"Can Artificial Intelligence replace the Engineer?" is not a simple question. Such a question subtly wraps itself both with semantic ambiguities and, at deeper levels, with several problems that philosophers have attempted to solve throughout history, especially the classic "mind–body problem." To aid the analysis of such an intriguing question, it is helpful first to identify precise meanings of the words involved and then to identify two distinct interpretations of the prompt. These two interpretations, interestingly enough, mimic the classic philosophical dualism of form versus function. By looking at each interpretation separately, the analyses will reveal support for a two-fold response to the prompt: (1) Artificial Intelligence is indeed capable of replacing the Engineer; but (2) such a replacement cannot occur in today's society.
The first task at hand is to identify specific meanings of the phrases in the prompt: 'the Engineer,' 'replace,' 'Artificial Intelligence,' and 'can.' 'The Engineer,' for purposes of this analysis, refers to the global group of humans who hold jobs in engineering. Engineers commonly perform such tasks as problem solving, creative design, prototype testing, field work, and working with groups and communicating with peers (Florman 1996). 'Replace' in this analysis means "become an acceptable substitute for" in economic and social senses. By becoming an acceptable (i.e. cost effective and socially admitted) substitute for the Engineer, Artificial Intelligence would effectively replace the Engineer in the work force, ideally performing all of the duties of an Engineer as a human would—or perhaps better than a human would. 'Artificial Intelligence,' often abbreviated as 'AI,' refers to the branches of computer science research that attempt to: (1) investigate, replicate, and possibly replace (i.e. surpass) human intelligence; and (2) create intelligent behavior using an artificial (i.e. man-made) medium such as a modern digital computer (Forester 1994). It is important to note that the second branch of AI does not necessarily attempt to replicate human intelligence specifically; any intelligent behavior is a target (Mainzer 1997). In the sense of the prompt, Artificial Intelligence refers not to the branches of computer science research themselves but to a tangible result of such research (e.g. a robot or some other physical device) capable of replacing the Engineer as described above. Finally, 'can' will be dealt with shortly by separating the analysis into two distinct parts. The first deals with possibility and the second with plausibility, both of which are important components of the total concept of 'can.'
The second task before analysis begins is to identify interpretations of the prompt. The first interpretation of the prompt is based on the form of the original question, which concerns capability (or, equivalently, possibility). A question like, "Is Artificial Intelligence capable of replacing the Engineer?" captures the meaning of this first interpretation. This interpretation is certainly interesting, and many people have proposed clever arguments on both sides of the question; however, an affirmative answer emerges through today's complex analysis techniques. As with many debates about form, though, the relevance of such an interpretation and its implications are limited in practice. The second interpretation of the prompt is based on the functionality (or, equivalently, plausibility) of the original question and is therefore much more relevant to our current situation. Such an interpretation deals with the 'may' part of the meaning of the word 'can.' A question like, "Is it plausible that Artificial Intelligence will replace the Engineer?" successfully captures the meaning of this second interpretation. Thorough analysis will show that the answer to this question is no. Each interpretation will be analyzed as thoroughly as possible, and then the broader implications of the results will be discussed, thus completing the analysis.
AI is, in the framework of today's emerging analysis techniques, capable of replacing the Engineer. Engineers, as defined earlier, perform a set of tasks that are generally understood to rely on the cognitive processes of problem solving, decision making, and language generation and comprehension (Matlin 1998). If AI researchers discover a way to create an AI that is sufficiently empowered to perform all of these cognitive processes as well or better than humans, it follows that such an AI is capable of replacing the Engineer in an economic sense, at least (especially as more engineers start to telecommute). Then if AI is capable of replacing the Engineer economically, two problems remain in this analysis before a total replacement can occur. The first problem is the premise that it is possible to discover an AI that is "sufficiently empowered" to actually perform the cognitive tasks of problem solving, decision making, and natural language processing. The second problem is that of the social acceptance of an "empowered" AI: If such an AI is created, it will have to be accepted socially before a true replacement of the Engineer can occur.
Humans are capable of performing cognitive tasks like problem solving, decision making, and natural language processing. Most people say without hesitation that humans are therefore intelligent, or at least display intelligent behavior. If an Artificial Intelligence were to perform the same tasks as humans as well as an average human could perform them, such an AI would be able to pass what is known as the Turing Test, and should therefore generally be regarded as intelligent (Turing 1950). In the early days of AI research, many computer scientists were overly optimistic that the Turing Test would be passed within years or, at most, decades. Edward Feigenbaum and Julian Feldman (1995) wrote in 1963, "[n]ot a single piece of evidence, no logical argument, no proof or theorem has ever been advanced which demonstrates an insurmountable hurdle along the continuum [leading to the capabilities demonstrated by human intelligence]," exemplifying many researchers' thoughts at the time. While their claim is not entirely untrue, it certainly fails in the framework of traditional computer science analysis, as partially illustrated with the publication of Perceptrons in 1969 and the subsequent decline in AI research during the 1970's. One of the most successful attacks on traditional computer science analysis has been John Searle's Chinese Room Argument, which essentially states that a general parallel or serial computer running a software program that manipulates symbols is not capable of cognition (Searle 1990). One of the main points Searle makes about computers is that they only manipulate formal symbols, and thus have no "causal powers" (Searle 1990).
Klaus Mainzer, an authority on complexity theory and complex or chaotic analysis techniques, presents a new viewpoint that differs fundamentally from traditional computer science techniques. Complexity theory proposes that intentionality (e.g. Searle's "causal powers") "has not fallen from heaven as a miraculous feature to guide and distinguish human mind from nature. It is a global pattern emerging in particular complex systems under certain conditions" (Mainzer 1997). Complex analysis encourages humans to look again at the basic assumptions that are present regarding brains and the nervous system. Groups of biological (as opposed to computer–simulated) neurons, acting together in humans to produce self–referential systems of great complexity and power, are only specific instances of complex systems that could be capable of producing behaviors that seem intelligent. Artificial (i.e. not produced as a result of natural evolution) complex systems are possible in principle (Mainzer 1997), and such systems would possess the ability to perform complex cognitive tasks intentionally through phase transitions of the complex system (Mainzer 1997; Hofstadter 1981).
Even though it is possible for complex AI systems to perform the cognitive tasks of the Engineer, the social acceptance of an AI remains as the last obstacle before AI is fully capable of replacing the Engineer. Just as human societies in the Western world have progressed from believing the sun revolved around the earth to believing that evolution is a natural process that applies to all living beings, a social transformation regarding thought as a mechanistic phenomenon seems inevitable at this point in the twentieth century, especially from the complex systems viewpoint. In this century, quantum mechanics has moved physical thought from strictly deterministic to probabilistic, changing our definition of 'mechanism' (Arbib 1985) and removing the strict predictability of digital computers from designs of the future, especially complex analog machines that are only starting to develop. In the Western world, a new world view is developing (Barbour 1993), and part of that world view is necessarily the mechanization—in the quantum, nondeterministic sense—of thought. With a different world view regarding thought and mechanization, it becomes plausible for AI to replace the Engineer on a social level; that is, if AI becomes an acceptable presence in society through a change in thought, it will become a socially acceptable substitute for the Engineer.
Although the complex systems approach and an emerging world view indicate that AI is capable of replacing the Engineer, the second interpretation of the prompt is at least as important as the first in determining if AI really can (in the full sense of the word 'can') replace the Engineer. There are two major aspects of the second interpretation that demand attention: technological and ethical. AI research is limited technologically (i.e. current resources and knowledge are not sufficient for developing an AI capable of replacing the Engineer) and should be limited ethically as well (Mainzer 1997). Analysis of this second interpretation will reveal that AI may not replace the Engineer—at least not in the near future.
The technical aspect of the second interpretation is basically a question of whether a "sufficiently empowered" AI can be developed using today's resources and knowledge. The answer is no, but there is a subtle paradox that could prohibit such a development in the future: AI researchers are generally thought of as Engineers themselves. Thus, it is paradoxical that a computer science researcher would voluntarily create a technological marvel capable of taking his place both economically and socially! Klaus Mainzer (1997) puts the dilemma into words: "we must not forget that we have to decide the direction and the ethical goals the technical development will aim at." The 'we' in Mainzer's statement should be thought of as all citizens, since every citizen, including AI researchers, will be affected by technological developments as profound as the development of an AI capable of replacing the Engineer.
The social aspect of the second interpretation is probably the more important of the two in today's world. Robert Trappl (1985) provides a good introduction to the fundamental problems at hand: "A concerned AI researcher should ask, before starting work and while working: What are the potential impacts of my work? And then decide whether she or he can justifiably continue." By creating an AI that is powerful enough to replace the Engineer, computer science researchers would be creating a situation in which humans lose control (and therefore freedom) in their own societies. As Mainzer noted earlier, researchers and laypeople are responsible for determining the ethical goals of any research. Our current ethics does not seem capable of including AI, and so such research must be carefully examined before we allow it to continue.
Interesting examples of these social and ethical limits for AI development come from knowledge bases developed to aid doctors in the diagnosis and treatment of patients. MYCIN, a knowledge base used to help prescribe antibiotics for bacterial infection, is one such expert system. An ethical question arises out of the use of such a knowledge base: Who is responsible for the diagnoses of a doctor who uses MYCIN to treat patients? In our current ethical pattern of thought, neither the machine (the computer) nor the program (MYCIN) can be held responsible. Thus, the responsibility falls on the programmer, the person who provided the data for the knowledge base, or the doctor himself, with the doctor being the most logical choice in a general situation (Barbour 1993). Complex moral decisions such as these are not handled well in today's ethics, and thus we must not allow AI research to approach its capabilities, which would confuse an already complicated moral situation.
In summary, there are serious limits present that should prevent the development of an AI capable of replacing the Engineer. The fact that it is possible, in theory, to create such an AI does not imply that it is necessary or inevitable that such a development will occur. There are many factors—including economics, society, and ethics—involved in this situation (Nilsson 1985). For several reasons, then, Artificial Intelligence may not replace the Engineer in today's world.
AI is capable of replacing the Engineer. However, AI may not replace the Engineer in today's world and world view. In light of these two conclusions, there is some consensus about what should happen in the near future. Many people (e.g. Mainzer 1997; Barbour 1993) suggest the cooperation of humans and technology, much like the cooperation in today's society, but probably becoming far more complex as time progresses. Mainzer (1997) makes explicit the need for "[i]nterdisciplinary cooperation of biotechnology, computational neuroscience, and engineering" in developing a moral and beneficial system of research. Barbour (1993) suggests, with reference to the MYCIN program and other medical expert systems, that "computers should be used as auxiliary tools to assist rather than replace the judgement of human medical experts." Such a cooperative effort would benefit doctors while providing a morally sound support for further AI research. Other potential arenas for cooperation are abundant.
While treatment of the prompt is necessarily brief in some areas, the results of the analysis generally support the twofold thesis presented earlier. Artificial Intelligence will experience many changes and improvements in the years to come. We must be aware of these developments in AI as individuals and as a society and actively work to create an ethical and societal framework that is mutually beneficial both to the development of AI and to the Engineer.
Arbib, M. A. (1985) "On Being Human in the Computer Age." In: The Impacts of Artificial Intelligence, R. Trappl (ed.), Amsterdam: Elsevier Science Publishers, B. V.
Barbour, I. G. (1993) Ethics in an Age of Technology. New York: HarperCollins Publishers.
Feigenbaum, E. A. and Feldman, J. (eds.). (1995) Computers and Thought. Menlo Park, CA: AAAI Press/MIT Press.
Florman, S. C. (1996) The Introspective Engineer. New York: St. Martin's Press.
Forester, T. and Morrison, P. (1994) Computer Ethics: Cautionary Tales and Ethical Dilemmas in Computing. Cambridge, MA: MIT Press.
Hofstadter, D. R. (1981) "Who Shoves Whom Around Inside the Careenium?" In: Metamagical Themas: Questing for the Essence of Mind and Pattern, New York: Basic Books, Inc.
Mainzer, K. (1997) Thinking in Complexity: The Complex Dynamics of Matter, Mind, and Mankind. Third edition. Berlin: Springer-Verlag.
Matlin, M. W. (1998) Cognition. Fourth edition. New York: Harcourt Brace & Company.
Nilsson, N. J. (1985) "Artificial Intelligence, Employment, and Income." In: The Impacts of Artificial Intelligence, R. Trappl (ed.), Amsterdam: Elsevier Science Publishers, B. V.
Searle, J. R. (1990) "Is the Brain's Mind a Computer Program?" Scientific American 262: 26–31.
Trappl, R. (1985) "Impacts of Artificial Intelligence: An Overview." In: The Impacts of Artificial Intelligence, R. Trappl (ed.), Amsterdam: Elsevier Science Publishers, B. V.
Turing, A. M. (1950) "Computing Machinery and Intelligence." In: Computers and Thought, E. A. Feigenbaum and J. Feldman (eds.), Menlo Park, CA: AAAI Press/MIT Press, 1995.
Created and published