Sentience in Robots: Applications and Challenges

1 Trends & Controversies661094-7167/01/$ 2001 IEEEIEEE INTELLIGENT SYSTEMSWho Wants a Nosy Butler?Paolo Dario, Scuola Superiore Sant Anna,Pisa, ItalyFor years, AI research has focused ondeveloping entities with intelligence com-parable to that of humans that is, with thecapability of reasoning and managingknowledge. The Turing test demonstrateshow researchers have pursued machineintelligence. Recently, AI research haschanged its focus. Recognizing (as RodneyBrooks at MIT first proposed) that interac-tion with the physical world is critical todeveloping intelligence, many researchgroups have tried to develop humanoid bodies. According to this approach,human-like intelligence relates not only toreasoning but also to learning, perceiving,and interpreting the physical world and tointeracting with the world and goals are much more difficult toimplement in machines than pure reason-ing.

2 As Tommaso Poggio pointed out,humans have developed reasoning rulesonly in the last few millennia; perceptionseems so natural to us because nature hasrefined it over millions of years. In fact,we ve accomplished what was AI s biggestchallenge for decades defeating a humanchampion in the game of chess. AI s new,much more challenging goal is to develophumanoid robots that can play, and possi-bly win, against a human soccer when AI is connected to robotics which provides a physical body formachines its goals are still to developentities that can autonomously perceive,reason, learn, and act, thus exhibitinghuman-like aspects of intelligence. Butshould these entities really be autonomous?Real-world Applications of robotics impliesusing robots in close interaction withhumans. In this nonabstract context, intelli-gence means interacting with humans andtheir environment safely, effectively, andpleasantly.

3 In most cases, this interactionrequires cooperation between the robot andhuman, rather than autonomous in rehabilitation roboticshave analyzed in depth the need for semiau-tonomous robots and human robot coopera-tion in real-world Applications . My grouphas obtained experimental evidence that theacceptability of personal robots (robots thatassist humans) is lower when the system isautonomous, especially in everyday tasks atSentience in Robots: Applications and Challengesand user-friendly robots. We also thoughtit would be instructive to know what themost experienced and distinguishedresearchers from the intelligent-roboticsfield think are the remaining first three contributions comefrom well-known roboticists. Paolo Dariodiscusses the acceptability of personalrobots in human societies.

4 He concludesthat human robot cooperation shouldreceive particular attention, keeping inmind the different abilities of humansand robots. Robin Murphy, activelyinvolved with using robot teams forurban search and rescue, discusses theneed for various levels of semisentiencein this context. Pete Bonasso, who hasworked in space robotics for a long time,believes the need for Sentience is a func-tion of how closely the robot must workwith humans while carrying out the task. Since its early stages, AI has been con-cerned with knowledge representation,reasoning, and problem solving. Thesetopics make up the core of what is oftencalled classical AI. For various reasons, thegeneral-purpose techniques of this long-standing AI tradition are seldom found inrobots. We were, therefore, curious toknow if researchers still believed thatthese techniques applied to robots.

5 Bern-hard Nebel, a theoretical AI researcherwho leads a champion RoboCup team,thinks that such representation and rea-soning techniques will eventually findtheir way into robots. He also discusseswhy this is taking so long to occur in awidespread fashion. And Nils Nilsson,three decades after building Shakey,thinks that logic-based representationand reasoning will be an essential com-ponent of sufficiently sentient robots although different variants might berequired for different kinds of tasks. Rationality, emotions, and what else?For Rodney Brooks, the real challengenow is developing robust methods forvisual object recognition, a skill that iseffortless for children but very difficultfor robots. He argues that this ability isessential for any reasonable understand-ing of the world. Lu s Seabra Lopes & Jonathan H.

6 ConnellThe articles in this special issue on semisentient robots provide an excellent viewof current research. However, we thought you might also appreciate views fromother researchers and practitioners directly involved with robotics Applications , to findout what they think about the applicability of this new generation of more adaptablehome. Who feels confident with or likes an autonomous machine that moves aroundthe house and makes decisions? Further-more, why should we let a robot waste efforttrying to recognize our favorite drink whenwe could help it (and our safety) by simplyconfirming that the object it is painfullytrying to identify is really a bottle?If the goal is to develop robots that canparticipate in human society and move andact in domains as complex as human envi-ronments, these robots should be able tocommunicate with humans using human-friendly modalities.

7 They should possesssome form of reasoning capabilities and beable to learn their user s preferences andadapt to the changing environment. Thismeans research must consider the differ-ences between natural and artificial intelli-gence and develop machines that can syn-ergistically exploit such complementarityin their interaction with Doesn t MeanBeing a Team MemberRobin Murphy, University of South FloridaOur work with rescue workers has led to asurprising discovery about intelligent robotsfor urban search and rescue. Although USAR robots need intelligence to operate in thecomplex, wireless-hostile environment of acollapsed building, it s not clear that the res-cuers need robots to be recognizably approach to USAR robotics concen-trates on giving the robot the ability to workin the existing technical-rescue-team struc-ture.

8 In this approach, the robot s relation-ship with humans is similar to that of a res-cue dog only one or two humans who havebonded with the robot care for and handle a relationship presumes recognizablesemisentience humans treat the robot as anintelligent entity or near , one motivation for USAR robots is that there aren t enough confined-space rescue workers certified for entry intoa collapsed structure to respond to a majordisaster. Robots that less-specialized rescueworkers can operate could alleviate the man-power problem. The catch with the robots-as-a-tool scenario is that a fire rescue workermight receive only 30 minutes trainingevery few months on a particular rescue , the rescue worker treats the robot as ageneric tool, not a near peer, and up to 10different workers might use a robot in a , the workers are unlikely to use thesame robot each shift.

9 There s no advantagein having a social bond, such as is inducedwhen a human recognizes a robot as semi-sentient. Clearly, we need semisentience for boththe robot -as-a-peer and robot -as-a-tool toperform their task competently. It is thehuman s perception of the robot s sentiencethat is different. To be treated as a teammember, a robot must be able to communi-cate naturally, explain itself, and adapt tohumans but precisely because the robot issemisentient, it can also take advantage ofhumans willing to adapt to it. To be a usefultool, a robot must be able to accomplish thesame tasks but with a wider range of users,who have less motivation and time to coad-apt to a the end, robots may have to be muchsmarter to appear comfortably Language BeatsProgrammingPete Bonasso, Johnson Space Center, NASAS entient robots (those capable of not onlyperception but also feelings) are useful onlywhen they work in close proximity withhumans.

10 Perhaps someday we ll show thatemotional biasing of robot behaviors is themost efficient control paradigm. Until then,the only real value of sentient robots is tocommunicate better with humans. Humansunderstand each other more completely byusing communications channels beyondlanguage such as speech prosody, facialexpressions, and body language. So, it seemsonly reasonable to endow robots with morethan just the ability to interact through menusor simple spoken commands. Given that theemotional model and its subsequent outputaccurately reflect the robot s state and itsunderstanding of the world, human robotteam efforts should prove more efficient. Forspace Applications , examples of such robotsmight be the control computers on the spacestation (the present-day incarnation of theHAL 9000) or human rover teams exploringand mapping planetary surfaces (the year2015 version of Star Wars C3PO).