Capability threshold
A machine did something previously beyond machines.
60 milestones.
BrainGate researchers decode imagined handwriting from neural signals to enable high-speed text communication
In May 2021, Francis R. Willett and colleagues at Stanford University and Howard Hughes Medical Institute published results showing that a BrainGate2 intracortical electrode array could decode imagined handwriting movements in a paralysed person at 90 characters per minute with 94.1% raw accuracy, substantially exceeding prior neural-interface typing rates.
Cornell University researchers demonstrate electrically actuated microscale origami robots with onboard CMOS control
On 17 March 2021, researchers at Cornell University published a demonstration of self-folding microscale robots, roughly 100–250 microns in size, driven by platinum-based shape-memory actuators and controlled by onboard complementary metal-oxide-semiconductor (CMOS) circuits, enabling untethered, electrically commanded origami-style locomotion at the micron scale.
Google Announces TPU v4 Tensor Processing Unit
Google announced its fourth-generation Tensor Processing Unit (TPU v4) at Google I/O in 2021. The chip, designed specifically for large-scale machine-learning workloads, offered substantially higher performance than its predecessor and was made available to researchers via Google Cloud.
Cavatappi Artificial Muscles from Drawing, Twisting, and Coiling Polymer Tubes
In 2021, researchers led by Tzu-Ching Shyu at the University of Illinois at Urbana-Champaign published findings in Science Robotics describing cavatappi artificial muscles, twisted polymer actuators made from polyethylene-based tubes, capable of contracting up to 45% of their length, lifting approximately 300 times their weight per unit volume, and actuating within 23 milliseconds per 1% length change.
Swinburne University of Technology Researchers Demonstrate Optical Neuromorphic Processor Using Micro-Comb Photonic Chip
In January 2021, a Swinburne University of Technology-led international team published results in Nature demonstrating an optical neuromorphic processor built on a photonic micro-comb chip, achieving a processing speed of 10 TOPS from a single integrated device and performing image classification tasks at high speed.
Milad Abolhasani and colleagues demonstrated Artificial Chemist 2.0, an autonomous flow chemistry platform for quantum dot synthesis
In December 2020, Milad Abolhasani's group at North Carolina State University published Artificial Chemist 2.0, an autonomous flow chemistry system combining machine learning with robotic synthesis to navigate a space of approximately 20 million quantum dot formulations and produce a target material within roughly 30 minutes of initiating a search.
CURIAL: An AI System to Detect COVID-19 in Emergency Department Patients Using Routine Blood Tests and Vital Signs
In July 2020, researchers at the University of Oxford, led by Dr Andrew Soltan and Professor David Clifton, announced CURIAL, a machine-learning model trained on routine blood tests and vital signs from 115,000 hospital presentations that could identify COVID-19 patients in emergency departments within one hour and with accuracy exceeding 90%.
OpenAI released GPT-3 via private beta API
In May–June 2020, OpenAI published the GPT-3 language model in a paper by Tom B. Brown and colleagues, and began distributing private beta API access. GPT-3's 175 billion parameters made it substantially larger than any publicly described language model at the time, enabling strong few-shot performance across diverse language tasks.
MIT Researchers Use Machine Learning to Identify Halicin, an Antibiotic Effective Against Drug-Resistant Bacteria
On 20 February 2020, James Collins and colleagues at MIT published research in Cell describing a deep-learning model trained to predict antibiotic activity; the model identified halicin, a compound previously investigated for diabetes treatment, as a potent broad-spectrum antibiotic capable of killing several drug-resistant bacterial strains.
Microsoft Research released Turing Natural Language Generation (T-NLG), a 17-billion-parameter language model
In February 2020, Microsoft Research announced Turing Natural Language Generation (T-NLG), a 17-billion-parameter autoregressive language model trained using the Megatron-LM framework. At the time of release it was the largest publicly disclosed language model and achieved state-of-the-art results on question-answering and summarisation benchmarks.
Textron Systems Unveils Ripsaw M5 Robotic Combat Vehicle
In October 2019, Textron Systems unveiled the Ripsaw M5 Robotic Combat Vehicle at the Association of the United States Army Annual Meeting in Washington, D.C., demonstrating an unmanned ground vehicle with autonomous navigation, 360-degree situational awareness, and configurable mission payloads.
Analogue resistive memory circuit solves linear algebra problems in one step
In March 2019, Daniele Ielmini and colleagues at Politecnico di Milano published results in PNAS demonstrating a crosspoint resistive-memory circuit that solves linear systems, matrix eigenvector problems, and differential equations by physical analogue relaxation, substantially reducing the energy and latency costs of conventional iterative digital solvers.
OpenAI Releases GPT-1: Improving Language Understanding by Generative Pre-Training
In June 2018, Alec Radford, Karthik Narasimhan, Tim Salimans, and Ilya Sutskever at OpenAI published 'Improving Language Understanding by Generative Pre-Training', introducing GPT-1, a 117-million-parameter Transformer pretrained on BooksCorpus via unsupervised language modelling and fine-tuned on downstream tasks, outperforming task-specific models on several NLP benchmarks.
BWIBots: A Platform for Robots Operating in Human-Inhabited Buildings
In February 2017, researchers at the University of Texas at Austin's Building-Wide Intelligence project published a description of the BWIBots platform in the International Journal of Robotics Research, detailing mobile robots designed to operate autonomously in office buildings, accept natural-language commands, and learn cooperative tasks through human interaction.
Caffe2Go: Facebook's On-Device Neural Style Transfer for Mobile Video
In November 2016, researchers at Facebook AI Research published Caffe2Go, a compressed deep-learning framework that ran neural style-transfer models entirely on iOS and Android devices without sending video frames to a server, enabling real-time artistic video effects on mobile hardware.
Nanorobotic drug delivery using flagellated bacteria by team from Polytechnique Montréal
In August 2016, a team led by Sylvain Martel at Polytechnique Montréal published research in Nature Nanotechnology demonstrating that swarms of approximately 100 million flagellated bacteria, each carrying a magnetosome acting as a compass, could transport and release an anticancer drug payload preferentially within the hypoxic regions of a tumour in a mouse model.
IBM TrueNorth: a neuromorphic integrated circuit with one million programmable neurons
In August 2014, researchers at IBM Research published a description of TrueNorth, a neuromorphic chip containing one million programmable spiking neurons and 256 million synapses on a 4096-core CMOS integrated circuit, built under the DARPA SyNAPSE programme.
SoftBank Robotics and Aldebaran Unveil Pepper, a Humanoid Robot with Emotion Recognition
In June 2014, SoftBank Robotics and its subsidiary Aldebaran Robotics unveiled Pepper, a 1.2-metre humanoid robot equipped with an emotion-recognition system capable of detecting human facial expressions, voice tone, and body language, intended for retail and customer-service deployment.
Xinlei Chen, Abhinav Shrivastava and Abhinav Gupta at Carnegie Mellon University present NEIL (Never-Ending Image Learner) at ICCV 2013
In December 2013, Xinlei Chen, Abhinav Shrivastava and Abhinav Gupta at Carnegie Mellon University presented NEIL (Never-Ending Image Learner) at ICCV 2013, a continuously running system that autonomously mined semantic relationships between visual concepts from unlabelled web images without human supervision.
Spaun: A Large-Scale Functional Brain Model Capable of Performing Multiple Cognitive Tasks
In November 2012, Chris Eliasmith and colleagues at the University of Waterloo published a description of Spaun (Semantic Pointer Architecture Unified Network) in Science, presenting a 2.5-million-neuron brain model capable of performing eight distinct cognitive tasks, including digit recognition, question answering, and list recall.
Google Brain Unsupervised Neural Network Learns to Detect Cats from YouTube Frames
In June 2012, Quoc V. Le and colleagues at Google Brain published research showing that a 1,000-machine, 16,000-core neural network trained without labels on 10 million YouTube thumbnail images spontaneously developed a neuron selectively responsive to human and cat faces, demonstrating large-scale unsupervised feature learning from unlabelled video data.
BrainGate2 Participants Use Thought-Controlled Robotic Arm to Reach and Grasp
In May 2012, researchers in the BrainGate2 clinical trial, led by Leigh Hochberg and colleagues at Massachusetts General Hospital, Brown University, and affiliated institutions, demonstrated that two participants with tetraplegia could use neural signals decoded from a 96-electrode intracortical array to control a robotic arm and perform reach-and-grasp tasks without manual assistance.
IBM Simulates 4.5 Percent of Human Brain Activity Using Blue Gene Supercomputer
In November 2011, IBM researchers led by Dharmendra Modha at IBM Research Almaden demonstrated a cortical simulation on the Blue Gene/P supercomputer that modelled approximately 4.5 percent of human-scale neural activity, using 147,456 processors to represent 1.617 billion neurons and 8.87 trillion synapses.
AeroVironment Nano Hummingbird: DARPA-funded Flapping-Wing Micro Air Vehicle
In February 2011, AeroVironment unveiled the Nano Hummingbird, a DARPA-funded flapping-wing micro air vehicle weighing 19 grams, less than a AA battery, that carried a video camera and used control systems to mimic hummingbird flight, including hover and omnidirectional movement.
Microsoft Research Develops Real-Time Human Pose Estimation for Kinect
In 2011, Jamie Shotton and colleagues at Microsoft Research Cambridge published a method for real-time human pose estimation from a single depth image, using randomised decision forests trained on synthetic data. The technique powered the skeleton-tracking feature of Microsoft Kinect and was presented at CVPR 2011.
Electronic Skin with Pressure Sensing Developed at University of Tokyo
In December 2008, Takao Someya and colleagues at the University of Tokyo published research in Nature Materials describing a flexible electronic skin using carbon nanotube composite films, enabling large-area pressure sensing suitable for robotic tactile feedback and wearable physiological monitoring.
LOPES Exoskeleton Robot for Interactive Gait Rehabilitation
Researchers at the University of Twente, led by Jan F. Veneman and colleagues, developed LOPES (Lower Extremity Powered ExoSkeleton), a treadmill-based robotic exoskeleton for interactive gait rehabilitation following stroke, with the system described fully in a 2007 IEEE publication. The robot combined powered hip and knee actuation with impedance control to support or resist patient movement during walking.
Autonomous Robotic Cardiac Surgery Guided by Machine Learning
In May 2006, a robotic surgical system at the University of Toronto, trained on data from more than 10,000 prior operations, performed an autonomous 50-minute cardiac procedure on a beating human heart, demonstrating machine-learning-guided autonomy in a clinical surgical setting.
Silicon Retina with Ganglion Cell Spiking Outputs as Neural Prosthesis
Around 2006, researchers developed a silicon retina implemented as an analogue VLSI chip that modelled four primary retinal ganglion cell types and generated 3,600 spiking outputs, designed as a neural prosthesis matched to the physical dimensions of the biological retina.
BigDog Quadruped Robot by Boston Dynamics
Around 2005, Boston Dynamics, under DARPA funding, developed BigDog, a four-legged quadruped robot capable of traversing rough terrain and carrying payloads of approximately 154 kg. BigDog demonstrated dynamic balance and legged locomotion at a scale and capability not previously achieved in a field robot.
Honda Unveils ASIMO Version with Running Capability
In December 2004, Honda demonstrated an updated version of its ASIMO humanoid robot capable of running at 3 km/h, a significant advance over prior versions that could only walk. The update also improved ASIMO's ability to navigate dynamically changing environments.
KHR-1 Humanoid Robot Platform
Researchers at the National Institute of Advanced Industrial Science and Technology (AIST) and Kawada Industries presented KHR-1 in 2004, a full-body humanoid robot platform with 30 degrees of freedom designed as a shared research base for bipedal locomotion and whole-body motion studies.
Pocket Delta: Miniaturised Parallel Delta Robot for High-Speed Micro-Assembly
In 2004, researchers at the Laboratoire de Robotique de Versailles demonstrated the Pocket Delta, a miniaturised parallel delta-structure robot capable of high-speed, sub-millimetre precision movements for micro-assembly tasks, advancing the application of parallel kinematic mechanisms to small-scale automated manufacturing.
Zoe: Autonomous Astrobiology Field Robot for the Atacama Desert
In 2004, a team from Carnegie Mellon University's Field Robotics Center, NASA Ames Research Center, and the University of Tennessee deployed the Zoe rover autonomously across Chile's Atacama Desert, demonstrating robotic detection of subsurface biological life with relevance to astrobiology and future Mars exploration.
Fujitsu Laboratories Develops Dynamically Reconfigurable Neural Network for Humanoid Robot Motor Learning
In March 2003, Fujitsu Laboratories announced a dynamically reconfigurable neural network system enabling humanoid robots to learn motor coordination from experience, reducing learning time that previously required days or months and substantially cutting the volume of motion-control software needed.
DARPA Centibots Project
DARPA funded the Centibots project, a multi-institution effort to demonstrate coordinated autonomous mapping and search using up to 100 mobile robots simultaneously, with results reported from 2003. It addressed whether large teams of robots could self-organise without continuous human supervision.
Intelligent Room and Affective Computing Agents at MIT AI Laboratory
In 1998, researchers at the MIT Artificial Intelligence Laboratory, including Rodney Brooks and Cynthia Breazeal, developed the Intelligent Room project alongside work on emotionally expressive robotic agents, combining layered behaviour-based architectures with computer vision to enable a physical space to perceive and respond to human occupants.
Cleverbot Developed by Rollo Carpenter
Rollo Carpenter, a British AI developer, launched Cleverbot as a publicly accessible web application in 1997, extending his earlier Jabberwacky project. Cleverbot learned conversational responses directly from accumulated human inputs rather than from a hand-coded rule base, and went on to accumulate hundreds of millions of logged exchanges.
Nomad Robot Field Experiment, Atacama Desert
In June 1997, Carnegie Mellon University deployed the Nomad robot in the Atacama Desert, Chile, in a NASA-funded field experiment testing long-range autonomous and teleoperated rover navigation over approximately 220 kilometres of terrain, directly informing future planetary exploration rover design.
WebCrawler Developed by Brian Pinkerton at the University of Washington
In spring 1994, Brian Pinkerton at the University of Washington deployed WebCrawler, a web search engine that used an automated crawler to index the full text of web pages, enabling users to search document content rather than titles or URLs alone.
TD-Gammon Developed by Gerald Tesauro at IBM
In 1992, Gerald Tesauro at IBM Thomas J. Watson Research Center developed TD-Gammon, a backgammon program that trained itself through self-play using temporal-difference learning applied to a multilayer neural network, reaching a standard of play close to that of strong human experts.
Yann LeCun Applies Backpropagation to Handwritten ZIP Code Recognition at AT&T Bell Labs
In 1989, Yann LeCun and colleagues at AT&T Bell Labs published 'Backpropagation Applied to Handwritten Zip Code Recognition', demonstrating that a convolutional neural network trained with backpropagation could read handwritten postal ZIP codes with high accuracy, establishing a template for practical deep learning in computer vision.
NETtalk Neural Network Developed by Terrence J. Sejnowski and Charles Rosenberg
Terrence J. Sejnowski of the Salk Institute and Charles Rosenberg of Princeton University developed NETtalk, a feedforward neural network trained to convert English text to speech, publishing the principal account in Complex Systems in 1987. The network learned pronunciation from examples alone, demonstrating that a multi-layer perceptron could acquire a complex linguistic skill without hand-coded rules.
Ernst Dickmanns and VaMoRs Autonomous Van Test, Bundeswehr University Munich
In 1986, Ernst Dickmanns and colleagues at Bundeswehr University Munich demonstrated VaMoRs, a Mercedes-Benz van retrofitted with cameras and real-time computer vision, capable of autonomous driving on traffic-free roads at speeds up to approximately 96 km/h, marking one of the earliest working autonomous vehicle demonstrations.
WABOT-2 Humanoid Robot Demonstrated at Waseda University
In 1984, researchers at Waseda University in Japan completed WABOT-2, a humanoid robot capable of reading printed musical scores, communicating with a human performer via speech, and playing an electronic organ using its fingers and foot pedals at a level comparable to an average adult pianist.
HAM-ANS Natural Language Dialog System (University of Hamburg, 1983)
In 1983, Wolfgang Hoeppner, Katharina Morik, and Heinz Marburger at the University of Hamburg presented HAM-ANS, a natural language dialog system enabling users to query databases in ordinary German, demonstrated at the GWAI-83 workshop and published in Springer's Lecture Notes in Computer Science series.
CADUCEUS Medical Expert System (Pople, University of Pittsburgh)
Harry Pople at the University of Pittsburgh developed CADUCEUS (originally called INTERNIST-1) during the late 1970s and early 1980s, publishing a detailed account in 1982. The system encoded diagnostic knowledge for several hundred internal medicine diseases and was among the most comprehensive medical diagnosis programs of its era.
Stanford Cart Slider Modification by Hans Moravec
During the late 1970s, Hans Moravec at Stanford University extended the Stanford Cart, originally built by James L. Adams in the 1960s, with a sliding camera mount, enabling the robot to navigate autonomously across a chair-filled room in 1979 using stereo vision, a milestone in autonomous vehicle research.
AM: Automated Mathematician Developed by Douglas Lenat
In 1976, Douglas Lenat at Stanford University completed AM (Automated Mathematician), a heuristic program that explored elementary mathematics by modifying and combining LISP-coded concepts, discovering properties such as prime numbers without being explicitly programmed to do so.
MYCIN: A Rule-Based Expert System for Infectious Disease Diagnosis, Developed at Stanford University
Beginning around 1972, Edward Shortliffe at Stanford University developed MYCIN, a rule-based expert system written in Lisp that used approximately 600 if-then rules to diagnose bacterial blood infections and recommend antibiotic treatments adjusted for patient body weight, establishing a widely studied model for clinical decision support.
Stanford Cart Road-Following Experiment Under Les Earnest
In late 1971, Les Earnest at the Stanford Artificial Intelligence Laboratory adapted the Stanford Cart for autonomous road-following, using a television camera and a low-power radio control link to guide the vehicle along a road at approximately 0.8 mph (1.3 kph).
SHRDLU Natural Language Understanding Program Developed by Terry Winograd at MIT
In 1970, Terry Winograd at the Massachusetts Institute of Technology completed SHRDLU, a natural language understanding program that allowed a user to converse in English about a simulated world of coloured blocks, demonstrating that a computer could parse and respond to complex grammatical instructions within a constrained domain.
2001: A Space Odyssey Released, Featuring HAL 9000
In April 1968, Stanley Kubrick and Arthur C. Clarke's film 2001: A Space Odyssey introduced HAL 9000, a fictional shipboard AI capable of natural speech, lip-reading, chess play, and autonomous decision-making. The portrayal became a lasting cultural reference in debates about machine intelligence, autonomy, and the risks of over-reliance on automated systems.
Analogy: A Program That Solves Geometric Analogy Problems by Thomas C. Evans
Thomas C. Evans, working at MIT, developed ANALOGY, a program capable of solving geometric analogy problems of the type found in standard IQ tests. First presented in dissertation form in 1963 and formally published in 1968, it demonstrated that a computer could perform a structured form of relational reasoning.
SIR (Semantic Information Retrieval) by Bertram Raphael, MIT
In 1964, Bertram Raphael at MIT completed his doctoral thesis describing SIR (Semantic Information Retrieval), a LISP program that answered English-language questions about simple relational facts by constructing an internal semantic model. It was an early demonstration that a computer could parse and reason over a restricted subset of natural language.
Rancho Arm Connected to Computer at Stanford Artificial Intelligence Laboratory
Around 1963, engineers at Rancho Los Amigos Hospital collaborated with Stanford Artificial Intelligence Laboratory to interface the six-degree-of-freedom Rancho Arm, an electrically powered prosthetic limb, with a computer, producing one of the earliest demonstrations of computer-controlled multi-jointed robotic manipulation.
Stanford Cart Radio-Link Configuration (1963)
In September 1963, researchers at Stanford University fitted the Stanford Cart with an analogue computer and radio-control links, allowing a remote operator to steer the vehicle using a television camera feed and a displayed target dot, in an early attempt at closed-loop vision-guided vehicle control.
IBM Shoebox Speech-Recognition Device Demonstrated
In 1962, IBM publicly demonstrated the Shoebox, a hardware device capable of recognising sixteen spoken English words, the digits zero through nine plus six arithmetic commands, and using that input to drive a calculating machine in real time, showing that limited-vocabulary speech recognition was mechanically feasible.
James Slagle Completes SAINT: A Symbolic Automatic Integrator
In 1961, James R. Slagle at MIT completed SAINT (Symbolic Automatic INTegrator), a heuristic program that solved indefinite integration problems at the level of a first-year calculus student, demonstrating that symbolic mathematical reasoning could be encoded in a computer program.
The Logic Theorist: first automatic theorem-proving program, developed by Newell, Shaw, and Simon
In late 1955, Allen Newell, J. Clifford Shaw, and Herbert A. Simon at RAND Corporation and Carnegie Institute of Technology completed the Logic Theorist, a computer program that proved mathematical theorems from Whitehead and Russell's Principia Mathematica by heuristic search, establishing that symbolic reasoning could be mechanised.