Although there is some disagreement among computer scientists, computer engineers, and the general public as to an exact definition of Artificial Intelligence (abbreviated as AI), a very broad definition of the concept of AI will serve this analysis well. AI, most people agree, refers to the branches of research in computer science, computer engineering, and electrical engineering that attempt to produce a machine of some sort that is capable of displaying intelligent behavior. Intelligent behavior could be displayed in any number of ways, including playing chess, making an informed decision about a situation in which expert knowledge is necessary, creating optimal routes for a robot, recognition and generation of speech, and so on. One of the ultimate goals of AI is, in my opinion, to create an artificially intelligent machine that is capable of performing any task requiring intelligence that humans might perform: an android.
The foundations for AI were laid primarily by early philosophers. The Greeks and early Christians contributed some ideas to humanity, notably the concepts of the Christian "Golden Age" of the future and the Greek hierarchical orderings of Nature. The Golden Age was developed as a spiritual target of material progress on Earth, but many AI researchers might envision the Golden Age as the time when the ultimate computer or android is created, bringing about a fundamental change in the global human civilization, not just the Western world. The Greek concept of a natural order underlying all things contributes many important ideas to the rigidly defined hierarchies of computer systems today.
Not until the Scientific Revolution, however, were ideas about the nature of mind and body presented. Descartes, Lamettrie, and many others debated and presented their pictures of the interaction between our conscious selves (the mind) and our physical selves (our bodies). Descartes and others presented the now-obsolete concept of a dualistic mind–body system, in which the mind and the body are two separate entities and have no physical relation to one another. Lamettrie upheld the concept of man as a mechanistic, deterministic system, perhaps exemplifying the ideals of his world view.
After the philosophers of the Scientific Revolution laid the theoretical groundwork, the Industrial Revolution of the early 1800's ultimately produced the hardware necessary to test the theories of people like Descartes and Lamettrie. Charles Babbage and Lady Lovelace worked together to produce the first "computer" late in the 19th century, and not long after that the first digital computers began to appear. In the 1950's, the creation of the transistor reduced the size, cost, and power consumption of beasts like ENIAC from only a few years before. In the 1970's, the development of the microprocessor launched the personal computer industry. In the late 1980's and early 1990's, as processor speeds continued to grow at an exponential rate, the Internet and the World Wide Web began to connect geographically distant locations nearly instantaneously over wide area networks. Today the computer industry is beginning to affect nearly all aspects of life in the Western world.
As computer hardware became increasingly complex, work in AI fields increased in intensity. Alan Turing developed the modern framework for AI research during the 1930's. John McCarthy coined the phrase "artificial intelligence" at a conference in the 1950's. Work on artificial neural networks began in the early 1960's and proceeds with great vigor today. Work in other AI fields, including visual processing, natural language processing, game playing, and problem solving also began in the middle of the century and continues today with strong support for future research. AI research promises to bring many new discoveries to the computer industry and to the global society as well.
In the late 1970's and early 1980's, philosophers like Douglas Hofstadter, John Searle, and Daniel Dennett began to produce sophisticated arguments related to the mind–body problem. Searle claimed that the task of "strong AI" (the attempt to formalize intelligence, or create an intentional machine) was impossible. Searle's Chinese Room argument against strong AI has undergone many intellectual tests, but the only way to refute the argument definitively is to develop a machine with intentional states.
It is now 1999. The development of AI and AI research leaves us at the threshold of both a new millennium and a new world view, surrounded by many questions and few answers. We stand on a diving board left to us from the second world view, of which Descartes, Lamettrie, Babbage & Lovelace, and even Turing and McCarthy are a part. The view of our world as a strictly anthropocentric, deterministic universe is failing under the strong light of quantum physics and complex systems theory. The view of our world as a complex, interconnected system is becoming more prevalent, and it is in this third world view that many ethical issues arise from the possibilities of AI.
In my mind, the single most important ramification of AI research is the potential to demystify human thought. It is often mentioned how the view of our world by society has changed through the development of significant scientific theories. Copernicus and Galileo helped demystify the motion of the planets and stars, effectively removing humanity from its privileged place at the center of the universe. Darwin's solidification of theories of evolution helped destroy the biological superiority of humans over other forms of life. Freud helped undermine the deterministic, controlled nature of our behavior by introducing the subconscious as a necessary part of the human psyche. AI research has the potential to subvert human superiority by revealing that our intellect—our ability to generate and comprehend speech, our ability to perceive and manipulate objects in the three–dimensional universe that we experience—might not be unique. If AI succeeds in some of its goals, human superiority could receive yet another blow through the removal of intelligence as a uniquely human trait.
Even if AI falls short of showing that intelligent artificial systems are possible, there are many other ethical issues ready to surface from the use of AI in new technologies. Some of these ethical issues come up when analyzed in light of the North Carolina Progress Board's report, Measuring Our Progress: Targets for the Year 2010. Other ethical issues do not come up when analyzing the document. I will attempt to first identify ethical issues that arise from within the framework of the report's four main topic areas (economy, education, environment, and health), and then I will address the ethical issues that remain hidden from these four topic areas. After attempting to identify ethical issues involved with AI technologies, I will take a step back to consider why these issues need to be considered in the first place, both by individuals and by a state like North Carolina in 1999.
Although some of the ethical issues surrounding AI will be identified here, it is important to realize that AI is a very new technology. Ethical issues involving AI will not attempt to be resolved since we have yet to get a handle on the daunting task of identifying the issues themselves! Instead, I will try to present my viewpoint on future directions of this new technology; at certain points I will identify issues that need to be resolved before further research in AI should be allowed to continue.
AI has many ethical ramifications for today's Western societies. The four areas established by the Progress Board as indicators of progress in North Carolina—economy, education, environment, and health—each bring to light ethical issues that AI could expose, if it is allowed to develop freely.
Each main section below lists some of the ethical issues that I have considered in thinking about AI and its uses in North Carolina in the upcoming years. There are two important points to note before beginning, however. The first is that circling over all of the issues presented below are always issues of justice. If beneficial AI technology becomes available for improved economy, education, environment, or health care, such technology must be freely available to all citizens of the state who might benefit from its use. If the technology is available but distributed unfairly, there is a distinct potential that justice will be violated. As a specific example, consider an AI program designed to help young students in grade school. If the technology is only available for schools with a certain budget, then only the students who happen to live in those school districts will benefit from the technology. This creates an unfair disadvantage for those children whose families do not live in the right school districts, and those disadvantages could possibly lead to further inequalities later in life.
The second important point to keep in mind is that this discussion is far from complete—in fact, it is far from beginning, in a very fundamental sense. AI is an incredibly new technology (although the philosophy behind the computers has been debated for centuries), and as a result, we have yet to even identify areas that could present ethical or moral problems in the future. With that in mind, I will proceed with identifying what I think are important ethical issues regarding AI in North Carolina.
According to Measuring Our Progress, one of the targets for North Carolina for the year 2010 is to have "a growing, dynamic economy" (goal 1) and "high quality jobs, distributed more widely" (goal 2). These two goals certainly involve ethical issues when AI technologies become widespread in the work force.
Although the first goal does not appear to present any difficult ethical issues (Who would argue with having a growing, dynamic economy?), the second goal presents some problems that seem unsolvable at the moment. One of the major issues that AI technology has started to bring to light is that of workers' rights. AI technologies today generally come in the form of "expert systems," which are basically databases storing facts about a certain limited realm of a field (e.g., symptoms and treatments for certain diseases) combined with rules of inference that allow the expert system to gather some data about a situation and give suggestions about potential actions to take. Expert systems, as far as I know, have not significantly started taking the places of human experts in the work force, but they certainly have the potential at some point in the future.
The ethical issues that arise are questions of rights and freedom: (1) Does a human have more right to have a given job as an expert than a computer system does? (2) Who is responsible for the decisions of an AI system? To answer the first question, most people would say yes on two counts: firstly, a computer cannot have rights; and secondly, a person would be better as an all–around reasoner than a computer system. These are difficult issues to face, indeed! The first issue has been the subject of intense debate for centuries; there are some people who reject the thought of a sentient machine out of hand, while others argue adamantly that there is nothing special about humans that requires society to give them preferential treatment. I am afraid that I cannot add anything constructive to the debate, although I lean personally toward saying "no" to the question. It seems to me that, just as our global and biological superiorities were confronted by Copernicus and Darwin, our intellectual superiority will be challenged soon by AI. If a computer system were as capable of performing a job as a human, I would say that it should have equal rights to that job. The argument that humans would serve as better general reasoners seems quite valid to me at this point: no expert systems exist that show proficiency in as many areas as a human can. Again, though, these are difficult questions and are not intended to be answered in a short paper.
As to the second question, that of responsibility, equally difficult issues arise. Suppose the computer system made some mistakes in judgment about some situation, and some young boy lost an arm due to an improperly diagnosed bacterial infection. Who would be to blame? We cannot blame the expert system because it feels no sense of remorse for its actions (we think). Could we blame the programmer of the system? Unlikely, but plausible. Could we blame the human experts who provided the information for the database of the computer system? Again, I think that is unlikely but certainly plausible. Might we blame the doctor who used the program? This seems like the most logical choice, but in any case (programmer, experts, or doctor) the person you punish for the faulty decision of the computer system might be unfairly targeted. This is a major ethical dilemma!
I hope that it is easy to see from these questions and examples that bringing AI into the work force involves some major ethical problems. Since these problems are unsolved and seemingly unsolvable today, I recommend that use of AI in the work force be highly scrutinized. Human workers should, if at all, use AI only for assistance or for a second or third opinion. In all cases, though, many human experts should be consulted as well.
The two goals for North Carolina set forth in Measuring Our Progress both deal with education at or above the high school level. The first goal is "higher expectations for high school and beyond," and the second is "colleges and universities for a knowledge-oriented economy and society." These two goals, while not directly related to an ethical issue produced by AI technology, are intimately involved with some of the most fundamental issues that AI might bring to light in the coming years.
Young North Carolinians today generally spend the first 14 or 16, if not the first 18 or even 22 years of their lives in schools in North Carolina. These years are without a doubt the most important years for the formation of ideas, opinions, the ability to reason effectively, the skills needed to communicate clearly, and nearly all other cognitive functions. AI could affect the development and education of young people in many ways; in fact, AI is not really even beginning to show its application areas in education!
One of the big ways in which I can imagine AI influencing education is through animated learning agents. NC State's own IntelliMedia Initiative is currently working on creating an "intelligent user agent" who teaches users about networking. Eventually, the aim of the IntelliMedia Initiative is to extend the user agents to create a learning tool that will be generally knowledgeable about many fields of science, math, and other disciplines. Creating such an educational tool has obvious benefits for students and for teachers; after all, what educator or pupil could not use a little more help every now and then? However, there are ethical dilemmas involved in using such intelligent agents as educators. One potential problem is the uniformity of education that comes from an AI source. All students would learn from the same program, creating essentially a collective knowledge base (kind of similar to the subliminal learning tapes in Brave New World, only with less malicious intent).
Unfortunately, it can only be speculated at this point as to the ethical issues that AI might create regarding the environment. Measuring Our Progress identifies four main goals for North Carolina and its environment, but I cannot foresee any of the four as involving major ethical issues arising from the use of AI systems.
The only environmental impact that I can imagine as a result of AI influence is positive. If AI systems are designed that are capable of "smart" analysis of pollution trends, they might give valuable insight into new ways to solve environmental problems. If not, we will just continue our current strategies for finding ways to combat environmental degradation.
Measuring Our Progress identifies four goals relating to the health of its citizens: "All North Carolina children should have a healthy start in life," "All North Carolina children enter school ready to learn," "All North Carolina families are able to provide a nurturing environment," and "North Carolina citizens are healthy." AI has the potential to vastly improve both family health and personal (physical) health through the use of expert medical systems, and hopefully through sophisticated computer simulations at some point in the future.
Of course, there are some very important ethical issues that arise in the medical field that we must explore before allowing AI to come into full use. The main issue, already hinted at earlier in the paper, is that of responsibility: Who is responsible if a medical AI technology makes a mistake and causes the loss of a human life? There are several options, and each one could be viable depending on the circumstances of the AI mistake. The doctor could be held responsible for not questioning the judgment of an expert medical system, but if that is the case, what is the point of having the expert system around in the first place? The programmers could be held responsible for writing negligent decision algorithms, but in most cases the algorithms are always undergoing development and refinement anyway. Programmers would have to stop exploring the vast frontiers of computer science to concentrate on an algorithm, make sure it is complete, and test that it works appropriately in all situations. I think actually convincing programmers to stop exploring and ensure stability of their algorithms would be comparable to herding sheep without a sheep dog: some of the programmers would drift off, entranced by the world of opportunity contained within the undiscovered algorithms out there. Indeed, this "herding of sheep" may be necessary if untested AI technologies start to proliferate (especially using the Internet as a fast mode of transmission between one programmer and another). The people who supply the expert system with their knowledge could be held responsible for an AI mistake, but I think this group should be the last to be considered when such a mistake occurs. Even in the relatively small realm of medical expert systems, the possibilities for ethical problems are numerous!
These possibilities should be enough to keep unrestricted AI development at bay. In the meantime, we need to work to somehow gather an ethic together that effectively deals with the responsibility of AI creations. Who is responsible for the decisions that a machine makes? If we decide that the machine is responsible, we are basically admitting that humans are just another link in the evolutionary chain, that human thought is no more special than a machine's processing of electrons. If we decide that people are responsible, we are playing the fool to our technological developments by not allowing them the freedom to develop as they might otherwise.
The final dilemma is that, for medical expert systems as an example, AI technologies have the potential to be beneficial to our society as a whole. If a medical expert system were paired with a human doctor or team of doctors, the computer might come up with possible diagnoses of medical problems that the doctors were unable to see; however, if the computer makes a bad decision, we would hope that one of the doctors would notice. So AI technologies are potentially very helpful, but we need to develop a sufficient ethic before we can use them effectively.
There are many philosophical issues that relate to AI which are not brought out in analyzing the report of the North Carolina Progress Board. The biggest issue is commonly called the "mind–body" problem, and it basically tackles the question of whether the mind is separate from the brain somehow (called dualistic thought), or whether the mind and brain are really one single entity, or just two ways of looking at the same entity (called monistic thought). AI has the potential to shed new light on the mind–body problem. If an AI develops that sufficiently demonstrates the ability to think and act in a conscious manner, we might gain some insight as to the nature of mind and brain. This depends, of course, on the exact physical makeup of the AI that could develop, and on how well we as humans understand its inner workings.
Many people seem uninterested in the mind–body problem until they realize that a solution to the problem (potentially coming in the form of an AI) could signal the beginning of a new pattern of thought regarding human intelligence. If machines can think, then what makes humans different? Many people point to our "soul" as something that humans possess that machines would not, even if they did seem intelligent or conscious. The ethical issues coming from this problem are so grand that they hardly seem like ethical issues. What happens to our sense of social responsibility if our thoughts are not unique in the world? Are we free if our thoughts, actions, and decisions can be imitated by a machine? These are complicated issues, and I do not even know where to begin to look for the answers. Again, though, this paper is only supposed to identify some of the first signs of the path that we must follow to arrive at an ethical understanding of AI technologies.
The main reason why AI is an ethical issue, as the analysis of Measuring Our Progress revealed in brief outlines, deals with freedom and responsibility. I think the biggest question we need to answer before seriously developing AI technologies is that of the responsibility of a machine capable of making decisions. Who (or what) is responsible for decisions?
The reason this question is so important is its scope. As I explained in the previous section, if AI technology is developed that is complex enough to make independent decisions, it would threaten humanity's own conceptions of thought, freedom, and responsibility. AI technologies could throw humanity's entire image of itself on end, and we must consider those potential outcomes before seriously pursuing AI research.
However, all this is not to say that I think we should not pursue AI research at all. I think AI research could lead to liberating technologies like the user agents and expert systems discussed above. AI could also potentially liberate humans by revealing some of the mysteries surrounding thought and decision-making. I have mentioned these topics in a negative light earlier, but doing research in these areas is certainly appealing to a large number of people, especially computer science researchers! What I am calling for here is a serious look into the ethics surrounding such AI developments before those developments get out of hand. Once a sufficient framework is established that encompasses the problems discussed earlier in the paper, I am confident that AI research will lead humanity in fascinating directions.
Created and published