The first Machine Learning Workshop, the seed of ICML
A 1980 workshop at Carnegie Mellon became the series that grew into ICML, one of the three top machine-learning conferences.
XCON/R1, the first big commercial expert system
R1, later called XCON, was a rule-based expert system that configured DEC's VAX computer orders and became one of the first to pay off in business.
David Cope's EMI composes music in the style of the masters
David Cope began Experiments in Musical Intelligence, a program that analyzed a composer's catalog and wrote convincing new works in that style.
Hopfield networks revive interest in neural computation
John Hopfield's PNAS paper showed how networks of simple units can store memories as stable collective states, reenergizing the field.
Japan launches the Fifth Generation Computer Systems project
Japan MITI launched a ten-year national project in 1982, run by the ICOT institute, to build knowledge-processing computers based on logic programming.
DARPA launches the Strategic Computing program
In 1983 DARPA launched Strategic Computing, a billion-dollar plan for military machine intelligence: an autonomous tank, a pilot's associate, battle management.
Braitenberg's Vehicles
Braitenberg's 1984 thought experiments show simple sensor-motor machines producing behavior that looks like fear, aggression, and love.
Cyc begins, the common-sense knowledge project
Douglas Lenat began Cyc in 1984, an ambitious decades-long effort to hand-encode the everyday common-sense knowledge that AI systems lacked.
Racter and The Policeman's Beard Is Half-Constructed
A book of prose and poetry credited to the Racter program is published and billed as the first book written by a computer.
Rumelhart, Hinton, and Williams popularize backpropagation in Nature
The 1986 Nature paper made backpropagation the standard way to train multilayer neural networks with hidden units.
The second AI winter
In the late 1980s the commercial AI boom collapsed as the Lisp-machine market crashed and expert systems failed their hype, starting a long downturn near 1987.
LeCun applies backpropagation to handwritten zip code recognition
LeCun and colleagues at Bell Labs trained a convolutional network on real US Postal Service digits, a landmark for practical deep learning.
IBM reframes translation as statistics, not rules
In 1990 an IBM team published a statistical approach to machine translation, treating language as a problem of probability and data, not grammar rules.
Continuous Interleaved Sampling Cochlear Implants
A 1991 Nature paper introduced the CIS sound-processing strategy that made cochlear implants deliver high speech recognition.
First Loebner Prize contest
The first Loebner Prize held a public Turing-test contest in Boston, the longest-running attempt to stage Turing's imitation game as a competition.
TD-Gammon teaches itself backgammon
Gerald Tesauro's neural network learned backgammon at near-champion level through self-play using temporal-difference learning.
Vinge: The Coming Technological Singularity
Vernor Vinge's 1993 essay predicting that creating superhuman intelligence would soon end the human era, popularizing the singularity.
NVIDIA is founded to build 3D graphics chips
On April 5, 1993, Jensen Huang, Chris Malachowsky and Curtis Priem founded NVIDIA to bring 3D graphics to PCs - the company now central to AI compute.
The FERET program standardizes face recognition testing
Starting in 1993 the US FERET program built a shared face database and independent tests that turned face recognition into a measurable science.
START question answering system goes on the web
MIT's START, built by Boris Katz, connects to the web and answers plain-English questions, a precursor to modern QA and assistants.
GroupLens and the first automated collaborative filtering system
In 1994 GroupLens used ratings from like-minded readers to predict which Usenet articles a person would enjoy.
No Hands Across America (1995)
CMU researchers drove a vision-steered minivan from Pittsburgh to San Diego in 1995, autonomous for over 98 percent of 2,850 miles.
Support-vector networks: the method that out-ran neural nets
Cortes and Vapnik's 1995 support-vector network became the dominant classifier of the 1990s and 2000s, eclipsing neural nets until the deep-learning revival.
RoboCup founded
RoboCup launched robot soccer as a grand-challenge benchmark, with the stated goal of beating the human World Cup champions by 2050.
Deep Blue defeats world chess champion Kasparov
IBM's Deep Blue became the first computer to beat a reigning world chess champion in a full match, defeating Garry Kasparov 3.5 to 2.5.
Logistello sweeps the Othello world champion
In August 1997 Michael Buro's program Logistello beat Othello world champion Takeshi Murakami 6-0, ending human supremacy at the game.
Long Short-Term Memory networks introduced
Hochreiter and Schmidhuber publish LSTM, a recurrent network that remembers across long gaps, solving the vanishing-gradient problem.
The first International Planning Competition
The 1998 International Planning Competition introduced PDDL and a shared benchmark, turning AI planning into an empirically measured field.
LeNet-5 and gradient-based document recognition
LeCun and colleagues publish the LeNet-5 convolutional network, showing end-to-end gradient learning for reading handwritten documents.
Remote Agent takes control of Deep Space 1
In May 1999 NASA's Remote Agent became the first AI software to autonomously command a spacecraft in deep space.
NVIDIA ships the GeForce 256, the first chip called a GPU
In 1999 NVIDIA shipped the GeForce 256 and coined the term GPU, the parallel chip architecture that would later power deep learning.
Kismet and the birth of social robotics
MIT's Kismet, a robotic head built by Cynthia Breazeal, used expressive faces and voice to interact socially, helping launch the field of social robotics.
OpenCV, the open-source computer vision library, begins at Intel
OpenCV launched in 2000 as an Intel research effort and grew into the world's most widely used open-source computer vision library.
MIRI is founded (as the Singularity Institute)
The Machine Intelligence Research Institute, one of the first organizations focused on AI alignment, was founded in 2000 by Eliezer Yudkowsky.
Honda unveils ASIMO, the walking humanoid robot
On November 20, 2000, Honda unveiled ASIMO, a 120 cm humanoid robot whose predictive walking control gave it an unprecedented human-like gait.
Black & White ships a creature that learns from the player
Lionhead's Black & White ships an AI creature that learns its behavior from watching and being corrected by the player.
Hawk-Eye ball tracking debuts in cricket
Hawk-Eye's camera-based ball tracking first aired during an England-Pakistan Test at Lord's in April 2001.
Berners-Lee describes the Semantic Web
A 2001 Scientific American article by Berners-Lee, Hendler, and Lassila laid out the Semantic Web, a vision of web data that machines could understand.
SmarterChild, the instant-messenger chatbot millions chatted with for fun
In 2001 SmarterChild launched on AOL Instant Messenger and grew to over 17 million users who mostly just chatted with it for fun.
Viola-Jones real-time face detection
The 2001 Viola-Jones detector made real-time face detection practical and became the technology behind the autofocus box in digital cameras.
iRobot launches the Roomba and autonomy enters the home
iRobot's Roomba put a small, self-navigating robot in millions of households, the first truly mass-market consumer robot.
Amazon's item-to-item collaborative filtering
Amazon's 2003 item-to-item algorithm scaled recommendations to millions of products by relating items, not users.
BigDog, Boston Dynamics' first legged robot to leave the lab
Boston Dynamics' BigDog, a DARPA-funded rough-terrain quadruped, was the company's first legged robot able to balance and walk outside the lab.
CIFAR Funds Deep Learning Through the Winter (2004)
In 2004 Canada's CIFAR launched a program directed by Geoffrey Hinton that sustained Hinton, Bengio, and LeCun through the neural-network winter.
DARPA Grand Challenge 2004 - the race nobody finished
DARPA's first Grand Challenge offered $1M for an autonomous desert drive; no vehicle finished and the best managed 7.5 miles.
General Game Playing competition launched
Stanford's AAAI competition challenged programs to play any game from its rules alone, rewarding general skill over hand-built game-specific engines.
Damian Isla presents the Halo 2 behavior-tree AI
Bungie's Damian Isla details the Halo 2 behavior-tree architecture at GDC 2005, popularizing the technique in games.
Stanley wins the DARPA Grand Challenge
Stanford's self-driving car Stanley completed a 132-mile desert course to win the 2005 DARPA Grand Challenge, a turning point for autonomous vehicles.