Marvin Minsky

Steps toward Artificial Intelligence

The problems of heuristic programming-of making computers solve really difficult problems-are divided into five main areas: Search, Pattern-Recognition, Learning, Planning, and Induction. A computer can do, in a sense, only what it is told to do. But even when we do not know how to solve a certain problem, we may program a machine (computer) to Search through some large space of solution attempts. Unfortunately, this usually leads to an enormously inefficient process. With Pattern-Recognition techniques, efficiency can often be improved, by restricting the application of the machines methods to appropriate problems. Pattern-Recognition, together with Learning, can be used to exploit generalizations based on accumulated experience, further reducing search. By analyzing the situation, using Planning methods, we may obtain a fundamental improvement by replacing the given search with a much smaller, more appropriate exploration. To manage broad classes of problems, machines will need to construct models of their environments, using some scheme for Induction. Wherever appropriate, the discussion is supported by extensive citation of the literature and by descriptions of a few of the most successful heuristic (problem-solving) programs constructed to date.

A Proposal for the Dartmouth Summer Research Project on Artificial Intelligence, August 31, 1955

The 1956 Dartmouth summer research project on artificial intelligence was initiated by this August 31, 1955 proposal, authored by John McCarthy, Marvin Minsky, Nathaniel Rochester, and Claude Shannon. The original typescript consisted of 17 pages plus a title page. Copies of the typescript are housed in the archives at Dartmouth College and Stanford University. The first 5 papers state the proposal, and the remaining pages give qualifications and interests of the four who proposed the study. In the interest of brevity, this article reproduces only the proposal itself, along with the short autobiographical statements of the proposers.

K‐Lines: A theory of Memory

Most theories of memory suggest that when we learn or memorize something, some drepresentation of that something is constructed, stored and later retrieved. This raises questions like: How is information represented? How is it stored? How is it retrieved? Then, how is it used? This paper tries to deal with all these at once. When you get an idea and want to remember it, you create a K-line for it. When later activated, the K-line induces a partial mental state resembling the one that created it. A partial mental state is a subset of those mental agencies operating at one moment. This view leads to many ideas about the development, structure and physiology of memory, and about how to implement framelike representations in a distributed processor.

Music, mind, and meaning

Speculating about cognitive aspects of listening to music, this essay discusses: how metric regularity and thematic repetition might involve representation frames and memory structures, how the result of listening might resemble space-models, how phrasing and expression might evoke innate responses and finally, why we like music — or rather, what is the nature of liking itself.

Form and Content in Computer Science (1970 ACM turing lecture)

The trouble with computer science today is an obsessive concern with form instead of content. No, that is the wrong way to begin. By any previous standard the vitality of computer science is enormous; what other intellectual area ever advanced so far in twenty years? Besides, the theory of computation perhaps encloses, in some way, the science of form, so that the concern is not so badly misplaced. Still, I will argue that an excessive preoccupation with formalism is impeding our development. Before entering the discussion proper, I want to record the satisfaction my colleagues, students, and I derive from this Turing award. The cluster of questions, once philosophical but now scientific, surrounding the understanding of intelligence was of paramount concern to Alan Turing, and he along with a few other thinkers--notably Warren S. McCulloch and his young associate, Walter P i t t s made many of the early analyses tha t led both to the computer itself and to the new technology of artificial intelligence. In recognizing this area, this award should focus attention on other work of my own scientific family--especially Ray Solomonoff, Oliver Selfridge, John McCarthy, Allen Newell, Herbert Simon, and Seymour Papert, my closest associates in a decade of work. Papert s views pervade this essay. This essay has three parts, suggesting form-content confusion in theory of computation, in programming languages, and in education.

Universality of Tag Systems with P = 2

By a simple direct construction it is shown that computations done by Turing machines can be duplicated by a very simple symbol manipulation process. The process is described by a simple form of Post canonical system with some very strong restrictions. This system is monogenic: each formula (string of symbols) of the system can be affected by one and only one production (rule of inference) to yield a unique result. Accordingly, if we begin with a single axiom (initial string) the system generates a simply ordered sequence of formulas, and this operation of a monogenic system brings to mind the idea of a machine. The Post canonical system is further restricted to the “Tag” variety, described briefly below. It was shown in [1] that Tag systems are equivalent to Turing machines. The proof in [1] is very complicated and uses lemmas concerned with a variety of two-tape nonwriting Turing machines. The proof here avoids these otherwise interesting machines and strengthens the main result; obtaining the theorem with a best possible deletion number P = 2. Also, the representation of the Turing machine in the present system has a lower degree of exponentiation, which may be of significance in applications. These systems seem to be of value in establishing unsolvability of combinatorial problems.

Jokes and the Logic of the Cognitive Unconscious

Freud’s theory of jokes explains how they overcome the mental “censors” that make it hard for us to think “forbidden” thoughts. But his theory did not work so well for humorous nonsense as for other comical subjects. In this essay I argue that the different forms of humor can be seen as much more similar, once we recognize the importance of knowledge about knowledge and, particularly, aspects of thinking concerned with recognizing and suppressing bugs — ineffective or destructive thought processes. When seen in this light, much humor that at first seems pointless, or mysterious, becomes more understandable.

A FRAMEWORK FOR REPRESENTING KNOWLEDGE

Briefly describes frame systems as a formalism for representing knowledge and then concentrates on the issue of what the content of knowledge should be in specific domains. Argues that vision should be viewed symbolically with an emphasis on forming expectations and then using details to fill in slots in those expectations. Discusses the enormous problem of the volume of background common sense knowledge required to understand even very simple natural language texts and suggests that networks of frames are a reasonable approach to represent such knowledge. Discusses the concept of expectation further including ways to adapt to and understand expectation failures. Argues that numerical approaches to knowledge representation are inherently limited.

A Selected Descriptor-Indexed Bibliography to the Literature on Artificial Intelligence

This listing is intended as an introduction to the literature on Artificial Intelligence?i.e., to the literature dealing with the problem of making machines behave intelligently. We have divided this area into categories and cross-indexed the references accordingly. Large bibliographies, without some classification facility, are next to useless. This particular field is still young, but there are already many instances in which workers have wasted much time in rediscovering (for better for for worse) schemes already reported. In the last year or two this problem has become worse, and in such a situation just about any information is better than none.

Unrecognizable Sets of Numbers

When is a set <italic>A</italic> of positive integers, represented as binary numbers, “regular” in the sense that it is a set of sequences that can be recognized by a finite-state machine? Let π <subscrpt><italic>A</italic></subscrpt>(<italic>n</italic>) be the number of members of <italic>A</italic> less than the integer <italic>n</italic>. It is shown that the asymptotic behavior of π <subscrpt>A</subscrpt>(<italic>n</italic>) is subject to severe restraints if <italic>A</italic> is regular. These constraints are violated by many important natural numerical sets whose distribution functions can be calculated, at least asymptotically. These include the set <italic>P</italic> of prime numbers for which π <subscrpt><italic>P</italic></subscrpt>(<italic>n</italic>) @@@@ <italic>n</italic>/log <italic>n</italic> for large <italic>n</italic>, the set of integers <italic>A</italic>(<italic>k</italic>) of the form <italic>n<supscrpt>k</supscrpt></italic> for which π <subscrpt><italic>A</italic>(<italic>k</italic>)</subscrpt><italic>n</italic>) @@@@ <italic>n<supscrpt>P/k</supscrpt></italic>, and many others. The technique cannot, however, yield a decision procedure for regularity since for every infinite regular set <italic>A</italic> there is a nonregular set <italic>A′</italic> for which | π <subscrpt><italic>A</italic></subscrpt>(<italic>n</italic>) — π <subscrpt><italic>A′</italic></subscrpt>(<italic>n</italic>) | ≤ 1, so that the asymptotic behaviors of the two distribution functions are essentially identical.