Future of Human like Machines
Abstract: The field of robotics is developing rapidly which makes the gap between the human beings and the robots to be reduced. Even though scientists have done marvelous researches in robotics, still we are far behind from making robots with the human intelligence and the natural features. This review paper discusses about the robots and their future; also how the artificial intelligence has been attached to machines. What will be the future of the man and machine interaction is another topic discussed in this paper.
1. Introduction
Robotics is the science and technology of robots. Robotics requires a combination of electronics, mechanics and computer science. This field has developed a lot during the last few decades, and will be a very close subject to our lives in the near future.
Science fictions introduce human like machines as servants of mankind. Those machines gradually become more and more intelligent as they try to get the ruling power of humans and try to rule the planet and whole universe. That will become the biggest threat to the humans. That is how the science fictions say. Will that be a reality? What will be the future of those human like machines? Do not forget many of the predictions done by the science fiction writers have become reality.
As the science and technology improves, human like machines are becoming so close to our lives. This is the time period where those machines are servants of human beings. Because of those machines our lives are becoming so easy. So many new things in this field are invented by the scientists. Machines are getting more and more intelligent as the artificial intelligence is improving. It is really hard to predict what will be the future of the robotics.
This paper discusses particularly about the social robots. Social robot is the name given to robots that are made to interact with human beings and to support humans in some of their tasks. How the artificial intelligence is attached to machines in order to make them intelligent, is another area discussed in this paper. To build systems that can reach the brain level of humans is the main problem of artificial intelligence since its birth. As the artificial intelligence improves, robots will have more and more intelligent level, which will then be more than the humans. Then some argue that there will be a problem when robots interact with humans in the society. What should be the limitations of the robots and what moral ethics should the robots have are some other area that is discussed by this paper.
The rest of the paper is structured as follows: in Section 2, an overview to the robotics is given. In that section new trend in robotics is discussed, that is humanoids. Section 3 is dedicated to present the concept of social robots. In section 4, about the artificial intelligence and how it is used to transform smart machines to intelligent once is discussed. Section 5 considers about the human-machine interaction and the issues within social robots. Finally, in section 6 some conclusions and guidelines for future work are presented.
2. Overview of Robotics
The word robot can refer to both physical robots and the virtual software agents which have the artificial intelligence. The whole world will be full of robots and most of our activities will be controlled automatically by the intelligent robots. Sometimes their brain will be much more advanced than the human brain in the future. Therefore, it is important for all of us to be aware of the future of human like machines.
Robotics is a combination of different fields like mechanics, electronics and computational intelligence (Artificial intelligence). Because of that studying about robotics will help to gain a lot of knowledge about those areas. And also robotic is interrelated with Information Technology. Especially, it is very interesting to get the knowledge of how the robots work and about the electronics side. One field that is mostly emphasized on this paper is how the robots get their artificial intelligence and how the development of artificial intelligence will help to make the future of the robotics.
The newest trend of robotics is to make a human like machine which is very much similar to human beings. These robots are known as humanoid robots or anthropomorphic robots. If a robot is to be called a humanoid it has to have some features. It has to be self maintainable. It has to do autonomous learning. It should not be troublesome to the human beings and the environment. A robot which satisfies all these conditions has not been developed yet. But scientists are working on achieving that goal. As a result of that they have developed so many robots that are very close to human nature. American Doctor Sinthiya Brisil’s Kismet is a robot which can be called as a humanoid because it can react and express its feelings [4]. And there is another type of robots which are not similar to human beings physically, but doing a great service to the humans. Those robots are known as social robots. AIBO robot (2000) is a social robot which has the advantage of being biologically-inspired [6].
Humanoid robots can be classified in to three categories. Research, Consumer Products and Fiction are those [4]. Research robots are created in educational and industrial research laboratories, and Pearl is an example. Robots for sale purposes such as ASIMO robot are in the Consumer Product category. Fiction robot category consists of robots from television and films. Japanese Professor Hiroshi Ishiguro has recently developed a robot named as Repliee in his Intelligent Robotics Laboratory which is very similar to a young woman. As it is the first robot to be so similar to a human body, when it is sitting, talking and expressing its feelings you would find it is so difficult to differentiate it from a living woman. This is just one example of the amazing, exiting and ever improving world of robots.
The importance of robotics comes with so many fields. Especially, we are using robots to do the tasks that human beings can not do. Traversing around the endless space discovering new planets, going through the core of the earth, getting into terrorist areas as military weapons, radioactive waste clean-up are some of those activities which are too dangerous for humans. Robots are used in those activities. And they are used in tasks that should be performed with more precision such as automobile production which will take so much of time if it is to be done by humans. Medicine, astronomy, military are some of the fields where robots are important apart from doing the normal routine work of humans. An American scientist Raymond Kerswill says by the year 2029 the human like machines will be able to reach the intelligent level of human beings. Studying about robots and the future of them will be such an amazing journey.
3. Social Robots
A Social Robot is the kind of robots who interacts with human beings to serve them in some of their activities. Social robot should be an autonomous robot which can interact with humans or other physical agents (other social robots). A social robot should have a physical embodiment; otherwise it shouldn’t be called as a social robot. B. R. Duffy and others have introduced the Social Robot Architecture which is an agent based system itself [1], to develop social robots with physical embodiment and sociality. Concept of embodiment, sociality, agent based robotics, concepts of a social robot and the development of the Social Robot Architecture are the main topics explained in this below part in this paper.
Concept of embodiment says that social robots must have a bodily representation to get inputs from the environment and give outputs to the environment while interacting with human beings. Even though some argues that artificial intelligence and the body of robots are completely different things, Johnson and Lakoff argue that our ability to understand and reason abstractly depends heavily on our bodily experience which makes high level intelligence depend on embodiment [5]. Robots which display their heads in the computer screen cannot be considered as a social robot and they are in the base of being a robot. Social robots should have sensors to get the input and motors and physical hardware to react to different situations.
Sociality means the ability to create interactive relationships. The robots face the question of being social. An agent based robot should be able to communicate with other robots or humans. A social robot should be aware of their motives, beliefs, desires and intentions and should be able to analyze behavior of others. The social intelligence hypothesis or Machiavellian intelligence hypothesis considers for the development of artificially intelligent robot it should be embodied both in a physical environment and in a social environment. Mutual understanding between each other in a group is required for the establishment and continuation of the group and it is equally applied to the robot society also.
An agent is a system which observes and acts upon the environment, which can either be a software system or a physical robot. An agent based robot consists of physical devices which gather information through sensors and acts through actuators.
Duffy and others have described societal robotics and social robotics as two different things [1]. Integration of robotic entities to society is societal robotics; allowing or empowering robots to work to a goal while interacting with each other in a complex, dynamic and social environment is social robotics. They believe service robots which do a one task represent the societal robots. Social robotics deals with the multi-robot interaction that combines communication, co-ordination, communities, identity, relationships, and reactive or proactive behavioral models.
The researchers have developed the Social Robot Architecture which answers the key issues in developing social robots. The four layers of that architecture are physical, reactive, deliberative and social [1]. Motors to react, sensors to gather information and the physical appearance of the robot are considered in this layer. Series of fundamental behaviors are implemented at the reactive layer. Deliberative layer comprises of a Belief-Desire-Intention (BDI) architecture developed using Agent Factory. This layer deals with converting agent events into beliefs, storing the knowledge and providing when it is necessary. In the social layer agents can interact using an Agent Communication Language (ACL) called Teanga. This architecture facilitates the development of teams of robots of different kinds and the robots with different sensors and actuators can be readily integrated into the system.
4. How AI transform smart machines to intelligent machines
Artificial intelligence can be defined as implementing human intelligence on computer like machines. AI is the branch of computer science that deal with writing computer programs which can solve problems creatively or which have the power of pattern detection, reasoning and planning etc. Due to the development in the computer science during the last few decades, it is obvious that the artificial intelligence is no more just a dream. But building of intelligent machines that reach human intelligence, still remains as the biggest challenge of artificial intelligence since its birth. What are the reasons for us to be still far from having human like intelligent machines is one question that is addressed by the below section of this research paper. Different learning strategies and context types involved in learning process are explained below. Raducanu and Vitria have presented a study for cognitive development in a social robot through an incremental learning algorithm applied to the problem of face recognition [2].
Even though, now machines can do remarkable things like beating master chess players, working as troops in the battlefield, going into space etc, they still cannot reach the human brain level by doing things like interpreting objects in an image, answer a random question etc, which can be done by a normal person. This paper explains that lack of high-skilled people and necessary tools is not the reason for this matter. This failure is due to the fragmented study of artificial intelligence and due to the lack of consideration of the cognitive factors which are responsible for generating intelligent behavior. The cognitive perspective differs a lot from the classical approach of intelligent systems. Cognitive perspective considers about a system which learns continuously and contextually by itself while finding online resources, whereas in classical paradigm programmer give everything to the system.
An intelligent machine should be able to gather some knowledge in order to improve its behavior over time. That is what meant by learning. There is a very close relationship between data acquisition, knowledge representation and learning strategies. Symbolic representation, behavior-based systems and the cognitive approach are the learning strategies considered in this paper.
Symbolic representation is the traditional model of artificial intelligence which does a reasoning process by the inference engine based on the knowledge base built by the programmer as a “world-model”. Behavior-based system was developed to overcome the limitations of the symbolic representation. A system’s behavior could be represented by the process of converting its input data to output, without the need of a centralized representation (knowledge base). In this model high level behaviors were obtained from a set of low level behaviors formed by a set of agents that would interact with each other. In cognitive approach of learning the system will learn new concepts based on previous data and the new experience gathered over time. In cognitive learning the knowledge is in long-term memory and in short-term memory. Context is discussed in this research paper as part in cognitive learning process. Context is referred to as the history of what gathered as knowledge over time while doing interactions, which is in the space of shared knowledge.
As a case study of cognitive development of robots Raducanu and Vitria have introduced an incremental learning applied to the problem of face recognition for a social robot [2]. This study has been performed using an AIBO robot which represents the class of social robots. For a social robot it is very important to identify the faces of human beings and identify their feelings. To carry out the study Raducanu and Vitria have implemented the incremental LDA (linear discriminant analysis) algorithm named as IncLDA. IncLDA algorithm was tested by using about 6000 images of persons taken by the camera of AIBO robot [2]. The common recognition rate achieved was around 95% which is a quite a good result. This cognitive model could be developed to also have the forgetting ability, which enables to eliminate items from the database.
5. Human-machine interaction and fundamental issues in social robotics
The development of robotics due to the development of computer science and engineering has made the gap between the human being and a machine to be reduced by a considerable margin. But still the scientists haven’t been able to build machines which are up to the intelligent level of humans. The below section on this research paper considers about the relationship between the humans and robots. Up to what extent the robots will be able to reach the physical nature of human body is another question answered by this. Even though there is a fear that robots will be able to go beyond the humans and be a threat to humans, scientists are still trying to improve the abilities of robots. Duffy has discussed a number of fundamental distinctions between robots and humans. Also the ways to address those issues are discussed [3].
The fundamental difference between man and physical humanoid robotic systems is that of existence. That is discussed in this paper as the body dilemma. The difference between the animal systems and the mechanical systems is explained by Duffy using the concepts of autopoiesis and allopoiesis [3]. Autopoiesis means self-creating or self-producing. Allopoiesis means a concatenation of processes. The animal systems are autopoietic and the mechanical systems are allopoietic. The components of a natural system can grow and evolve, ultimately growing from a single cell or the mating of two cells. To reduce this gap between the machines and the natural systems, machines should be able to learn and adapt at hardware level as well as in software level. Evolvable hardware comes in to act with this matter. But the practical reality is that in order to realize a physical robotic system, a set of actuators, sensors and control mechanism has to be integrated, which makes it is impossible to be a real biological system.
The ultimate target of the robotics is to realize a humanoid robot that is incapable of being perceived as different from a normal human being. But should it be the ultimate goal of robotics? Can building a mechanistic digital synthetic version of man be a cheat when man is not mechanistic, digital nor synthetic? Anthropomorphism means giving the objects the human form. But a successful social robot may have to have both anthropomorphism and the maximum functionality required. In order to maximize the social interactions robots should be able to diagnose the human’s emotional state.
Expressing their intentions, being aware of their selves and others, and the power of making free choices unconstrained by external agencies are important features associated with human beings, which should be possible to realize those features in a machine. Duffy argues that it becomes a less of an issue whether the machine actually has these properties or whether they appear to have them [3]. This makes social robots to have emotional capabilities.
In the near future man will face the dilemma of integrating the machines into our society. Obviously machines will have more decision making power where there is large amount of data to be processed. But will they have the instincts to make decisions in the society? The problem arises when there is a conflict between the natural instincts of human and the logical solutions of the robots.
Moral rights and duties are the next question faced in the human-machine interaction. Duffy explains about the issue of moral rights from two perspectives. First is whether a machine should be programmed to be morally capable when interacting with the humans. The second perspective is whether it is necessary to have human capabilities to be able to assess morality. If the robots are accepted as just machines then moral rights and duties won’t come in to act, but when the robots come into the society of humans with the features and responsibilities of human then the moral rights are so important.
The goal of scientists should be to take more advantages of robots while empowering them, without just constraining them thinking that they may take over the world. Robots should not be considered as just some machinery, but it is a role.
6. Discussion
Robotics is the science and technology of designing and developing robots. Robotics requires knowledge with a combination of electronics, mechanics and computer science. The word robot can refer to both physical robots and the virtual software agents which have the artificial intelligence. Social robot is the name given to robots that are made to interact with human beings and to support humans in some of their tasks. The newest trend of robotics is to make a human like machine which is very much similar to human beings which are called as humanoid robots or anthropomorphic robots. Artificial intelligence is attached to machines in order to make them intelligent. To build systems that can reach the brain level of humans is the main problem of artificial intelligence since its birth. As the artificial intelligence improves, robots will have more and more intelligent level, which will then be more than the humans. Even though, now machines can do remarkable things we haven’t been able to make machines which have features of natural systems and which are up to the brain level of humans. As a case study of cognitive development of robots Raducanu and Vitria have introduced an incremental learning applied to the problem of face recognition for a social robot which has proven to be very effective. They have implemented the incremental LDA algorithm named as IncLDA. IncLDA algorithm was tested by using about 6000 images of persons taken by the camera of AIBO robot. The ultimate target of the robotics is to realize a humanoid robot that is incapable of being perceived as different from a normal human being. In the near future man will face the dilemma of integrating the machines into our society. Obviously machines will have more decision making power where there is large amount of data to be processed.
As further work I would like to do some more study on this area. I would like to prepare a review paper with more details particularly regarding humanoids, and also about the engineering part of robotics.
Acknowledgement
Mr. B. L. D. Senevirathne as my supervisor for this independent study helped me a lot to prepare this review paper. And Prof. Asoka Karunananda gave the idea for me to select this area for the study and his guideline for the preparation of the review paper was helpful to me. I would like to thank them all and also to my friends who helped me in finding relevant information.
References
[ 1 ] What is a Social Robot? By B.R. Duffy, C.F.B. Rooney, G.M.P. O’Hare, and R.P.S. O’Donoghue
[ 2 ] Learning to Learn: From Smart Machines to Intelligent Machines by B. Raduncanu and J. Virtria
[ 3 ] Fundamental Issues in Social Robotics Brian R. Duffy
[ 4 ] All Robots Are Not Created Equal: The Design and Perception of Humanoid Robot Heads by Carl F. DiSalvo, Francine Gemperle, Jodi Forlizzi and Sara Kiesler
[ 5 ] Metaphors We Live By, by Lakoff G. and Johnson M.
[ 6 ] AIBO robot, http://www.sony.net/Products/aibo/index.html
[ 1 ] What is a Social Robot? By B.R. Duffy, C.F.B. Rooney, G.M.P. O’Hare, and R.P.S. O’Donoghue
[ 2 ] Learning to Learn: From Smart Machines to Intelligent Machines by B. Raduncanu and J. Virtria
[ 3 ] Fundamental Issues in Social Robotics Brian R. Duffy
[ 4 ] All Robots Are Not Created Equal: The Design and Perception of Humanoid Robot Heads by Carl F. DiSalvo, Francine Gemperle, Jodi Forlizzi and Sara Kiesler
[ 5 ] Metaphors We Live By, by Lakoff G. and Johnson M.
[ 6 ] AIBO robot, http://www.sony.net/Products/aibo/index.html
No comments:
Post a Comment