Human Super Intelligence

Transcription

1 Int. J. Emerg. Sci., 1(2), , June 2011 ISSN: IJES Human Super Intelligence Alexander A. Antonov Research Centre of Information Technologies TELAN Electronics, P.O. Box 73, Kiev, 03142, Ukraine. Abstract. The manuscript will be interesting for scientists in any field, as it explains how human super intelligence can be implemented in the near future and enable people to successfully solve multi-factor tasks currently impossible for human intelligence. Indeed, all dependencies discovered in all sciences so far, with a rare exception (e.g., the third Kepler s law with Newton s correction), are described with functions of no more than three variables, as people think with visual (therefore, not more than three-dimensional) images. But certainly, there are numerous phenomena and processes in nature which are described by functions of a larger number of variables. New knowledge obtained with the help of human super intelligence will allow significantly improving the life quality, solving regional and global environmental issues, solving the energy supply problems and many others. It is necessary to develop the human multi-factor intelligence systems based both on super computers and personal computers. Keywords: Intelligence, Thinking, Knowledge, Technological Singularity, Super Intelligence, Super Knowledge 1. INTRODUCTION 1.1. Origination of the problem Shortly after computers were invented, people wanted to teach them to think. The first goal that was set was to teach computers what humans learn in the first year of their life: to speak and understand human speech, to recognize people, to identify the things and animals around them, to maintain equilibrium and to walk, as people believed that these were the simplest intellectual tasks. Later on, the list was supplemented with the task of translating from one language into the other, and some other tasks. As for the more complicated intellectual tasks performed by adults, they were supposed to be solved later. These tasks made up the subject of investigation of the so-called artificial intelligence (AI). Although huge amount of work has been done over the past 50 years, and sometimes it was even successful, the above primary tasks have not been solved yet. Indeed, there is no computer which could tell a cat from a dog, machine translation leaves much to be desired, etc. These results lead us to believe that the initial goal definition was somewhat wrong. 164

2 Alexander A. Antonov 1.2. Research problem Therefore, the task of the given paper is to demonstrate that: the initial problem definition of the AI was incorrect, as it was supposed to solve human intellectual tasks (which are mostly multi-factor) with the more primitive low-factor computer means: the definition of the concept of technological singularity as a result of successfully completed solution of the AI problem is untenable, because it is impossible to solve the multi-factor intellectual human problems with the low-factor purely computer means; numerous problems faced by people and requiring a solution are mostly multi-factor, and, therefore, can be solved only using the systems of humancomputer super intelligence based on human multi-factor thinking; moreover, they can be solved already in the near future. Comprehension of the correct problem definition for the development of super intelligence will allow eventually understanding the actual ways of solving multifactor tasks, and solving them quickly. 2. HUMAN INTELLIGENCE The mistake in target setting is accounted for by the fact that, when defining the problem of AI, there was no clear understanding of what intelligence is. Moreover, it is still not understood what intelligence, both human and computer, actually is [1], [2], [3], and there is no generally acceptable definition of it. Thus, let us first of all give some consideration to human intelligence. We shall refer to human intelligence as a scope of all types of human thinking. That is, we shall include into intellectual human activity not only rational thinking, but also emotional thinking, unconscious thinking, intuitive thinking, and automatic control of biological systems. This definition in no way intends to be universal, and does not take into account all peculiarities of the notion. It just gives a more precise definition of an engineering problem we shall speak about later. Rational low-factor thinking, which, in the opinion of many, is the only type of human thinking, helps the people to learn the outside world and make quick decisions ensuring their survival. Rational thinking is based mainly on the analysis of visual images, and is, therefore, not more than three-dimensional. It is not able to understand objects or processes having a larger number of dimensions. In order to make sure this is true, try to imagine, for instance, a four-dimensional cube (or, in other words, the hypercube, tesseract, octachoron). Subconscious or unconscious multi-factor thinking is the main type of human thinking, although, as follows from its name, people are usually not aware of it, as this thinking is performed mainly in sleep. Actually, people sleep in order to make it possible for the unconscious thinking to complete its job. In sleep, nothing disturbs people from making a powerful multi-factor analysis of information accumulated in the wakeful state, regarding both the outside world and the state of one s own body. 165

3 International Journal of Emerging. Sciences, 1(2), , June 2011 As for rational thinking, it merely allows formulating the results of multi-factor unconscious thinking with the help of low-factor (not more than three-dimensional) images, which people can exchange with each other. However, multi-factor thinking is used in the wakeful state, as well. For instance, when having a conversation, no one formulates his or her thoughts due to a conscious solution (i.e. with the help of rational thinking) of logic-mathematical problems relevant to the topic of the conversation. In fact, the rational low-factor thinking of a vigil person constantly interacts with the multi-factor unconscious thinking and is based on it. The utmost importance of the overall multi-factor human thinking can be illustrated with an old Sufi parable about the three blind sages and an elephant. It reads that, wishing to understand how the elephant looks, the first wise man touched the animal s leg and said that an elephant is like a trunk of a large tree. The second one felt the side of the elephant and said that an elephant resembles a wall. The third sage touched the elephant s tail and said that an elephant is like a large snake. Similarly, it is very difficult for people to understand the essence of multi-factor (more than three-factor) objects and processes surrounding them with the help of only the rational thinking. In the case when it is impossible to realize the results of the unconscious multifactor thinking in the form of low-factor images, a person in the wakeful state is simply prompted to do something without any reasons with the help of emotional thinking (usually solving the day-to-day routine tasks) and intuitive thinking (solving in the exceptional cases the life-and-death problems). And, finally, all day round the human system of automatic control of complicated biological systems works, including blood circulation, digestion, immune system, respiratory system, etc. Thus, vital activity of a human being is provided for by a whole number of different and complicated types of thinking, which solve different tasks. People remember the results of all these types of thinking and accumulate the knowledge, both the conscious low-factor and the unconscious multi-factor. 3. COMPUTER INTELLIGENCE 3.1. Artificial intelligence All the foregoing allows explaining the difficulties that arose when solving the tasks initially set for the problem of artificial intelligence. The matter is that, as it turns out, a child in the first year of his life, even when he is awake, uses only the multifactor unconscious thinking for self-learning. Thus, interaction with other people is of great importance for the intellectual development of children. Later on, as people get biologically older, they start using the simpler low-factor rational thinking, in addition to the multi-factor unconscious thinking. This is why, in particular, it turned out to be so difficult to turn Mowgli back to the human way of life. 166

4 Alexander A. Antonov Thus, naturally, the significantly simpler low-factor thinking, both human and computer, is unable to comprehend the results of the much more complicated multifactor thinking. As it turned out, the initial tasks set for the AI involved the solution of multifactor intellectual tasks with the computer low-factor means. This is why they have not yet been solved, and will not be solved any time soon. They may never be solved, as in the process of its development the humankind will eventually shift to the use of other cognition tools, more perfect than the AI. Nevertheless, research of the AI has become a powerful incentive to a deeper understanding of human intelligence. Human intelligence has actually been the key object of investigation within the AI problem. Comparison of computers and human brain allows better understanding of the latter. Comparison of human and artificial intelligence allows for better objective evaluation of computer achievements based on the main criteria, which is the degree of intellectuality, and not the separate technical parameters, however brilliant they seem to engineers. Therefore, the fact of comprehending and formulating the problem of artificial intelligence should be considered the greatest achievement of human mind, disregarding the practical results obtained. At the same time, we cannot ignore the certain anti-humane aspect of the AI systems, developed, in fact, to the detriment of the people, since their task is to solve intellectual tasks instead of people, and, therefore, objectively oust people out of the intellectual activity. As for practical results, after the failures in solving the seemingly simplest human tasks initially set for the GOFAI (Good Old-Fashioned Artificial Intelligence), which have not been solved until now [4], [5], [6], [7], the scope of tasks considered within the problem of artificial intelligence has been enlarged. Some of them are successfully solved now, as they can be given better formal description by the rational human thinking. Here belongs development of expert systems of real time, investigation of neural networks, studies in robotics, and some others. It would probably be useful to make a small commentary to the problem of artificial intelligence. As specialists in different fields of science are involved in the research, including electronic engineers, software developers, mathematicians, neurophysiologists, philosophers, psychologists, linguists, etc., there are various interpretations of this problem; they suggest different methods and means of solving it. For example, the representatives of the so-called neurophysiologic approach believe that first it is necessary to study the human brain, to understand the ways it solves different intellectual tasks, to find out what human intelligence actually is, and only then to develop technical devices emulating and improving the human brain. Adherents of the so-called cybernetic approach believe that technical devices improving the work of the human brain can and must be developed without emulating its work. They refer to the fact that, for instance, ground vehicles created by engineers use wheels, and do not emulate the movement of terrestrial living creatures using legs. 167

5 International Journal of Emerging. Sciences, 1(2), , June 2011 In other words, the criteria of simple and complicated in nature and in human machine civilization differ greatly, as the environment and conditions of their existence are different, and this circumstance cannot be ignored Technological singularity However, despite huge difficulties in practical implementation of GOFAI, rapid improvement of technical parameters of computers has led to the opinion that soon a computer civilization may emerge on our planet [8], [9], [10]. This scenario of AI development came to be known as technological singularity. This term is defined as hypothetical, at some point explosive acceleration of the scientific and technical progress having unpredictable results. Moreover, there are several scenarios of transition to the computer civilization: Super intelligence will be the result of improvement of computers; Self-developing information networks (e.g. the Internet) can develop into super intelligence; Intelligence of computer users will develop to the extent of super intelligence due to improvement of the human-computer interface; Nanotechnologies and bioengineering will improve the biological nature of the mankind to such an extent that humans will become super intelligent Computer super intelligence The definition of super intelligence as an alternative to human intelligence was first suggested in Ref [11]. In Ref. [12], [13] and others, possible ways of implementing computer super intelligence are considered. Ref. [14] states that computer super intelligence will definitely emerge by 2030, irrespective of the will of people. And this circumstance raises serious concerns [15], [16], [17]. Ref. [18] wittingly suggests that the first super intelligent computer will be the last invention of the mankind. The concerns mentioned above are caused by: the fact that computer super intelligence is very likely to develop; absence of any specific and persuasive positive scenarios of relationship development between the computer and human civilizations. However, the issue of the greatest concern is that basically we are dealing with complete misunderstanding of the computer super intelligence phenomenon and its frightening differences from human intelligence. Indeed, no one even makes any suppositions as to what these differences are. Therefore it becomes clear that we have lost control over the situation. The validity of the above can be easily verified if we try to answer the following questions: In what way will the new super knowledge differ from human knowledge? Is a man actually able to comprehend this knowledge if, supposedly, a friendly computer super intelligence (or any other friendly civilization) would be willing to share its super knowledge with the mankind? 168

6 Alexander A. Antonov And what is actually super knowledge? If the new biotechnologies and nanotechnologies are somehow able to improve the information characteristics of humans, the question is which ones and in what way? If the new human-computer interfaces are created, the question is how and in what way will they allow people to comprehend super knowledge? 4. HUMAN SUPER INTELLIGENCE So, does the human civilization have no way out, and it is doomed? No. To understand the answer, let us figure out how exactly super intelligence and super knowledge differ from the existing human intelligence and existing human knowledge. In our opinion: super knowledge is multi-factor knowledge (not yet obtained by people), i.e. knowledge of multi-factor objects and processes, including, for instance, biological nature of a man, the mechanisms of human illnesses, the mechanism of ageing, the global and regional economy, the weather and the environment, state and corporate governance, etc; super intelligence is the ability to comprehend the above mentioned super knowledge and use it in the best possible way (e.g., in practice, or by improving one s own intellectual abilities through the use of computer training, or in any other way). Thus, the prospects of further existence of human civilization depend only on the ability or inability of men to comprehend and use multi-factor information. Fortunately, as we have mentioned above, human intelligence is based on multifactor thinking. Therefore, it can serve as the basis of the new multi-factor human super intelligence, contrary to the above mentioned low-factor super intelligence. In order to realize this possibility, multi-factor super intelligence [19] should use not only human thinking, but computer resources [20], [21] as well, both in the form of computers and computer networks. For the primary discussion of the problem of multi-factor super intelligence, the variant of its implementation of the basis of super computers is more convenient as it allows omitting many insignificant technical peculiarities. Indeed, let us suppose for simplicity that we already have a super computer with the best possible features we can imagine. Let us also assume that it stores all the conscious knowledge accumulated by the mankind (which is currently only lowfactor). We shall not discuss when, how, on what basis and at what cost these results can be achieved. This is not relevant right now. What is relevant is the answer to the question how the human-computer super intelligence will be able to solve multi-factor tasks by developing multi-factor knowledge based on the available low-factor knowledge (let us recall the three blind sages and the elephant). Without trying to embrace the un-embraceable, hereinafter we shall discuss only the solution of scientific multi-factor tasks, as they are probably of the greatest practical interest and are most understandable. 169

7 International Journal of Emerging. Sciences, 1(2), , June 2011 Indeed, simplistically the process of scientific intellectual activity (or scientific creative activity) can always be clearly divided into two stages which are substantially different: detection of factors which significantly influence the process under investigation, and screening of irrelevant or insignificant factors; determination of mathematical (or other cause-and-effect) relation between the revealed significant factors. Meanwhile, the process under investigation may be both low-factor and multifactor. It depends on the problem statement, financing, the researcher s talent, competitive environment, and many other circumstances. And it turns out that in terms of solving low-factor problems with the help of rational human thinking, the first stage is the most difficult, i.e. detection of significant factors influencing the results of the process under investigation. Finding mathematical dependence between the revealed significant factors (e.g. based on the results of experimental studies) in this case is relatively a much simpler task. In terms of solving multi-factor problems, the complexity of performing the tasks corresponding to both stages increases to such an extent that solving such a problem becomes virtually impossible for a human. This is why in exact sciences, for instance, there are no laws, with only a few exceptions (e.g., the third Kepler s law with Newton s correction) described with functions of five and more variables, although such laws certainly exist in nature. However, laws described with functions of three or less variables are, most likely, almost completely learned. For a computer having the relevant hardware and software, the number of factors is rather a quantitative than a qualitative difference. This is why a computer, together with its user, is able to successfully analyze multi-factor processes. However, without human intelligence such intellectually demanding work of a computer is most likely impossible to be implemented in the foreseeable (quite distant) future. Thus, it should be admitted that the danger of computer civilization appearing in the nearest time is exaggerated. Furthermore, just like we need different types of business, both small and large, we need various types of scientific research, both teamwork and individual. Science having no alternatives, in the same way as business lacking competitive environment, cannot develop. Creative work is always inherently individual. Only on condition of making the most favorable conditions for realization of the researchers creative potential is it possible to develop the human-computer multifactor super intelligence. This is why the network implementation of human-computer super intelligences would be most suitable and easier accessible for the majority of scientists. That is, the human-computer multi-factor super intelligence should mostly be implemented by users based on their PCs and their own databases. The new information network suggested in [20] is the most convenient one for the development of this network implementation of the human-computer multi-factor super intelligence, as it offers its users the special computer intelligence service and guarantees complete information security. Therefore, this information network should be deployed as soon as possible, first in the form of regional implementations, and later on, as the demand for the services it offers increases, in the form of a global implementation. 170

8 Alexander A. Antonov 5. CONCLUSION Summarizing the results, to make the comparison of the artificial intelligence and the human-computer intelligence more convenient, let us list their differences in a table. Table 1. Сomparison of the artificial intelligence and the human-computer intelligence Differences The objective Artificial intelligence (leading eventually to the development of the computer super intelligence and the computer civilization). To develop the software allowing computers to independently solve intellectual problems better than people and, eventually, instead of people. Human-computer intelligence (resulting eventually in the development of the human super intelligence and the human-computer civilization). To develop the software allowing people to use their multi-factor thinking to the best efficiency with the help of computer resources. Possible result of meeting the objective Relations of the humans and the computers Peculiarities of intellectual activity Peculiarities of the tasks performed Knowledge obtained Significant restriction of the human ability to intellectual work (as software developers, engineers, doctors, etc.). Computers dominate, humans maintain them Computers process information with the low-factor means (i.e. the software developed to solve the problems comprehended by the low-factor human rational thinking). Use and develop the abilities of the rational human thinking. Low-factor knowledge being mostly of scientific interest and of very limited practical interest. Significant improvement of the intellectual level of the tasks successfully performed by people. Humans dominate and use computers Computers process information with the multi-factor means (i.e. the software developed to implement the humancomputer multi-factor tinking). Develop the abilities of the human multi-factor thinking. Multi-factor super knowledge being of significant practical interest, as it allows solving numerous problems medical, physical, economic, etc. It follows from the above that solution of the suggested problem of human super intelligence will result in the emergence of the new human-computer civilization in the near future. Within this civilization people would master new types of thinking. The main peculiarity of the new types of thinking would be their multi-factor nature. The new thinking would also allow mastering the multi-factor knowledge 171

9 International Journal of Emerging. Sciences, 1(2), , June 2011 which is now beyond human comprehension: physical, technical, medical, biological, economic, geoclimatic and other knowledge, and use it efficiently. REFERENCES 1. Eysenck HJ, Intelligence: A new look. Transaction Publishers, New Brunswick, NJ, Miles TR, On defining intelligence. The British Journal of Educational Psychology 1957, 27(3): Available at: 3. Suvorov VV, Intelligence is the axiom of psychic reality Herald of the Moscow State University. Psychology 1999, 14(3): Available at: 4. Dreyfus HL, What computers can't do: The limits of artificial intelligence, MIT Press, New York, Haugeland J, Artificial intelligence: The very idea. MIT Press, Cambridge, Mass, Luger GF, Artificial intelligence. Structures and strategies for complex problem solving, Addison Wesley, Newell A, & Simon HA, Human problem solving, Prentice Hall, Englewood Cliffs, NJ, Moravec H, Mind children: The future of robot and human intelligence. Cambridge, MA: Harvard University Press, MA, Moravec H, When will computer hardware match the human brain? Journal of Evolution and Technology 1998, 1: Available at: Vinge V, The coming technological singularity: How to survive in the post-human era, Paper presented in VISION-21 Symposium, NASA Lewis Research Center and the Ohio Aerospace Institute, Available at: Bostrom N, How long before superintelligence? Linguistic and Philosophian Investigations, 2006, 5(1): Available at: Legg S., Machine super intelligence, Unpublished doctoral dissertation, University of Lugano, Switzerland, Available at: Stoklosa T, Super intelligence, Nature 2010, 467: 878. doi: /467878a 14. Kurzweil R, The singularity is near: When humans transcend biology. Viking, NY, Bostrom N, Existential risk: analyzing human extinction scenarios and related hazards, Journal of Evolution and Technology, 2002, 9(1). Available at: Joy WN, Why the future doesn t need us, Wired magazine, 2000, April: Available at: html 17. Yudkowsky ES, Artificial intelligence as a positive and negative factor in global risk. Singularity Institute for Artificial Intelligence, CA, Palo Alta, Available at: 172

10 Alexander A. Antonov 18. Good IJ, Speculations concerning the first ultra intelligent machine, Advances in Computers 1965, 6: Available at: Antonov AA, Human-computer super intelligence, American Journal of Scientific and Industrial Research 2010, 1(2): doi: /ajsir Dix A, Finlay JE, Abowd GD, and Beale R, Human-Computer Interaction,, Prentice- Hall, NJ, Sears A, & Jacko JA, The Human-Computer Interaction Handbook: Fundamentals, Evolving Technologies and Emerging Applications Taylor & Francis, Antonov AA, Safe global/regional informational network, European Journal of Scientific Research 2009, 28(1): Available at: 173