Looking over Marvin Minsky’s classic The Society of Mind after over a decade and a half from my last reading of this seminal book on artificial intelligence, and reflecting on the intervening crazy evangelism of Kurzweil, Kaku, Brynjolfsson and MacAfee, and not a few others who seem certain of the appearance in this century of Strong Artificial Intelligence–riffs on Vernor Vinge’s “Singularity”–I appreciated yet again Minsky’s honesty: we don’t yet even have much of a workable theory of the mind. Those ubiquitous color-coded brain scans lighting up active areas for various kinds of thinking…so what? A theory they aren’t. (What they are is a staple of the fatuous “brain fitness” programs on TV.)
One evening in the mid 1980s, I had dinner in Cambridge with some friends, including Minsky. An enthusiastic waiter praised each entree’ selection, conveying that a choice among such delights would be agonizing. Finally, Minsky asked, “What’s the worst thing on the menu?”
Minsky’s The Society of Mind sums up years of practical thinking about the tactics of thinking: how do humans choose among options, find meaning, and infer answers? And though unavoidably flawed in design, his book as a whole creates with wit and grace a profound metaphor of the human mind that has considerably influenced many others who think about thought, both human and artificial.
It has been over a quarter of a century since The Society of Mind came out. (Minsky has since written The Emotion Machine: Commonsense Thinking, Artificial Intelligence, and the Future of the Human Mind , a companion piece to The Society of Mind and which extends his thinking about our cognition and, among other things, theorizes about the nature and role of emotions in human thought, a book that should be considered in its own dedicated review.) Certainly in a historic endeavor such as artificial intelligence, a quarter of a century is an instant. The Society of Mind is a theory with an indefinite life.
The Society of Mind raises large, urgent, and profound matters about our identity, our possible journey as individuals and as a species, and the mystery of our defiance of our limitations imposed by astonishing, impersonal, and constraining natural selection. I think finally Minsky hopes human psychology will be saved from Accident.
Indeed, Minsky is passionately confident of a certain vision, one that presages the creation of more powerful minds:
“There is not the slightest reason to doubt that brains are anything other than machines with enormous parts that work in perfect accord with physical laws.”
Essentially, Minsky’s idea of mind is this: When we think, we do so through the interaction of many “agents,” each of which performs specific functions in response to certain stimuli which originate in the world outside us or from regions within our own minds. Furthermore, these mental agents interact as a society in which combinations of operations create the ability to apprehend significance, find analogies, appreciate beauty, represent physical objects, and perform the rest of the forms of our thought. Hardly a traditionally poetic concept of the mind, Minsky’s idea of cognition leads him to dozens of clever ideas about how the processes of thought can be modeled. He is today a conceptual leader in artificial intelligence where people need tractable ideas about cognition which promote productive programming.
He argues that the major reason we now can only speculate about the mind is our inexperience in dealing with the enormous levels of complexity of our mental operations. Minsky believes that robust artificial intelligence programs of the future, running on far more powerful computers than today’s, may give us such experience, and hence greater insight into our minds, because of the great abilities of this later technology for powerful simulations.
But such hopes are not of immediate importance to Minsky. First he wants to represent viable processes of thought in such a way as to guide, inspire and enable computer scientists, computer technologists, hardware designers, computer architects, and programmers to proceed with confidence to build an artificial intelligence that emulates and, in some dimensions, far surpasses mysterious thinking as we experience it. Human thought is wondrous but painfully dark and limited. Minsky wants to get on with building models of thought processes that help make artificial intelligence work in factories, space exploration, household cleaning, prognostication, grocery stores, government, education and wherever else it can improve the quality of life.
The enormous, historic transnatural thrust comes later.
Hence The Society of Mind, though it is a very clever, wonderfully aphoristic, and often dazzling book of philosophy, theory, speculation, and even literature, is less concerned with philosophically probing the riddles of the mind than with promoting confident, purposeful engineering. Technologists essentially must have a workable idea of what they are doing to do it well. And so along with its large dimensions, Minsky, because he believes that the confused, eclectic and bootstrapped concepts about thought which most artificial intelligence researchers have been forced to rely on are not “doable,” has written an extraordinarily sophisticated, subtle and elegant How-To book. This hardly diminishes his book; indeed, precisely because of its technological inspiration, its elegant thrust toward engineering, The Society of Mind is quintessentially a serious, definitive book in the history of such inquiry.
Consider, for example, Minsky’s idea of “chaining.” Imagine a bridge. The essential arch form of a bridge is also that of a tunnel. You can chain several bridge forms together side-by-side and create the structure of an arcade. If you chain bridge forms together end-to-end in a long series, you have the structure of a road. If you stack bridge forms, you build shelves. The form of a bridge is also that of a table. And so on. For a computer programmer trying to simulate vision or build smart graphics, “chaining” is an interesting and useful metaphor of how certain processes of thinking will work. Many mental operations such as “chaining” can be understood, says Minsky, as the functional working of his various mental “agents.” Computer people will feel at ease with the numerous diagrams suggesting how these mindless agents with specialized functions can be wired together in different ways to allow a thinking entity to perform a great variety of functions from recognizing objects to building structures to thinking up new ideas.
A more subtle example of the engineering thrust of The Society of Mind lies in Minsky’s treatment of metaphor as an analogy. He is everywhere concerned with efficient, clever and practical ways to do difficult mental things such as remember large bodies of knowledge; represent cause and effect; form conclusions; and weigh various courses of action. For instance, he alludes to this clever metaphorical device of memory: a person can more easily remember a complicated list or a lengthy oratory by imagining that certain portions are always to be found in certain rooms along certain hallways or on certain floors of an imaginary building through which he strolls when needing to recall the knowledge. Such clever tactics to inspire the engineering of intelligent machines bring confidence in part because they are based on modern insights into cognition. There is increasingly compelling research in psychology which shows that regardless of the mysteries of human thought, ingenious metaphors and analogy are pervasive in many successful human mental operations. For example, think briefly about such indispensable but simple everyday examples of metaphor as: Argument as War, Time as Money, More as Up, Better as Up, Bad as Down: “As I was winning the argument, Bob lost ground”; “Sally is spending time wastefully”; “Mary rose in stature”; “Harry fell into disgrace.” Metaphors are hardly unique to poetry.
A complex and impressive example of how metaphor and analogy are liberating artificial intelligence technology can be seen in Douglas Lenat’s well-known Project CYC, a crucial and pioneering multi-year project to build a large encyclopedic base of common sense and specific knowledge. Considered by some to be an artificial intelligence project which must succeed for the field to advance significantly, CYC is being designed in part to allow computer programs to conduct smart and enormous searches to discover new designs, scenarios, concepts, heuristics and other applications: in sum, to accomplish superhuman learning. Fundamental to such searching are processes of analogizing. I’ve merely mentioned the term, “analogy,” but in the hands of skilled programmers the concept leads technically to a practical way to accomplish a considerable and critical forward step in artificial intelligence; forms of analogizing are, in short, doable for the programmer. It is important to understand that without the technologically philosophical approach of ingenious engineer-thinkers like Minsky (and colleagues such as Lenat), such crucial applied thinking about thinking would not be available for building a successful artificial intelligence.
The Society of Mind also has bearing on ideas of computer hardware. Consider the conceptual compatibility of Minsky’s thought with the building of more and more powerful computing. Minsky’s mental agents simultaneously process various forms of diverse knowledge; and, of course, knowledge should be processed in parallel fashion in networks of distributed, cooperative processors.
It is therefore unfortunate that some readers, particularly nonspecialists, may find themselves discouraged by the organization of The Society of Mind. Minsky has primarily presented his thoughts in the form of single ideas each of which is described in a self-contained fashion and numbered after the manner of technical manuals: 3.1, 5.2, 14.3, etc. Fortunately, this arrangement fails to hide the wit and wisdom of the book, but for a short while makes for slow going until the coherence and narrative interest of Minsky’s vision soon envelop the reader. Minsky confesses that he could not seem to find a more immediately dramatic way to convey his idea of the mind. Perhaps neither he nor anyone else now living could find a way short of that future when machine minds are at work which are far more complex and powerful than today’s machine intelligence.
The Society of Mind is a courageous and essential visionary book by a most daring and original thinker. For that reason alone, it is an indispensable book for all who care deeply about how we think and how we might extend our thinking.