The Mind's I: fantasies and reflections on the self and the soul Douglas R. Hofstadter and Daniel C. Dennett 1981 (ISBN 0553345842)
This book is an invitation and a challenge to the reader to join in exploration of the human mind and soul. During the 1900's science began its invasion of this territory. Previously, philosophers and theologians had control over discussion of minds and souls and feared no empirical challenges. Now, neurobiologists work to reveal the mechanisms of mind, and artificial intelligence researchers strive to produce increasingly sophisticated computational models of mental processes.
The book contains 27 chapters, each with a previously published work by people like Alan Turing, Richard Dawkins, Raymond Smullyan, John Searle, Stanislaw Lem, Thomas Nagel, as well as Hofstadter and Dennett. Most chapters also contain a commentary by Hofstadter and/or Dennett. Dennett and Hofstadter are both supporters of the idea that we can learn much about human minds and souls as we continue to explore human mentality in terms of what is broadly called information processing. Dennett and Hofstadter both like the idea that the wonders of human mentality can be accounted for by mechanical brain processes. If human mental processes reflect physical brain processes that can be interpreted in terms of information processing, then there is nothing theoretical to prevent us from building human-like mental processes into our mechanical devices. A few views that run counter to this notion are included in this book mainly as targets for refutation by Dennett and Hofstadter.
Chapter 4 reprints Alan Turing's famous "Turing test" article from 1950 (in the journal Mind. The Turing test for machine intelligence is an operational test that basically asks if a machine can use human language well enough so as to be able to play complex language games. In the 1960s Joseph Weizenbaum's famous ELIZA program was able to surprise some people with the quality of its mindless generation of human language text. In the 1970s other programs such as Terry Winograd's SHRDLU showed the very real difficulty of providing machines with the semantic knowledge upon which human language behavior depends. No machine has ever come close to passing the Turing test, but Searle and others have indicated that they would not be impressed if a computer program did pass this test. It can be argued that merely behaving as if you have human-like intelligence in no way shows that you have human-like intelligence. Arguments against the utility of the Turing test usually revolve around the idea that a human-like computer system could be a "zombie" with no real understanding of the language games it plays. Other people who still see value in the Turing test believe that you cannot fake complex behaviors like human language use, that the only way to build a computer system that could use human language would be to build a system that actually understands human language. But nobody knows how. Neuroscience is still working to study how human children use the brain to learn about the world and how to talk about what they know. Certainly no artificial intelligence researcher has ever figured out how to get a computerized machine to replicate these tasks that are mere "child's play". All Dennett and Hofstadter could do in 1981 was lament about how little was known about how to make computers learn in the ways humans do. The two main efforts of artificial intelligence research towards machine learning since The Mind's I was published (trainable neural nets and expert systems) failed to bring us any closer to machine intelligence, so the challenge offered to the reader in this book still seems fresh and lively.
One if the central ideas explored in this book is stated clearly by Searle (Chapter 22, "Minds, Brains and Programs", originally an article published in The Behavioral and Brain Sciences, 1980): "...mental processes are computational processes over formally defined elements." Searle has objections to the idea that computer programs might ever produce mind, but the idea that mentality involves computation can be traced through the history of Western philosophy where it has long been explored in the context of trying to explain human reason in terms of formal logical systems. A dramatic and famous rejection of the formal systems idea was that of Ludwig Wittgenstein, a philosopher who Dennett respects. After first embracing the idea of reducing everything to logical atoms (Tractatus Logico-Philosophicus), Wittgenstein later rejected the idea that human language games should be formulated as formal systems (Philosophical Investigations). However, many philosophers and artificial intelligence researchers remain captivated by the formal systems approach. For example, Dennett has tried to help the MIT COG project develop formal computer programming methods towards the goal of producing human-like intelligence. In his book "Contemporary Philosophy of Mind", Georges Rey provides an example of continuing attempts to express human intelligence in machines through computational processes over formally defined elements. An alternative but minority approach has grown out of the work of people like Gerald Edelman and his student Olaf Sporns through which it is suggested that machine intelligence can most efficiently be achieved by creating autonomous robotic systems that can learn the way human children learn through interacting with their environment.