Can Machines Think?

This book reports on the results of the third edition of the premier conference in the field of philosophy of artificial intelligence, PT-AI 2017, held on November 4 - 5, 2017 at the University of Leeds, UK. It covers: advanced knowledge on key AI concepts, including complexity, computation, creativity, embodiment, representation and superintelligence; cutting-edge ethical issues, such as the AI impact on human dignity and society, responsibilities and rights of machines, as well as AI threats to humanity and AI safety; (...) and cutting-edge developments in techniques to achieve AI, including machine learning, neural networks, dynamical systems. The book also discusses important applications of AI, including big data analytics, expert systems, cognitive architectures, and robotics. It offers a timely, yet very comprehensive snapshot of what is going on in the field of AI, especially at the interfaces between philosophy, cognitive science, ethics and computing. (shrink)

Much of the basic non-technical vocabulary of artificial intelligence is surprisingly ambiguous. Some key terms with unclear meanings include intelligence, embodiment, simulation, mind, consciousness, perception, value, goal, agent, knowledge, belief, optimality, friendliness, containment, machine and thinking. Much of this vocabulary is naively borrowed from the realm of conscious human experience to apply to a theoretical notion of “mind-in-general” based on computation. However, if there is indeed a threshold between mechanical tool and autonomous agent (and a tipping point for singularity), projecting (...) human conscious-level notions into the operations of computers creates confusion and makes it harder to identify the nature and location of that threshold. There is confusion, in particular, about how—and even whether—various capabilities deemed intelligent relate to human consciousness. This suggests that insufficient thought has been given to very fundamental concepts—a dangerous state of affairs, given the intrinsic power of the technology. It also suggests that research in the area of artificial general intelligence may unwittingly be (mis)guided by unconscious motivations and assumptions. While it might be inconsequential if philosophers get it wrong (or fail to agree on what is right), it could be devastating if AI developers, corporations, and governments follow suit. It therefore seems worthwhile to try to clarify some fundamental notions. (shrink)

As important as it is to realize the potential of computer technology to improve education, it is just as important to understand how the social organization of schools and classrooms influences the use of computers, and in turn is effected by that technology in unanticipated ways. In Computers and Classroom Culture, first published in 1996, Janet Schofield observes the fascinating dynamics of the computer-age classroom. Among her many discoveries, Schofield describes how the use of an artificially-intelligent tutor in a geometry (...) class unexpectedly changes aspects like the level of peer competition and the teachers' grading practices. She also discusses why many teachers fail to make significant instructional use of computers and how gender appears to have a crucial impact on students' reactions to computer use. All educators, sociologists, and psychologists concerned with educational computing and the changing shape of the classroom will find themselves compellingly engaged. (shrink)

The purpose of this paper is to argue that, in order for the debate in Computing and philosophy to move forward with respect to its current state, the advances of Computer Science and Artificial Intelligence of the last decades must be taken into account. Computational reflection, one of these advances, is presented in detail and its philosophical implications are discussed, in contrast with old-fashioned views of computational systems such as those presented by Lucas’ papers on Minds and Machines.

This is the first of two volumes of essays in commemoration of Alan Turing, whose pioneering work in the theory of artificial intelligence and computer science continues to be widely discussed today. A group of prominent academics from a wide range of disciplines focus on three questions famously raised by Turing: What, if any, are the limits on machine `thinking'? Could a machine be genuinely intelligent? Might we ourselves be biological machines, whose thought consists essentially in nothing more than the (...) interaction of neurons according to strictly determined rules? The discussion of these fascinating issues is accessible to non-specialists and stimulating for all readers. (shrink)

Abstract: Many believe that, in addition to cognitive capacities, autonomous robots need something similar to affect. As in humans, affect, including specific emotions, would filter robot experience based on a set of goals, values, and interests. This narrows behavioral options and avoids combinatorial explosion or regress problems that challenge purely cognitive assessments in a continuously changing experiential field. Adding human-like affect to robots is not straightforward, however. Affect in organisms is an aspect of evolved biological systems, from the taxes of (...) single-cell organisms to the instincts, drives, feelings, moods, and emotions that focus human behavior through the mediation of hormones, pheromones, neurotransmitters, the autonomic nervous system, and key brain structures. We argue that human intelligence is intimately linked to biological affective systems and to the unique repertoire of potential behaviors, sometimes conflicting, they facilitate. Artificial affect is affect in name only and without genes and biological bodies, autonomous robots will lack the goals, interests, and value systems associated with human intelligence. We will take advantage of their general intelligence and expertise, but robots will not enter our intellectual world or apply for legal status in the community. -/- . (shrink)

Recent philosophy of mind has increasingly focused on the role of technology in shaping, influencing, and extending our mental faculties. Technology extends the mind in two basic ways: through the creative design of artifacts and the purposive use of instruments. If the meaningful activity of technological artifacts were exhaustively described in these mind-dependent terms, then a philosophy of technology would depend entirely on our theory of mind. In this dissertation, I argue that a mind-dependent approach to technology is mistaken. Instead, (...) some machines are best understood as independent participants in their own right, contributing to and augmenting a variety of social practices as active, though often unrecognized, community members. Beginning with Turing’s call for “fair play for machines”, I trace an argument concerning the social autonomy of nonhuman agents through the artificial intelligence debates of the 20th century. I’ll argue that undue focus on the mind has obscured the force of Turing’s proposal, leaving the debates in an unfortunate stalemate. I will then examine a network theoretic alternative to the study of multi-agent complex systems that can avoid anthropocentric, mind-dependent ways of framing human-machine interactions. I argue that this approach allows for both scientific and philosophical treatment of large and complicated sociotechnical systems, and suggests novel methods for designing, managing, and maintaining such systems. Rethinking machines in mind-independent terms will illuminate the nature, scope, and evolution of our social and technological practices, and will help clarify the relationships between minds, machines, and the environments we share. (shrink)

Why do the big philosophical questions so often strike us as far-fetched and little to with everyday life? Mary Midgley shows that it need not be that way; she shows that there is a need for philosophy in the real world. Her popularity as one of our foremost philosophers is based on a no-nonsense, down-to-earth approach to fundamental human problems, philosphical or otherwise. In _Utopias, Dolphins and Computers_ she makes her case for philosophy as a difficult but necessary tool for (...) solving some of the most pressing issues facing contemporary society. How should we treat animals? Why are we so confused about the value of education? What is at stake in feminism? Why should we sustain our environment? Why do we think intelligent computers will save us? Mary Midgley argues that philosophy not only can, but should be used in thinking about these questions. _Utopias, Dolphins and Computers_ will make fascinating reading for philosophers, educationalists, feminists, environmentalists and indeed anyone interested in the questions of philosophy, ethics and life. (shrink)

[This is the short version of: Müller, Vincent C. and Bostrom, Nick (forthcoming 2016), ‘Future progress in artificial intelligence: A survey of expert opinion’, in Vincent C. Müller (ed.), Fundamental Issues of Artificial Intelligence (Synthese Library 377; Berlin: Springer).] - - - In some quarters, there is intense concern about high–level machine intelligence and superintelligent AI coming up in a few dec- ades, bringing with it significant risks for human- ity; in other quarters, these issues are ignored or considered science (...) fiction. We wanted to clarify what the distribution of opinions actually is, what probability the best experts currently assign to high–level machine intelligence coming up within a particular time–frame, which risks they see with that development and how fast they see these developing. We thus designed a brief questionnaire and distributed it to four groups of experts. Overall, the results show an agreement among experts that AI systems will probably reach overall human ability around 2040-2050 and move on to superintelligence in less than 30 years thereafter. The experts say the probability is about one in three that this development turns out to be ‘bad’ or ‘extremely bad’ for humanity. (shrink)

There is, in some quarters, concern about high–level machine intelligence and superintelligent AI coming up in a few decades, bringing with it significant risks for humanity. In other quarters, these issues are ignored or considered science fiction. We wanted to clarify what the distribution of opinions actually is, what probability the best experts currently assign to high–level machine intelligence coming up within a particular time–frame, which risks they see with that development, and how fast they see these developing. We thus (...) designed a brief questionnaire and distributed it to four groups of experts in 2012/2013. The median estimate of respondents was for a one in two chance that high-level machine intelligence will be developed around 2040-2050, rising to a nine in ten chance by 2075. Experts expect that systems will move on to superintelligence in less than 30 years thereafter. They estimate the chance is about one in three that this development turns out to be ‘bad’ or ‘extremely bad’ for humanity. (shrink)

The declared goal of this paper is to fill this gap: “... cognitive systems research needs questions or challenges that define progress. The challenges are not (yet more) predictions of the future, but a guideline to what are the aims and what would constitute progress.” – the quotation being from the project description of EUCogII, the project for the European Network for Cognitive Systems within which this formulation of the ‘challenges’ was originally developed (http://www.eucognition.org). So, we stick out our neck (...) and formulate the challenges for artificial cognitive systems. These challenges are articulated in terms of a definition of what a cognitive system is: a system that learns from experience and uses its acquired knowledge (both declarative and practical) in a flexible manner to achieve its own goals. (shrink)

Report for "The Reasoner" on the conference "Philosophy and Theory of Artificial Intelligence", 3 & 4 October 2011, Thessaloniki, Anatolia College/ACT, http://www.pt-ai.org. --- Organization: Vincent C. Müller, Professor of Philosophy at ACT & James Martin Fellow, Oxford http://www.sophia.de --- Sponsors: EUCogII, Oxford-FutureTech, AAAI, ACM-SIGART, IACAP, ECCAI.

Two large lexicological projects for the Center for the Greek Language, Thessaloniki, were to be published in print and on the WWW, which meant that two conversions were needed: a near-database file had to be converted to fully formatted file for printing and a fully formatted file had to be converted to a database for WWW access. As it turned out, both conversions could make use of existing clues that indicated the kinds of information contained in each particular piece of (...) text, thus separating fields from each other and ordering them into a tree-like structure. This indicates that both forms of the dictionaries, print and database, stem from the same cognitive need to categorize information into a kind of information before further understanding – be this for a human reader or for a machine. (shrink)

This paper concerns “human symbolic output,” or strings of characters produced by humans in our various symbolic systems; e.g., sentences in a natural language, mathematical propositions, and so on. One can form a set that consists of all of the strings of characters that have been produced by at least one human up to any given moment in human history. We argue that at any particular moment in human history, even at moments in the distant future, this set is finite. (...) But then, given fundamental results in recursion theory, the set will also be recursive, recursively enumerable, axiomatizable, and could be the output of a Turing machine. We then argue that it is impossible to produce a string of symbols that humans could possibly produce but no Turing machine could. Moreover, we show that any given string of symbols that we could produce could also be the output of a Turing machine. Our arguments have implications for Hilbert’s sixth problem and the possibility of axiomatizing particular sciences, they undermine at least two distinct arguments against the possibility of Artificial Intelligence, and they entail that expert systems that are the equals of human experts are possible, and so at least one of the goals of Artificial Intelligence can be realized, at least in principle. (shrink)

This paper is a critical essay on the question "Can machines think?", with particular attention paid to the articles appearing in an anthology Minds and Machines, A. R. Anderson editor. The general conclusion of this paper is that those arguments which have been advanced to show that machines can think are inconclusive. I begin by examining rather closely a paper by Hilary Putnam called "Minds and Machines" in which he argues that the traditional mind-body problem can arise with a complex (...) cybernetic machine. My argument against Putnam's is that either there are no problems with computers which are analogous to the ones raised by mental states, or where there are problems with machines, these problems do not have at bottom the same difficulties that human experiences raises. I then continue by showing that a cybernetic machine is an instantiation of a formal system. This leads to a discussion of the relationship between formality and predictability in which I try to show that some types of machine are in principle predictable. In the next section I attempt to prove that any discussion of outward signs of imitative behavior presupposes that some linguistic theory, such as a type reduction, has been substantiated. The force of this argument is that such a theory has not in fact been substantiated. I offer some general theory about the complexity of concept-property relations. Finally I give a demonstration that no test or set of tests can be found that will be logically sufficient for the ascription of the concept "capable of thought." If this is successful, then I have shown that no test can be found, which when a machine is built to pass it, is logically adequate for saying that that.machine can think. This argument is offered as further criticism of the Imitation Game which A. M. Turing proposed as an adequate test for thinking subjects. Besides the specific conclusion that insufficient evidence, has been offered to say that machines can think, this paper offers a more general conclusion that most standard problems have at bottom a linguistic difficulty. However, this general conclusion is a broad speculative one to which the work in this paper, is only a small exemplification and as such reflects mainly the further ambitions of the author. (shrink)

This essay replies to some representative authors, canonic indeed, of the opposition to artificial intelligence (AI) : Hubert L. DREYFUS, with his study, systematic enough in its time, What Computers can't do. His phenomenological critic meets a numerous audience, even among computer scientists. Joseph WEIZENBAUM, research worker in artificial intelligence himself, gone over to the opposition with his book Computer Power and Human Reason. His objections are classically moral and anti-techno-scientific. The philosopher John R. SEARLE, whose book Mind, Brain and (...) Science reminds us most of the resistances of linguistic scientists against artificial intelligence, turning the real difficulties of formalization of language in ontological impossibilities. Terry WINOGRAD, another research worker of the early times of AI, and Fernando FLORES, former economy minister of Chile in 1973, both rebels against the computerization of the human's world - thanks to the phenomenological post-hedeggerian way of think -, in their common work, Understanding Computers and Cognition. -/- Phenomenological philosophy has found the perfect target with AI, which embodies the image of science for the post-romantic thinking. This image can be perfectly symbolized by the frescos, choosen for the cover of this very book, situated on the walls of the Sorbonne University Library stairs! Againts the arguments of this running down romantical thinking, this book make the effort to answer on every point. But this work invites us to give up the intellectual illusions too, against which the method of cognitive sciences and information processing can be used as riddle. Incidentally, the approach of this critical method itself can be considered too as an automatic knowledge management try. (shrink)