Prehistory

2010–20158 events

Editors' summary

Deep learning starts to work, and the talent gathers at the big companies

In September 2010, Demis Hassabis and others founded DeepMind in London, declaring they would build general intelligence and working on reinforcement learning through games. The following year Google Brain was formed inside Google, running large neural networks on the company's own compute.

In 2012, AlexNet won the ILSVRC image recognition contest by a wide margin. Deep learning had reached a usable standard, and the direction of research turned. In March 2013, Google acquired DNNresearch, the company of Geoffrey Hinton and his students; in January 2014 it acquired DeepMind. The competition to take on people by buying their companies had begun.

In February 2015, DeepMind's DQN appeared in Nature, playing Atari games at human level from the screen alone. In November, Google open-sourced TensorFlow, putting the tools of deep learning within anyone's reach.

On 11 December 2015, OpenAI was founded. Elon Musk, Sam Altman and others pledged $1 billion, and the organisation was set up as a non-profit that would publish its results openly — so that no single company would hold intelligence to itself.

Microsoft announced Cortana in April 2014, an attempt at a voice assistant across Windows and phones.

This block is written by the editors. It is kept separate from the sourced record below.

Record8 events

Sep 23
BusinessGoogle / Demis Hassabis / Shane Legg / Mustafa Suleyman
DeepMind is founded in London
DeepMind Technologies, which would become the core of Google's AI research, was founded in London by neuroscientist and former chess prodigy Demis Hassabis, Shane Legg, and Mustafa Suleyman. It was incorporated on 23 September 2010 under the name Friars 2022 Limited, renamed DeepMind on 15 November 2010 and formally launched. The company set out to combine neuroscience and machine learning in pursuit of general-purpose intelligence. Google acquired it in 2014, and in 2023 it merged with Google Brain to form Google DeepMind.
date unknown
ResearchGoogle / Andrew Ng / Jeff Dean / Greg Corrado
The Google Brain project begins
Google Brain, the company's large-scale neural network research team, got its start. Stanford professor Andrew Ng had begun what he called the Deep Learning Project at Google, joined by Google fellow Jeff Dean and researcher Greg Corrado; much of the early work was part-time, done in the 20 percent time Google allowed. It began inside Google X and was graduated back into Google proper after proving successful. In June 2012 the team published "Building high-level features using large scale unsupervised learning," in which a network spanning 16,000 processors, trained on 10 million images drawn from YouTube without labels, developed features that detected cats. The New York Times headline — "How Many Computers to Identify a Cat? 16,000" — helped bring deep learning to public attention. Google Brain went on to produce TensorFlow and the Transformer, and merged with DeepMind in 2023.
Mar 12
BusinessGoogle / Geoffrey Hinton / Alex Krizhevsky / Ilya Sutskever
Google acquires Hinton's DNNresearch
Google acquired DNNresearch, the company Geoffrey Hinton had incorporated in 2012 with his University of Toronto graduate students Alex Krizhevsky and Ilya Sutskever; terms were not disclosed. The three were the people who had dramatically raised the state of the art in object recognition with convolutional neural networks at the previous year's ImageNet competition, and Google took in both the research and the talent behind it. Krizhevsky and Sutskever joined Google, while Hinton split his time between the university and the company. Google bought the company built around the researchers rather than rights to the research itself. Sutskever left Google in 2015 to co-found OpenAI.
Jan 26
BusinessGoogle / Demis Hassabis / Shane Legg / Mustafa Suleyman
Google acquires DeepMind
Google disclosed its acquisition of DeepMind Technologies, the London AI research company. Terms were not officially released, but the price was reported at more than $500 million, with some accounts putting it between $400 and $600 million; Google CEO Larry Page was said to have led the negotiations. DeepMind had no product, but its work training deep reinforcement learning agents to play Atari games had drawn notice. Reporting at the time said Google agreed as part of the deal to establish an ethics board to guard against misuse of AI technology. The lab retained substantial independence and stayed in London, going on to produce AlphaGo in 2016 and AlphaFold2 in 2020. In 2023 it merged with Google Brain to become Google DeepMind, the group behind Gemini.
Apr 2
ProductMicrosoft / Satya Nadella / Joe Belfiore
Microsoft introduces Cortana
At its Build developer conference, Microsoft introduced Cortana, a voice-driven digital assistant shipping with Windows Phone 8.1. Built on the Bing search engine, it was designed to learn from a user's behavior and ask questions before assuming what they wanted; the name came from the AI character in the game Halo. Following Apple's Siri and Google's voice assistant, it represents the era of talk-to-it AI before generative models. Cortana was later wound down, and from 2023 Microsoft's assistant strategy was replaced by Copilot.
Feb 26
ResearchGoogle / Volodymyr Mnih / David Silver / Koray Kavukcuoglu
DeepMind's DQN reaches human-level play on Atari
Google DeepMind published "Human-level control through deep reinforcement learning" in Nature. Given only screen pixels and the game score as input, and using the same algorithm, network architecture, and hyperparameters throughout, a deep Q-network learned to play 49 different Atari 2600 games, reaching a level comparable to a professional human tester on many of them. The paper showed that combining deep learning with reinforcement learning worked in practice, and it demonstrated publicly why Google had bought DeepMind. The same deep reinforcement learning framework was used in AlphaGo the following year.
Nov 9
ProductGoogle / Jeff Dean / Rajat Monga
Google open-sources TensorFlow
Google open-sourced TensorFlow, its machine learning framework, under the Apache 2.0 license. Announced by Jeff Dean and Rajat Monga, the release opened up the second-generation machine learning system Google had been using internally, along with its library, tools, tutorials, and examples. It improved on the previous DistBelief system in speed, scalability, and production readiness — twice as fast on some benchmarks — and was the same technology behind image search in Google Photos and Smart Reply in Gmail. The release decisively established the practice of sharing research as working implementations rather than papers alone, and vastly widened access to deep learning. PyTorch would later contest its lead, but at this point TensorFlow became the industry standard.
Dec 11
BusinessOpenAI / Sam Altman / Elon Musk / Ilya Sutskever
OpenAI founded
OpenAI was announced as a non-profit AI research company, with Sam Altman and Elon Musk as co-chairs, Ilya Sutskever as research director, and Greg Brockman as CTO. Backers pledged a total of $1 billion, with the stated mission of developing AI for the benefit of humanity as a whole.