Stanford Artificial Intelligence Lab
STA N FOR D U N I V ER SI T Y S C HO OL OF ENGI N EER I NG / C OM PU T ER S C I ENC E DEPA RTM EN TA RTI FICI A L I NTELLIGENCE LA BFEBRUA RY 2 0191StanfordArtificialIntelligenceLab
2STA N FOR D U N I V ERSITYManeesh AgrawalaRon DrorNoah GoodmanMykel KochenderferProfessor, Computer Science;Director, the Brown Institute forMedia InnovationComputer graphics, humancomputer interaction,visualizationTo improve the effectiveness of mediaof all kindsAssociate Professor, ComputerScienceComputational biologyTo determine spatial structure anddynamics at the molecular andcellular levelsAssociate Professor, Psychology,Linguistics (courtesy), ComputerScience (courtesy)Computational psychology,machine learning, linguisticsTo understand cognition, languageand social behaviorAssistant Professor, Aeronauticsand Astronautics, ComputerScience (courtesy)Machine learning, decisiontheoryTo develop safe and efficientsystems for air traffic, drones, andautonomous vehiclesGill BejeranoAssociate Professor, ComputerScience, DevelopmentalBiology, Pediatrics(School of Medicine)Computational genomicsTo discover the relationship betweenthe genotype and phenotypeJeanette BohgAssistant Professor, ComputerScienceRobotics, machine learning,computer visionTo understand the fundamentalprinciples of complex, sensorimotorbehavior and how it can be generatedon robotsEmma BrunskillAssistant Professor, ComputerScienceMachine learning/deeplearningTo advance the frontiers ofreinforcement learningJohn DuchiAssistant Professor, ElectricalEngineering, StatisticsMachine learning,optimization and statisticsTo understand the limits andapplication of statistics and AILeo GuibasProfessor, Computer ScienceComputer vision, computergraphics, geometryTo accurately describe the realphysical 3D worldStefano ErmonThomas IcardAssistant Professor, ComputerScienceMachine learning, statistics,sustainabilityTo solve impactful problems insustainabilityAssistant Professor ofPhilosophy and (by courtesy) ofComputer ScienceMachine learning/deeplearning, big data, knowledgebases, logicTo understand how reasoning works,and how we might like it to workRon FedkiwProfessor, Computer ScienceComputer graphics, computervision, fluid dynamics, solidmechanics, biomechanicsTo design computational algorithmsMichael GeneserethAssociate Professor, ComputerScienceComputational logicTo solve problems in business, law,and game playingDan JurafskyProfessor, Linguistics,Computer ScienceComputational linguisticsTo solve problems and provideinsights in behavioral andsocial sciencesOussama KhatibProfessor, Computer ScienceRoboticsTo enable a new generation of robotsthat cooperate with humans and otherrobots in complex and unpredictableenvironmentsDaphne KollerAdjunct Professor, ComputerScience; CEO, insitroMachine learningTo solve real-world problemsinvolving complexity and uncertaintyAnshul KundajeAssistant Professor, Genetics(School of Medicine), ComputerScienceComputational biology,machine learningTo analyze all kinds of genomicand genetic data to understand generegulationJames LandayProfessor, Computer ScienceHuman-computer interaction,NLP, autonomous vehiclesTo design user-centered AI systemsthat augment and support peoplerather than replace them
3A RTI FICI A L I NTELLIGENCE LA BJure LeskovecChris ManningMarco PavoneSilvio SavareseAssociate Professor,Computer ScienceData mining, machine learningTo study the workings of large socialand information networksProfessor, Computer Science,Linguistics; Director, StanfordArtificial Intelligence LabNatural language processing,machine learningTo develop computers that can process,understand and generate humanlanguageAssistant Professor, Aeronauticsand AstronauticsRobotics; decision theory;autonomous vehiclesTo push the limits of robot autonomyvia principled decision-making andlearning algorithmsAssociate Professor, ComputerScienceComputer vision, robotics,geometry, machine learningTo push the limits of computer visionfor object, scene and human behaviorunderstanding and applications insocial robots and autonomous vehiclesFei-Fei LiProfessor, Computer Science,Psychology (courtesy)Computer vision, machinelearning, computationalcognitive neuroscienceTo help computers see better to helpand work with humansPercy LiangAssistant Professor, ComputerScience, Statistics (courtesy)Machine learning, naturallanguage processingTo build systems that allow humansand computers to communicateTengyu MaAssistant Professor, ComputerScience, StatisticsMachine learning/algorithmsTo understand and develop machinelearning algorithmsAndrew NgAdjunct Professor,Computer ScienceMachine learningTo solve problems in autonomousdriving, robots, image analysis andlanguageJuan Carlos NieblesSenior Research Scientist,Computer ScienceComputer vision, machinelearningTo allow computers to understandobjects, scenes, activities and events inimages and videosVijay PandeAdjunct Professor,BioengineeringStatistical mechanics,Bayesian statistics,biophysics, biochemistryTo push the limits of simulationand statistics to answer questions inbiophysics and biochemistryChristopher RéAssociate Professor, ComputerScienceDatabase, machine learning,theoryTo create the future of data systemsfor unstructured and structured dataDorsa SadighAssistant Professor, ComputerScience, Electrical EngineeringMachine learning/deep learning, robotics,autonomous vehiclesTo design algorithms for robots thatsafely interact with peopleKen SalisburyProfessor (Research), ComputerScienceRoboticsTo develop useful robots for surgery,imaging, haptics and personalassistanceSebastian ThrunAdjunct Professor, ComputerScience; CEO, UdacityRobotics, machine learningTo improve robotics, autonomousvehicles, smart homes, healthcare anddronesDan YaminsAssistant Professor, Psychology,Computer Science (courtesy)Computational neuroscienceTo reverse engineer the human brainand build more effective AI systemsJames ZouAssistant Professor ofBiomedical Data Science and (bycourtesy) of Computer Scienceand of Electrical EngineeringMachine learning,computational statistics,computational biology,AI for healthTo develop ML algorithms and theorythat are motivated by cutting-edgedevelopments in biotechnology andhealthcare
STA N FOR D U N I V ERSITYIntroductionDear Friends,ARTIFICIAL INTELLIGENCEWelcome to the Stanford Artificial Intelligence LabThe Stanford Artificial Intelligence Lab (SAIL)was founded by Prof. John McCarthy, one ofthe founding fathers of the field of AI. Whilethe discipline of AI has transformed in manyfundamental ways since its inception in the1950s, SAIL remains a proud leading intellectualhub for scientists and engineers, an educationmecca for students, and a center of excellencefor cutting edge research work. With thisbrochure, we hope to share with you some ofthe latest research and activities at SAIL.Reflecting on the history of AI, the pastfifty years are mostly what I call the “AI invitro” times, during which most AI researchwas conducted in academic laboratories.This is the time when AI researchers laidthe foundations for our fields, including thequestions we are pursuing, the methodologies,the measurements and metrics, and thepotential applications. AI grew from a small setContentsIntroduction . . . . . . . . . . . . . . . . . . . . . . . . . . . 4Machine Learning / Deep Learning . . . . . . . 6Natural Language Processing . . . . . . . . . . 10Computer Vision . . . . . . . . . . . . . . . . . . . . . . 14Robotics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18Genomics and Healthcare . . . . . . . . . . . . . 22Autonomous Vehicles . . . . . . . . . . . . . . . . . . 26AI Salon . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30Stanford AI4ALL . . . . . . . . . . . . . . . . . . . . . . . 32Faculty Focus . . . . . . . . . . . . . . . . . . . . . . . . . 34of ideas to important areas such as robotics,natural language processing, computer vision,computational genomics and medicine,among others.An important development in the field of AIwas the emergence of machine learning inthe 1980s. The recent convergence of moderncomputing hardware, big data and powerfulmachine learning algorithms has led tosome breathtaking breakthroughs in manyapplications of AI, from speech recognition, toimage recognition, to self-driving cars. We havenow entered the time of “AI in vivo,” where AItechnologies are changing people’s everydaylives as products of chatbots, photo organizers,assistive driving technology, and much more.My colleagues at SAIL have been at theforefront of this AI revolution, just like theywere the pioneers in establishing the AI fieldhalf a century ago. They continue to pushthe boundaries of fundamental research andinnovative applications in many different areasof AI, from teaching computers to translatebetween languages, to developing a swimmingrobot that can explore shipwrecks deep underwater following haptic instructions from anarchaeologist at the surface, to enablingcomputers to read satellite images that canhelp governments to monitor the state ofregional economy, to creating a smart visualrobot that can navigate among crowds insocially courteous ways that have been uniqueto humans, or to developing algorithms to sort
5A RTI FICI A L I NTELLIGENCE LA Bthrough millions of gene sequences to discoverthe genetic makeup of fatal diseases. And thelist goes on.Equally important to the mission of technologyinnovation, we believe in the education of thenext generation of technologists. As part of theStanford Computer Science Department, SAILfaculty and researchers have been teachingstudents across the Stanford campus andfrom outside. It is well known that ComputerScience has become the largest undergraduatemajor at Stanford since 2015. Less known is thefact that except for the introductory classes,all the most popular CS classes are now inAI: AI Principles and Techniques, MachineLearning, Natural Language Processing andComputer Vision. In the various research groupsof SAIL, undergraduates, including women andunder-represented minority students, are oftenworking alongside researchers on the mostcutting-edge research projects. In this brochure,we highlight two unique programs at SAIL thatare beloved by the students. One is AI Salon,the regular SAIL-wide discussion. The other isStanford AI4ALL, which was the first in a nowgrowing number of summer camps for highschoolers. Stanford AI4ALL targets high schoolgirls focusing on humanistic AI.This is a historic time for AI researchers,technologists, educators and students.There has never been so much excitementand hope for the potential promises ofAI. But equally important, there has neverbeen so much need to create benevolentAI technologies and to educate humanistic AItechnologists for our world. To address thisneed, SAIL is strongly involved in creating anew university-wide Human-centered ArtificialIntelligence Initiative and we will have muchmore to tell you about that in the coming year.We have come a long way, but the journeyahead of us is still long. None of today’sintelligent machines come close to the breadthand depth of human intelligence. So all of us atSAIL are striving to build better algorithms andmachines that will help humans to live better,safer, more productive and healthier lives.Sincerely yours,Christopher ManningDirector, Stanford Artificial Intelligence LabThomas M. Siebel Professor in Machine LearningThe SAIL research lab and communityare blossoming.
STA N FOR D U N I V ERSITYMACHINE LEARNING / DEEP LEARNINGMachine Learning / Deep LearningInvestigating how machines can learn to improve their perception,cognition, and actions with experience has become a bedrock discipline ofAI in recent years. Since 2010, it has moved out of academia, where it hadbeen explored for decades, and into industry where it is starting to remakehuman-computer interactions.Deep learning, a key sub area of machinelearning, is transforming tasks that longstymied researchers. For instance, speechrecognition used to be based on algorithmscreated by smart human programmers. Thealgorithms told computers to try to matchpatterns representing specific phonemes indigitized recordings. But in the last five years,as computers became more powerful, itbecame practical to build very large learningtools called deep neural networks. A neuralnetwork iteratively processes data, throughlayers of nodes. After listening to thousandsof recordings and reading the correspondingtranscripts, the computer learns for itselfwhat digitized sound waves correspond to theword “Trump.”The key to machine learning is data—lots of it.If there is lots of annotated data that can beexamined by a computer, the computer canlearn to identify and understand it. Captionedphotos, spoken language with transcripts, text,videos and genomes all provide rich libraries ofinformation that computers can understand.The deep learning revolution took off after 2009.One spark occurred when Stanford researcherFei-Fei Li released ImageNet, a free databaseof 14 million images that had been labeled bytens of thousands of Amazon Mechanical Turkworkers. When AI researchers started usingImageNet to train neural networks to catalogphotos and annotate their contents, machinelearning exploded.SAIL faculty working in machine learning, deep learning, big data, knowledge base, and logic include:Jeannette Bohg, Emma Brunskill, John Duchi, Stefano Ermon, Michael Genesereth, Noah Goodman, Leonidas Guibas,Thomas Icard, Mykel Kochenderfer, Daphne Koller, Jure Leskovec, Fei-Fei Li, Percy Liang, Tengyu Ma, Chris Manning,Andrew Ng, Christopher Ré, Dorsa Sadigh, Sebastian Thrun, Dan Yamins, James Zou.
7A RTI FICI A L I NTELLIGENCE LA BMachine learning replaces the complexity ofwriting algorithms that cover every eventualitywith the complexity of processing lots ofdata. Processing the data can be a simplertask. With sufficient computing resources,the machines can learn much faster thanprogrammers can teach them.Machine learning can be applied to tasksthat seem far afield. For example, Stanfordprofessor Stefano Ermon looks for societalproblems that can be addressed with machinelearning techniques. Identifying poverty levelsin underdeveloped countries is a difficult butimportant task for governments and povertyfighting organizations. Ground level surveysare expensive and difficult. Financial dataare scarce.Ermon obtained high resolution satellite imagesof villages in Africa where asset levels had beendocumented by surveys. He used those to traina machine learning system. The system wasable to spot features like roads, plowed fieldsand metal roofs as signs of relative wealth.After validating the program’s conclusions,Ermon published his results and made theprogram available to governments and NGOs,where it is starting to be used. Because satelliteimages cover every corner of the world and arefrequently updated, agencies can even studyareas like Somalia where warfare makes humansurveys too dangerous.He has used similar techniques to look atgrowth in farm fields. Other applicationsinvolve energy and the environment.He hopes to use machine learning to figureout ways to maximize battery life in electriccars. Other researchers are trying to crowdsource sufficient data to get new insights.One target: understanding bird migrationand species distribution by getting amateurbird watchers to report their observations.Ermon thinks machine learning could alsoprovide guidelines on how to manage fisheriesto benefit fishermen while preserving thepopulation stock.
8STA N FOR D U N I V ERSITYMachine Learning / Deep LearningStill, many humans are likely to be skepticalof relying on machines that have essentiallytrained themselves. When neural networkslearn things on their own, they sometimeserr. In 2015, an automated labeling feature ofGoogle Photos mistakenly identified someblack people as “gorillas.” Google says it hastweaked the feature, but it’s impossible topredict other unexpected results.In some cases, people will undoubtedly insiston a full understanding of the machinelearning process. As AI guides robots, carsand planes, if an accident occurs, andengineers can’t reconstruct the decisionmaking process, liability lawyers will have afield day.Machine learning will be a core componentof solving the next generation of AI problems.Robotics, natural language conversationsand understanding videos are all tasksthat require machine learning, probablyaugmented by other techniques that are stillunder development.
A RTI FICI A L I NTELLIGENCE LA BProf. Kochenderfer’s studentsuse partially observableMarkov decision processes(POMDPs) to representproblems that involve decisionmaking under uncertainty.Techniques developed byhis students are being usedto derive optimal controlstrategies for automaticallyfinding GPS jammers thatcould interfere with air trafficcontrol.Machine learning will bea core component of solvingthe next generation of AI problems.9
STA N FOR D U N I V ERSITYNatural Language ProcessingANATURAL LANGUAGE PROCESSINGrtificial intelligence has made dramatic progress in understandingspeech. Computers commonly make transcripts and provide phraseby-phrase translations. But understanding human language is athornier problem.Natural language processing is one of theliveliest areas of AI research, in part because itaddresses so deeply the issue of what makeshumans human. Computers can listen tospeech and learn the meanings of individualwords and phrases. Processing speech issusceptible to using neural networks thattrain themselves to recognize the streams ofbits that represent spoken words. But reallyunderstanding full sentences in context orhaving a dialogue is still a daunting task. AImethods pioneered at Stanford have made itpossible to extract the grammatical structure ofa sentence with high accuracy.Nevertheless, computers are getting better atnatural language. AI has made huge progressin improving language translation using deeplearning techniques and neural networks. Bylooking at hundreds of thousands or evenmillions of documents in English that have beentranslated to Chinese, a neural network canlearn how to translate a full sentence. StanfordNLP Group founder Chris Manning and hisstudents have helped advance this new NeuralMachine Translation paradigm.Sentiment tree. Jason Chuang and Richard SocherBy using large databases of newspaperarticles, neural networks can learn to generatesummaries or headlines for articles. That canmake it easier for humans to scan search resultsand find what they need.Researchers are also applying AI techniques tounderstand the social meanings behind writtenand spoken language, people’s opinions,attitudes, and emotions. For instance, Stanfordprofessor Dan Jurafsky works on sentimentanalysis, to understand how groups of peoplefeel about a subject discussed on social medialike Reddit or Twitter. One challenge is thatwords have sharply different connotations inSAIL faculty working in natural language processing include:Noah Goodman, Dan Jurafsky, James Landay, Percy Liang, Chris Manning, James Zou
11A RTI FICI A L I NTELLIGENCE LA Bdifferent contexts. For example, he notes, “soft”is a good thing when people are discussingpuppies; it’s negative when discussing footballplayers.Jurafsky and other researchers developed alexicon of accurately labeled sentiment wordsby identifying a few positive and negativewords that are roughly consistent acrosshundreds of domains. Then they used thoseseed words to let their algorithm create amuch larger lexicon of related words specificto each domain. Creating such lexicons isextremely time-consuming for humans, andit has been a barrier to progress in sentimentanalysis. One interesting byproduct of thetechnology is that it has made it practical to dosentiment analysis on historical documents likenewspapers and letters. That is tricky becausewords’ meanings change over time. “Awful”once was a positive word, connoting awe andmajesty, but today it has a negative se
ARTIFICIAL INTELLIGENCE LAB STANFORD UNIVERSITY SCHOOL OF ENGINEERING / COMPUTER SCIENCE DEPARTMENT 1 Stanford Artificial Intelligence Lab FEBRUA RY 2 019. 2 STANFORD UNIVERSITY Maneesh Agrawala Professor, Computer Science; Director, the Brown Institute for Media Innovation Computer graphics, human-computer interaction, visualization To improve the effectiveness of media of all kinds Gill .
SEISMIC: A Self-Exciting Point Process Model for Predicting Tweet Popularity Qingyuan Zhao Stanford University qyzhao@stanford.edu Murat A. Erdogdu Stanford University erdogdu@stanford.edu Hera Y. He Stanford University yhe1@stanford.edu Anand Rajaraman Stanford University anand@cs.stanford.edu Jure Leskovec Stanford University jure@cs.stanford .
Artificial Intelligence -a brief introduction Project Management and Artificial Intelligence -Beyond human imagination! November 2018 7 Artificial Intelligence Applications Artificial Intelligence is the ability of a system to perform tasks through intelligent deduction, when provided with an abstract set of information.
and artificial intelligence expert, joined Ernst & Young as the person in charge of its global innovative artificial intelligence team. In recent years, many countries have been competing to carry out research and application of artificial intelli-gence, and the call for he use of artificial
BCS Foundation Certificate in Artificial Intelligence V1.1 Oct 2020 Syllabus Learning Objectives 1. Ethical and Sustainable Human and Artificial Intelligence (20%) Candidates will be able to: 1.1. Recall the general definition of Human and Artificial Intelligence (AI). 1.1.1. Describe the concept of intelligent agents. 1.1.2. Describe a modern .
IN ARTIFICIAL INTELLIGENCE Stuart Russell and Peter Norvig, Editors FORSYTH & PONCE Computer Vision: A Modern Approach GRAHAM ANSI Common Lisp JURAFSKY & MARTIN Speech and Language Processing, 2nd ed. NEAPOLITAN Learning Bayesian Networks RUSSELL & NORVIG Artificial Intelligence: A Modern Approach, 3rd ed. Artificial Intelligence A Modern Approach Third Edition Stuart J. Russell and Peter .
Peter Norvig Prentice Hall, 2003 This is the book that ties in most closely with the module Artificial Intelligence (2nd ed.) Elaine Rich & Kevin Knight McGraw Hill, 1991 Quite old now, but still a good second book Artificial Intelligence: A New Synthesis Nils Nilsson Morgan Kaufmann, 1998 A good modern book Artificial Intelligence (3rd ed.) Patrick Winston Addison Wesley, 1992 A classic, but .
BCS Essentials Certificate in Artificial Intelligence Syllabus V1.0 BCS 2018 Page 10 of 16 Recommended Reading List Artificial Intelligence and Consciousness Title Artificial Intelligence, A Modern Approach, 3rd Edition Author Stuart Russell and Peter Norvig, Publication Date 2016, ISBN 10 1292153962
1. People who automatically get a Blue Badge 2. People who may get a Blue Badge. 3. What you need to apply for a Blue Badge 4. Providing proof of benefits. 5. Proof of eligibility 6. How you can provide documents. 7. Blue Badges for organisations 8. Related guidance. Blue Badges help people with disabilities or health conditions park closer to .
