Amazon launches Mechanical Turk, a marketplace for human microtasks
Amazon launched Mechanical Turk in 2005, an API that lets software farm out small tasks to human workers it billed as 'Artificial Artificial Intelligence'.
F.E.A.R. brings planning AI to shooters with GOAP
Monolith's F.E.A.R. ships Goal-Oriented Action Planning, giving shooter enemies a three-state FSM that plans with A*.
Google Translate goes live with statistical machine translation
Google launches a free online translator built on statistics rather than hand-coded rules, starting with Arabic and English.
Deep belief nets relaunch deep learning
Hinton, Osindero, and Teh show a fast greedy way to train deep networks layer by layer, reviving interest in deep learning.
Facebook launches the News Feed
Facebook's 2006 News Feed turned a static profile site into an algorithmic stream of friends' activity, sparking a user revolt.
Netflix offers $1 million to beat its recommender
Netflix launches a $1 million open contest to improve its movie recommendation system by 10 percent, mainstreaming the public machine learning competition.
NVIDIA releases CUDA
NVIDIA introduced CUDA, a platform that let developers run general-purpose code on GPUs, providing the compute engine that later powered deep learning.
Moses, the open-source statistical machine translation toolkit
Moses, released in 2007, gave any researcher or company a free state-of-the-art statistical machine translation system to train and run.
Checkers is solved
After 18 years of computation, Jonathan Schaeffer's team proved that checkers is a draw with perfect play, the most complex game solved at that time.
Boss wins the DARPA Urban Challenge
In 2007 Carnegie Mellon's Boss drove itself through mock city traffic to win the DARPA Urban Challenge, proving autonomy could handle real driving.
reCAPTCHA turns spam-blocking into book digitization
Luis von Ahn's reCAPTCHA, published in Science in 2008, used everyday CAPTCHA tests to crowd-transcribe scanned books at over 99 percent word accuracy.
Cleverbot launches as a learning web chatbot
Rollo Carpenter releases Cleverbot, a web chatbot that replies by reusing how people have talked to it before.
PARO, the therapeutic seal robot, is cleared as a US medical device
In 2009 US regulators recognized PARO, a sensor-laden robotic baby seal, as a neurological therapeutic device for dementia and elder care.
India launches Aadhaar, the world's largest biometric ID
India's Aadhaar program, begun in 2009, enrolled over a billion residents using fingerprints, iris scans, and a photo.
The unreasonable effectiveness of data
Three Google researchers argue that simple models on enormous data beat elaborate models on less - a 2009 data-first creed that predates the scaling laws.
ROS gives robotics a common open-source operating system
ROS, the open-source Robot Operating System from Willow Garage and Stanford, gave robotics a shared software platform first described in 2009.
ImageNet, a large-scale image database
Fei-Fei Li and colleagues introduce ImageNet, a labeled image database that became the proving ground for modern computer vision.
Kaggle launches
Kaggle turned machine learning into a competitive discipline, hosting public prediction contests that became a benchmark and a talent pipeline for the field.
The 2010 Flash Crash and the SEC/CFTC joint report
On May 6, 2010 US markets briefly collapsed and recovered within minutes, set off by an automated futures-selling algorithm.
Betterment launches and the robo-advisor era begins
Betterment debuted at TechCrunch Disrupt in May 2010, automating portfolio investing and helping start the robo-advisor industry.
Google reveals its self-driving car project
Google publicly disclosed that it had been quietly building self-driving cars that had already logged 140,000 miles on real roads.
IBM Watson wins Jeopardy!
IBM Watson question-answering system beat Jeopardy! champions Ken Jennings and Brad Rutter in 2011, showing computers could answer natural-language questions.
Apple puts a talking assistant in millions of pockets
Apple launches Siri with the iPhone 4S, putting a conversational voice assistant into the hands of millions of consumers.
Stanford's free online AI course launches the MOOC era
In fall 2011 Sebastian Thrun and Peter Norvig taught an online Stanford AI course to 160,000 students, seeding Udacity and Coursera.
Da Silva Moore: the first court approval of predictive coding
Magistrate Judge Andrew Peck issued the first US judicial opinion approving computer-assisted review for e-discovery.
Amazon acquires warehouse-robot maker Kiva Systems
In 2012 Amazon bought Kiva Systems for about $775 million, bringing fleets of mobile robots into its fulfillment centers.
Coursera founded, bringing universities onto one MOOC platform
Stanford professors Andrew Ng and Daphne Koller launched Coursera in 2012 with Princeton, Stanford, Michigan, and Penn as founding partners.
Google Launches the Knowledge Graph
Google introduced the Knowledge Graph, shifting search from matching keyword strings toward understanding real-world entities.
AlexNet wins ImageNet 2012
A deep convolutional neural network crushed the ImageNet contest, proving deep learning could outperform hand-built computer vision.
word2vec makes word embeddings practical
Mikolov and colleagues introduced word2vec, an efficient way to learn word vectors that capture meaning through geometry.
The Campaign to Stop Killer Robots launches
In April 2013 a coalition of NGOs launched a campaign calling for a pre-emptive ban on fully autonomous weapons that attack without human control.
ICLR is founded with open peer review
The first ICLR, founded by Bengio and LeCun in May 2013, gave deep learning its own conference and helped pioneer open, public peer review.
DARPA unveils Atlas, the disaster-response humanoid
DARPA and Boston Dynamics unveiled Atlas, a six-foot hydraulic humanoid robot built as a common platform for the DARPA Robotics Challenge.
Facebook AI Research (FAIR) is founded
In December 2013 Facebook created FAIR under Yann LeCun, building a major open-research deep-learning lab inside a consumer tech company.
Deep Q-Networks learn Atari from pixels
DeepMind trained a single neural network to play Atari video games directly from raw pixels, beating humans on several titles.
Alan Turing receives a posthumous royal pardon
On 24 December 2013 the Queen granted Alan Turing a posthumous royal pardon, 59 years after his conviction for gross indecency drove him to suicide.
Google acquires DeepMind
Google buys the London AI lab DeepMind in January 2014, placing a leading research lab inside a big-tech company.
Quakebot writes the LA Times earthquake story in minutes
An LA Times algorithm called Quakebot turned a USGS alert into a published earthquake story within minutes, an early case of automated newswriting.
Microsoft unveils Cortana
Microsoft reveals Cortana, its Bing-powered voice assistant for Windows Phone, with a Notebook that stores what the user cares about.
Microsoft launches XiaoIce, the social chatbot that became a companion to millions
In 2014 Microsoft released XiaoIce in China, an empathetic chatbot designed for companionship that grew to hundreds of millions of users.
Generative Adversarial Networks introduced
Ian Goodfellow and colleagues proposed pitting two neural networks against each other, launching the era of AI-generated images.
Microsoft Project Catapult puts FPGAs in the datacenter
Microsoft showed FPGAs deployed across 1,632 servers could speed Bing ranking by 95 percent - an early bet on specialized AI hardware beyond CPUs.
AP automates corporate earnings stories with Automated Insights
The Associated Press began using Automated Insights' Wordsmith software to write thousands of quarterly earnings stories automatically.
Bostrom's Superintelligence
Nick Bostrom's 2014 book Superintelligence brought existential risk from advanced AI to a wide audience and shaped how lab founders talk about safety.
IBM unveils TrueNorth, a million-neuron brain-inspired chip
In 2014 IBM revealed TrueNorth, a neuromorphic chip with a million spiking neurons and 256 million synapses running on 70 milliwatts.
NVIDIA releases cuDNN, the deep learning library under the frameworks
In 2014 NVIDIA released cuDNN, a library of GPU primitives for neural networks that became the engine beneath PyTorch, TensorFlow and Caffe.
Sequence to sequence learning
Sutskever, Vinyals, and Le showed neural networks could map input sequences to output sequences, enabling end-to-end translation.
Amazon introduces Echo and the Alexa assistant
Amazon introduces the Echo smart speaker and its voice assistant Alexa, creating the always-listening ambient assistant category for the home.