the expensive part.
2020#
Brown and colleagues show that making models bigger doesn't just improve them — it changes what they can do. Examples in the prompt start substituting for training, which nobody designed for. Kaplan's scaling laws arrive the same year and turn 'make it bigger' from a hunch into an equation. Large language models become a category.
2020#
Lewis and colleagues formalise fetching documents at query time and letting the model answer from them, rather than baking knowledge into weights. RAG becomes the standard answer to 'the model doesn't know our data' — and remains the most commonly misdiagnosed system in production.
2021#
Radford's team trains on image–caption pairs until pictures and words share one embedding space. That alignment is the groundwork under multimodal systems — and under text-to-image generation, which needs a way to connect a prompt to a picture.
2021#
Protein structure prediction — a fifty-year-old problem in biology — is substantially solved by a deep learning system. It's the clearest evidence to date that these methods do real scientific work rather than parlour tricks, and it's the result most likely to still matter in fifty years.
2022 – 2024
The public era
The technology stopped being a research topic and became a product category. The underlying models changed less in this period than the interfaces did.
2022#
Latent diffusion moves the denoising process into a compressed space, and suddenly image generation runs on consumer hardware. Within months it's in public hands. The copyright and consent arguments that follow are still unresolved.
2022#
Ouyang and colleagues train on human preferences instead of correct answers — people compare two outputs and say which is better, and the model learns to produce the preferred kind. RLHF is the step that turns a text predictor into something that answers your question. Underrated, and the reason the next entry landed the way it did.
2022#
The underlying model was not dramatically new. The interface was — a text box, no instructions needed, free. It reaches a hundred million users faster than anything before it, and AI stops being a topic and becomes a product category. The lesson most of the industry drew was about models. The lesson available was about interfaces.
2023#
Llama 2 makes capable models downloadable under a licence that permits most commercial use, and an ecosystem forms around running models yourself. The argument about what open means — weights without training data isn't open source — starts here and hasn't finished.
2023#
ReAct and tool use turn models into systems that act rather than answer. The demos are extraordinary and the production record is not, and the gap between them becomes the field's most expensive lesson: a system that's 90% reliable per step is a coin flip after seven steps.
2024#
Models that generate reasoning before answering shift spending from training time to serving time. It's a genuine change in the economics — capability becomes something you buy per request rather than something the model has — and it quietly retires the assumption that inference is the cheap part.
Common questions
When did artificial intelligence start?
As a named field, 1956 — the Dartmouth workshop coined the term. But the underlying ideas are older: McCulloch and Pitts modelled an artificial neuron in 1943, and Turing's paper asking whether machines can think was published in 1950.
Who invented AI?
No single person. The term was coined at the 1956 Dartmouth workshop organised by John McCarthy, Marvin Minsky, Nathaniel Rochester and Claude Shannon. The intellectual groundwork was laid earlier by Turing, McCulloch and Pitts among others.
What was the AI winter?
Periods when funding and interest collapsed — roughly 1974–1980 and 1987–1993. The trigger wasn't the technology failing outright. It was the technology failing to match what had been promised for it, most visibly after the 1973 Lighthill report led the UK to withdraw funding.
When did deep learning start working?
2012, with AlexNet winning the ImageNet competition by a wide margin. The techniques were decades old — backpropagation dates to 1986, convolutional networks to 1989. What changed was the arrival of large labelled datasets and GPUs fast enough to train on them.
When was ChatGPT released?
November 2022. Its underlying model was not a dramatic advance on what existed; the significant change was the interface and free public access, which took it to roughly a hundred million users faster than any consumer product before it.
What is the most important paper in modern AI?
Attention Is All You Need (Vaswani et al., 2017), which introduced the transformer. It removed recurrence from sequence modelling, which made training parallel and therefore scalable. Nearly every significant model since is built on that architecture.
Why do AI milestones keep 'not counting' afterwards?
Because intelligence has never had an agreed definition, so the frontier defines it. Chess was the benchmark until Deep Blue won in 1997, at which point it became 'just search'. The same happened with Go in 2016. Each time, people learned the task needed less than they'd assumed.
This stops short of the present on purpose. Recent releases are covered in the blog, where a piece carries a date and is allowed to age. A timeline should be the part that has settled — and the last few years haven't. If you want the underlying ideas rather than the chronology, start with the map.