Artificial Intelligence And Robotics

3y ago
62 Views
1 Downloads
5.17 MB
56 Pages
Last View : 15d ago
Last Download : 4m ago
Upload by : Maleah Dent
Transcription

Artificial Intelligence and RoboticsJavier Andreu Perez, Fani Deligianni, Daniele Ravi and Guang-Zhong Yang

// Artificial Intelligence and Robotics

Artificial Intelligence and Robotics //UKRAS.ORG

// Artificial Intelligence and RoboticsF ORE WOR DWelcome to the UK-RAS White PaperSeries on Robotics and AutonomousSystems (RAS). This is one of the coreactivities of UK-RAS Network, funded bythe Engineering and Physical SciencesResearch Council (EPSRC). By bringingtogether academic centres of excellence,industry, government, funding bodies andcharities, the Network provides academicleadership, expands collaboration withindustry while integrating and coordinatingactivities at EPSRC funded RAS capitalfacilities, Centres for Doctoral Training andpartner universities.cognitive abilities. Current AI technologiesare used in a set area of applications,ranging from healthcare, manufacturing,transport, energy, to financial services,banking, advertising, managementconsulting and government agencies. Theglobal AI market is around 260 billion USDin 2016 and it is estimated to exceed 3trillion by 2024. To understand the impactof AI, it is important to draw lessons fromit's past successes and failures and thiswhite paper provides a comprehensiveexplanation of the evolution of AI, its currentstatus and future directions.The recent successes of AI have capturedthe wildest imagination of both the scientificcommunities and the general public.Robotics and AI amplify human potentials,increase productivity and are moving fromsimple reasoning towards human-likeThe UK-RAS white papers are intended toserve as a basis for discussing the futuretechnological roadmaps, engaging thewider community and stakeholders, as wellas policy makers, in assessing the potentialsocial, economic and ethical/legal impactDr. Javier Andreu Perez,The Hamlyn Centre,Imperial College LondonDr. Fani Deligianni,The Hamlyn Centre,Imperial College Londonof RAS. It is our plan to provide annualupdates for these white papers so yourfeedback is essential - whether it is to pointout inadvertent omissions of specific areasof development that need to covered, or tosuggest major future trends that deservefurther debate and in-depth analysis.Please direct all your feedback to whitepaper@ukras.org. We look forward tohearing from you!Prof Guang-Zhong Yang, CBE, FREngChair, UK-RAS NetworkDr. Daniele Ravi,The Hamlyn Centre,Imperial College LondonProf. Dr. Guang-Zhong Yang,The Hamlyn Centre,Imperial College LondonOn behalf of the UK-RAS Network, established to provide academic leadership, expand collaboration with industrywhile integrating and coordinating activities at EPSRC funded RAS capital facilities, Centres for Doctoral Trainingand partner universities.

Artificial Intelligence and Robotics //EXE CUTI V E S UM M ARYIn 2016, Artificial Intelligence (AI) celebratedits 60th anniversary of the DartmouthWorkshop, which marked the beginningof AI being recognised as an academicdiscipline. One year on, the pace of AI hascaptured the wildest imagination of both thescientific community and the general public.The term AI now encompasses the wholeconceptualisation of a machine that isintelligent in terms of both operational andsocial consequences. With the predictionof the AI market to reach 3 trillion by 2024,both industry and government fundingbodies are investing heavily in AI androbotics. As the availability of informationaround us grows, humans will rely moreand more on AI systems to live, to work,and to entertain. Given increased accuracyand sophistication of AI systems, they willbe used in an increasingly diverse range ofsectors including finance, pharmaceuticals,energy, manufacturing, education, transportand public services. It has been predictedthat the next stage of AI is the era ofaugmented intelligence. Ubiquitous sensingsystems and wearable technologies aredriving towards intelligent embeddedsystems that will form a natural extension ofhuman beings and our physical abilities. WillAI trigger a transformation leading to superintelligence that would surpass all humanintelligence?This white paper explains the origin of AI,its evolution in the last 60 years, as wellas related subfields including machinelearning, computer vision and the rise ofdeep learning. It provides a rational viewof the different seasons of AI and how tolearn from these ‘boom-and-bust’ cyclesto ensure the current progresses aresustainable and here to stay. Along withthe unprecedented enthusiasm of AI, thereare also fears about the impact of thetechnology on our society. A clear strategyis required to consider the associatedethical and legal challenges to ensure thatsociety as a whole will benefit from theevolution of AI and its potential negativeimpact is mitigated from early on. To thisend, the paper outlines the ethical and legalissues of AI, which encompass privacy,jobs, legal responsibility, civil rights, andwrongful use of AI for military purposes.The paper concludes by providing a setof recommendations to the researchcommunity, industry, government agenciesand policy makers.To sustain the current progress of AI, it isimportant to understand what is sciencefiction and what is practical reality. Arational and harmonic interaction is requiredbetween application specific projectsand visionary research ideas. Neither theunrealistic enthusiasm nor the unjustifiedfears of AI should hinder its progress.They should be used to motivate thedevelopment of a systematic frameworkon which the future of AI will flourish.With sustained funding and responsibleinvestment, AI is set to transform thefuture of our society - our life, our livingenvironment and our economy.

// Artificial Intelligence and RoboticsRobotics and AI augment and amplify humanpotentials, increase productivity and are movingfrom simple reasoning towards human-like cognitiveabilities. To understand the impact of AI, it is importantto draw lessons from the past successes andfailures, as well as to anticipate its future directionsand potential legal, ethical and socio-economicimplications.”

Artificial Intelligence and Robotics //CONTENTS1I n t ro d u c t i o n22T h e b i rt h a n d b o o m o f A I33Q u e s t i o n i n g t h e i m p a ct of A I74A c l o s e r l o o k a t t h e e vol ut i on of A I4 .1Seasons of AI4 .2I n f l u e n c e o f f u n d i ng4 .3P u b l i c a t i o n v s Pat ent i ng999105Financial impact of AI126S u b f i e l d s a n d t e c h n o l ogi es t hat und erp i nni ngs A I187T h e ri s e o f De e p L e a r n i ng: ret hi nk i ng t he machi ne l ear ni ng p i p el i ne198H a rd w a re f o r A I229Robotics and AI241 0 P ro g ra m m i n g l a n g u a g es f or A I271 1 I m p a c t o f M a c h i n e Vi si on321 2 A rt i f i c i a l I n t e l l i g e n c e and t he Bi g Brai n361 3 E t h i c a l a n d l e g a l q u e st i ons of A I1 3 .1 E t h i c a l i s s u e s i n A I1 3 .2 L e g a l i s s u e s a n d q uest i ons of A I3939401 4 L i m i t a t i o n s a n d o p p o r t uni t i es of A I411 5 C o n c l u s i o n a n d re c o mmend at i ons43R e f e re n c e s44

1 // Artificial Intelligence and RoboticsWith increased capabilities and sophistication of AIsystems, they will be used in more diverse ranges ofsectors including finance, pharmaceuticals, energy,manufacturing, education, transport and public services.The next stage of AI is the era of augmented intelligence,seamlessly linking human and machine together.

Artificial Intelligence and Robotics // 21. INTRODUCTIONArtificial Intelligence (AI) is a commonly employed appellationto refer to the field of science aimed at providing machineswith the capacity of performing functions such as logic,reasoning, planning, learning, and perception. Despite thereference to “machines” in this definition, the latter couldbe applied to “any type of living intelligence”. Likewise, themeaning of intelligence, as it is found in primates and otherexceptional animals for example, it can be extended toinclude an interleaved set of capacities, including creativity,emotional knowledge, and self-awareness.The term AI was closely associated with the field of“symbolic AI”, which was popular until the end of the 1980s.In order to overcome some of the limitations of symbolic AI,subsymbolic methodologies such as neural networks, fuzzysystems, evolutionary computation and other computationalmodels started gaining popularity, leading to the term“computational intelligence” emerging as a subfield of AI.Nowadays, the term AI encompasses the wholeconceptualisation of a machine that is intelligent in termsof both operational and social consequences. A practicaldefinition used is one proposed by Russell and Norvig:“Artificial Intelligence is the study of human intelligenceand actions replicated artificially, such that the resultantbears to its design a reasonable level of rationality” [1]. Thisdefinition can be further refined by stipulating that the levelof rationality may even supersede humans, for specific andwell-defined tasks.Current AI technologies are used in online advertising,driving, aviation, medicine and personal assistance imagerecognition. The recent success of AI has captured theimagination of both the scientific community and the public.An example of this is vehicles equipped with an automaticsteering system, also known as autonomous cars. Eachvehicle is equipped with a series of lidar sensors andcameras which enable recognition of its three-dimensionalenvironment and provides the ability to make intelligentdecisions on maneuvers in variable, real-traffic roadconditions. Another example is the Alpha-Go, developed byGoogle Deepmind, to play the board game Go. Last year,Alpha-Go defeated the Korean grandmaster Lee Sedol,becoming the first machine to beat a professional player andrecently it went on to win against the current world numberone, Ke Jie, in China. The number of possible games in Gois estimated to be 10761 and given the extreme complexityof the game, most AI researchers believed it would beyears before this could happen. This has led to both theexcitement and fear in many that AI will surpass humans inall the fields it marches into.However, current AI technologies are limited to very specificapplications. One limitation of AI, for example, is the lack of“common sense”; the ability to judge information beyondits acquired knowledge. A recent example is that of the AIrobot Tay developed by Microsoft and designed for makingconversations on social networks. It had to be disconnectedshortly after its launch because it was not able to distinguishbetween positive and negative human interaction. AI is alsolimited in terms of emotional intelligence. AI can only detectbasic human emotional states such as anger, joy, sadness,fear, pain, stress and neutrality. Emotional intelligence is oneof the next frontiers of higher levels of personalisation.True and complete AI does not yet exist. At this level, AI willmimic human cognition to a point that it will enable the abilityto dream, think, feel emotions and have own goals. Althoughthere is no evidence yet this kind of true AI could exist before2050, nevertheless the computer science principles drivingAI forward, are rapidly advancing and it is important toassess its impact, not only from a technological standpoint,but also from a social, ethical and legal perspective.

3 // Artificial Intelligence and Robotics2. THE BIRTH AND BOOM OF AIThe birth of the computer took place when the firstcalculator machines were developed, from the mechanicalcalculator of Babbage, to the electromechanical calculatorof Torres-Quevedo. The dawn of automata theory canbe traced back to World War II with what was known asthe “codebreakers”. The amount of operations requiredto decode the German trigrams of the Enigma machine,without knowing the rotor’s position, proved to be toochallenging to be solved manually. The inclusion of automatatheory in computing conceived the first logical machines toaccount for operations such as generating, codifying, storingand using information. Indeed, these four tasks are the basicoperations of information processing performed by humans.The pioneering work by Ramón y Cajal marked the birth ofneuroscience, although many neurological structures andstimulus responses were already known and studied beforehim. For the first time in history the concept of “neuron”was proposed. McClulloch and Pitts further developed aconnection between automata theory and neuroscience,proposing the first artificial neuron which, years later, gaverise to the first computational intelligence algorithm, namely“the perceptron”. This idea generated great interest amongprominent scientists of the time, such as Von Neumann,who was the pioneer of modern computers and set thefoundation for the connectionism movement.1956 –when the term AI was first coined.The Dartmouth Conference of 1956 was organizedby Marvin Minsky, John McCarthy and two seniorscientists, Claude Shannon and Nathan Rochesterof IBM. At this conference, the expression“Artificial Intelligence” was first coined as the titleof the field. The Dartmouth conference triggered anew era of discovery and unrestrained conquestsof new knowledge. The computer programmesdeveloped at the time are considered by mostas simply "extraordinary"; computers solvealgebraic problems, demonstrate theorems ingeometry and learnt to speak English. At thattime, many didn’t believe that such "intelligent"behavior was possible in machines. Researchersdisplayed a great deal of optimism both in privateand in scientific publications. They predictedthat a completely intelligent machine would bebuilt in the next 20 years. Government agencies,such as the US Defence and Research ProjectAgency (DARPA), were investing heavily in thisnew area. It is worth mentioning, that some ofthe aforementioned scientists, as well as majorlaboratories of the time, such as Los Alamos(Nuevo Mexico, USA), had strong connections withthe army and this link had a prominent influence,as the work at Bletchley Park (Milton Keynes,UK) had, over the course of WWII, as did politicalconflicts like the Cold War in AI innovation.Since 2000, a third renaissance of theconnectionism paradigm arrived with the dawnof Big Data, propelled by the rapid adoption ofthe internet and mobile communication.

Artificial Intelligence and Robotics // 4In 1971, DARPA funded a consortium of leading laboratoriesin the field of speech recognition. The project had theambitious goal of creating a fully functional speechrecognition system with a large vocabulary. In the middleof the 1970s, the field of AI endured fierce criticism andbudgetary restrictions, as AI research development did notmatch the overwhelming expectations of researchers. Whenpromised results did not materialize, investment in AI eroded.Following disappointing results, DARPA withdrew funding inspeech recognition and this, coupled with other events suchas the failure of machine translation, the abandonment ofconnectionism and the Lighthill report, marked the first winterof AI [2]. During this period, connectionism stagnated forthe next 10 years following a devastating critique by MarvinMinksy on perceptrons [3].From 1980 until 1987, AI programmes, called “expertsystems”, were adopted by companies and knowledgeacquisition become the central focus of AI research. At thesame time, the Japanese government launched a massivefunding program on AI, with its fifth-generation computersinitiative. Connectionism was also revived by the work ofJohn Hopfield [4] and David Rumelhart [5].AI researchers who had experienced the first backlash in1974, were sceptical about the reignited enthusiasms ofexpert systems and sadly their fears were well founded.The first sign of a changing tide was with the collapse ofthe AI computer hardware market in 1987. Apple and IBMdesktops had gradually improved their speed and powerand in 1987 they were more powerful than the best LISPmachines on the market. Overnight however, the industrycollapsed and billions of dollars were lost. The difficulty ofupdating and reprograming the expert systems, in additionto the high maintenance costs, led to the second AI winter.Investment in AI dropped and DARPA stopped its strategiccomputing initiative, claiming AI was no longer the “latestmode”. Japan also stopped funding its fifth-generationcomputer program as the proposed goals were notachieved.In the 1990s, the new concept of “intelligent agent” emerged[6]. An agent is a system that perceives its environmentand undertakes actions that maximize its chances of beingsuccessful. The concept of agents conveys, for the firsttime, the idea of intelligent units working collaboratively witha common objective. This new paradigm was intended tomimic how humans work collectively in groups, organizationsand/or societies. Intelligent agents proved to be a morepolyvalent concept of intelligence. In the late 1990s, fieldssuch as statistical learning from several perspectivesincluding probabilistic, frequentist and possibilistic (fuzzylogic) approaches, were linked to AI to deal with theuncertainty of decisions. This brought a new wave ofsuccessful applications for AI, beyond what expertsystems had achieved during the 1980s. These new ways

5 // Artificial Intelligence and Robotics Birth ofPROLOG BKG backgammon AI Locomotion,problem solving(Shakey, Standford) Man-computerSymbiosis Semantic nets(Masterman) First game playing IBM (Arthur Samuel)1950 Turing test Assimov’sthree laws19541958 Bayesian methods Unimationrobot for GM Rise of LISP DartmouthConference Rosenblatt’sPreceptron Marvin Minsky’sframes, schemesand semantic links David Marr’svisual perception ELIZA (interactivedialogue,ARPANET)19621966 Negative reporton machinetranslation MacHack (Chess) Dendral expertsystems1970 Perceptronbook (Minskyand Papert’s)19741980 Birth ofLISP MYCIN Medicaldecision supportsyste SHRDLU Languageunderstandingand robotics DARPA cutsin academicresearch on AI ParallelComputingFigure 1.A timeline highlighting some of the most relevant events of AI since 1950. The blue boxes represent events that have had a positive impact onthe development of AI. In contrast, those with a negative impact are shown in red and reflect the low points in the evolution of the field, i.e. theso-called ”winters” of AI.of reasoning were more suited to cope with the uncertaintyof intelligent agent states and perceptions and had its majorimpact in the field of control. During this time, high-speedtrains controlled by fuzzy logic, were developed [7] as weremany other industrial applications (e.g. factory valves, gasand petrol tanks surveillance, automatic gear transmissionsystems and reactor control in power plants) as well ashousehold appliances with advanced levels of intelligence(e.g. air-conditioners, heating systems, cookers and vacuumcleaners). These were different to the expert systems in1980s; the modelling of the inference system for the task,achieved through learning, gave rise to the field of MachineLearning. Nevertheless, although machine reasoningexhibited good performance, there was still an engineeringrequirement to digest the input space into a new source, sothat intelligence could reason more effectively. Since 2000,a third renaissance of the connectionism paradigm arrivedwith the dawn of Big Data, propelled by the rapid adoptionof the Internet and mobile communication. Neural networkswere once more considered, particularly in the role theyplayed in enhancing perceptual intelligence and eliminatingthe necessity of feature engineering. Great advances werealso made in computer vision, improving visual perception,increasing the capabilities of intelligent agents and robotsin performing more complex tasks, combined with visualpattern recognition. All these paved the way to new AIchallenges such as, speech recognition, natural languageprocessing, and self-driving cars. A timeline of key highlightsin the history of

definition used is one proposed by Russell and Norvig: “Artificial Intelligence is the study of human intelligence and actions replicated artificially, such that the resultant bears to its .

Related Documents:

Arti cial Neural Networks Human and Arti cial Neurons Arti cial Neurons ANN Artchitecture A simple neuron Arti cial neuron is a device with many inputs and one output Two modes: Training Using Firing Ruledetermines when a neuron should re. Are very important in neural networks and accounts for their high exibility

CHAPTER 1. INTRODUCTION Arti cial neural networks (https://www.merriam-webster.com) de nes as arti cial intelligence: 1. A branch of computer science dealing with the simulation of intelligent behavior in computers;

The Future of Robotics 269 22.1 Space Robotics 273 22.2 Surgical Robotics 274 22.3 Self-Reconfigurable Robotics 276 22.4 Humanoid Robotics 277 22.5 Social Robotics and Human-Robot Interaction 278 22.6 Service, Assistive and Rehabilitation Robotics 280 22.7 Educational Robotics 283

Pour l’année 2018, la Conférence Nationale d’Intelligence Arti cielle et les Rencontres des Jeunes Chercheurs en Intelligence Arti cielle sont organisées conjointement. Elles ont lieu du 4 au 6 juillet 2018 à Nancy sur laPlate-Forme Intelligence Arti cielle (PFIA). Lors de la même semaine, la plate-

2. Chapter 3 reviews the state of the art of Arti cial Intelligence applications in wind energy. The prospects and challenges of applying Arti cial Intelligence in the wind energy sector are discussed in Chapter 4. Finally, some conclusions are given in Chapter 5. The funding from DTU Wind Energy is greatly appreciated. I also want to thank my .

A robust layered control system for a mobile robot. IEEE Journal of Robotics and Automation, 2(1):14{23. Brooks, R. (1991). Intelligence without representation. Arti cial Intelligence, 47:139{159. Cambridge Dictionary (2006). Cambridge online dictionaries. Dudek, G. and Jenkin, M. (2000). Computational Principles of Mobile Robotics.

Cho and Kasa, fiValidation.fl J. Bullard Discussion Parameters versus models Connections to escape dynamics Model validation Speci–cation testing Assignment of the PLM Nature of the validation dynamics Instability generated by rival models Arti–cial intelligence Comparison with arti–cial intelligence Statistical versus economic .

CS 188 Spring 2015 Introduction to Arti cial Intelligence Midterm 1 You have approximately 2 hours and 50 min