John McCarthy? Pamela McCorduck? Bertram Raphael? Donald Michie? Melanie Mitchell? Bertrand Meyer? Anonymous?
Question for Quote Investigator: Pioneering artificial intelligence (AI) researchers tackled a variety of challenging problems. One early goal was the development of symbolic mathematics systems capable of performing polynomial factorization, integration, and differentiation. Researchers made such great progress that this field was reclassified. It was no longer part of AI; instead, it became a subfield of algorithm design and analysis.
In 1997 the Deep Blue chess computer triumphed over world champion Garry Kasparov. The system employed a massive brute-force game-tree search. The victory was a milestone, but some researchers believed that the system was no longer part of AI research.
In general, if a problem is effectively solvable then it is no longer deemed an appropriate task for AI. Here are two versions of a comment about this phenomenon that is both wistful and mordant:
(1) As soon as it works, no one calls it AI anymore.
(2) If it works, it isn’t AI.
This notion has been attributed to several AI researchers including John McCarthy and Edward Feigenbaum. I am having difficulty finding solid citations. Would you please explore this topic?
Reply from Quote Investigator: This notion is difficult to trace because it can be expressed in many different ways. Below is an overview showing the evolution via key statements together with dates:
1971: AI is a collective name for problems which we do not yet know how to solve properly by computer [Attributed to Bertram Raphael by Donald Michie]
1979: Every time somebody figured out how to make a computer do something—play good checkers, solve simple but relatively informal problems—there was a chorus of critics to say, but that’s not thinking [Pamela McCorduck]
1979: AI is whatever hasn’t been done yet [Attributed to Larry Tesler by Douglas Hofstadter]
1982 May: If it’s useful, it isn’t AI [Anonymous]
1982 Sep: If you can understand how it works, it isn’t AI [Anonymous]
1983 May: If you do know what you’re doing (or if you find out), it isn’t AI anymore [Beau Sheil]
1984: If it works, it isn’t AI [Anonymous]
1984 Feb: Anything computers can’t yet do is AI [Anonymous]
1984 Sep: It it’s useful, it isn’t AI [Anonymous]
1985: If you understand how it works, it isn’t AI [Anonymous]
1985 Apr: When an AI idea is turned into a useful system, in some sense it isn’t AI anymore [Roger Schank and Larry Hunter]
1985 Jun: Once they are thoroughly solved, they are not AI anymore but just another computer program [Severo Ornstein]
1988: If it works, it isn’t AI [Attributed to Edward Feigenbaum]
2011 Oct: As soon as it works, no one calls it AI anymore [Attributed to John McCarthy by Bertrand Meyer]
2017: Intelligence is whatever machines haven’t done yet [Attributed to Larry Tesler by Garry Kasparov]
Below are details for selected citations in chronological order.
In 1970 a gathering of computer scientists was held in Menaggio, Italy and afterwards the proceedings was published under the title “Artificial Intelligence and Heuristic Programming”. Donald Michie of the University of Edinburgh wrote an article that referred to a definition of AI proposed by Bertram Raphael of SRI International. Boldface added to excerpt by QI:1
B. Raphael, at this meeting, has suggested that AI is a collective name for problems which we do not yet know how to solve properly by computer.
As soon as convincing success is attained in any particular domain, that domain is transferred by tacit consent into an appropriate established category of computer science — information retrieval, adaptive control theory, computational linguistics, optical character recognition, and so on, as the case may be.
Raphael’s definition fits into the family of expressions under examination in this article because it implies that active AI problems only have ineffective or impractical solutions. Tractable problems are no longer problems for AI.
In 1979 Pamela McCorduck published “Machines Who Think: A Personal Inquiry into the History and Prospects of Artificial Intelligence”. She discussed the change in status of problems that are solvable with computer programs:2
. . . it’s part of the history of the field of artificial intelligence that every time somebody figured out how to make a computer do something—play good checkers, solve simple but relatively informal problems—there was a chorus of critics to say, but that’s not thinking.
Also, in 1979 Douglas Hofstadter published “Gödel, Escher, Bach: An Eternal Golden Braid”. The author credited computer scientist Larry Tesler with a pertinent remark:3
There is a related “Theorem” about progress in AI: once some mental function is programmed, people soon cease to consider it as an essential ingredient of “real thinking”. The ineluctable core of intelligence is always in that next thing which hasn’t yet been programmed. This “Theorem” was first proposed to me by Larry Tesler, so I call it Tesler’s Theorem: “AI is whatever hasn’t been done yet.”
In May 1982 journalist Tom Alexander published an article about AI in the business magazine “Fortune”. Alexander stated that AI researchers lost interest in a topic after practical mastery had been achieved. He presented a germane anonymous adage:4
Researchers in the field of artificial intelligence have a two-edged saying—part rueful admission, part complaint: “If it’s useful, it isn’t AI.” Artificial intelligence (often called Al) is the discipline devoted to equipping computers with intellectual faculties such as the abilities to perceive, reason, and learn. It has traditionally attracted brainy computer scientists who were more inclined than most of their colleagues to pursue research with little thought to its practical application . . .
In September 1982 a piece in “Computing Reviews” discussed a book by Peter Laurie. The book discussed machine learning algorithms in the domains of image interpretation and chemical analysis. The reasoning steps of the algorithms were sometimes opaque. This prompted an anonymous adage:5
Artificial Intelligence, what is it? Peter Laurie quotes the old non-definition: “If you can understand how it works, it isn’t AI.”
In May 1983 a symposium on “Artificial Intelligence Applications for Business” was held at New York University. An article in the proceedings by Beau Sheil referred to a definition of intelligence he attributed to Marvin Minsky:6
I rather like Minsky’s definition of “intelligence” as a property we ascribe to intellectual behavior that we admire but do not understand . . .
The implication for AI programming of Minsky’s unsettling definition is that, if you are building an AI program, you are, by definition, building a program that you don’t understand. The going in position is that you don’t know what you’re doing. And there’s no reprieve. If you do know what you’re doing (or if you find out), it isn’t AI anymore, it’s something else, because it can’t possibly involve “intelligence”. Some topics have actually been reclassified out of AI, for just this reason.
In 1984 the book “The Computer Chronicles” by H. D. Lechner included an anonymous instance in this family of sayings:7
There has been a general feeling that AI was a science fiction, impractical kind of work that couldn’t ever be commercialized (a snide phrase widespread in industry at one time was: “If it works, it isn’t AI”).
In February 1984 “PC Magazine” printed the following in an article by journalist Jared Taylor:8
The question of what is—or is not—AI has gotten so fuzzy that some wags have decided that anything computers can’t yet do is AI. By this definition, speech recognition rates as AI because it is still hard to make machines do it properly.
In September 1984 Professor of Education Diane McGrath published an anonymous instance in the journal “Electronic Learning”:9
Some AI researchers have an expression: “It it’s useful, it isn’t AI.” This half-serious, half-facetious comment harks to what has been the traditional, research-oriented attitude towards AI.
In 1985 Peter Laurie published the book “Databases: How To Manage Information on Your Micro” which contained the following two anonymous sayings:10
There is a saying in the trade: ‘If you understand how it works, it isn’t AI.’ There is a shorter, cynical version too: ‘If it works, it isn’t AI.’
In April 1985 AI researchers Roger Schank and Larry Hunter published the following passage within the computer hobbyist magazine “Byte”:11
Many of our best AI ideas require a great deal of work before they can become useful applications. And when an AI idea is turned into a useful system, in some sense it isn’t AI anymore.
In June 1985 “The Atlantic” magazine quoted computer scientist Severo Ornstein who presented an instance:12
Ornstein says, “As a concept, AI is like jelly—when you push on it in one place, it just goes someplace else. It is really just a term we apply to problems that seem intractable. Once they are thoroughly solved, they are not AI anymore but just another computer program.”
In 1988 the book “Putting Artificial Intelligence To Work: Evaluating & Implementing Business Applications” by Seymour Schoen and Wendell G. Sykes attributed an instance to Edward Feigenbaum. The authors suggested that the remark was spoken during the 1970s:13
Professor Edward Feigenbaum, while explaining the meaning of AI to a distinguished and perplexed scientific review panel for a Department of Defense AI application development program in the late 1970s commented, “If it works, it isn’t AI.”
In October 2011 computer scientist Bertrand Meyer reminisced about John McCarthy who had died a week earlier. Meyer stated that he had heard McCarthy employ a version of the saying:14
Many breakthroughs in computer science, both in theory (advances in lambda calculus and the theory of computation) and in the practice of programming (garbage collection, functional programming languages), can directly be traced to work in AI. Part of the problem is a phenomenon that I heard John McCarthy himself describe: “As soon as it works, no one calls it AI any more.”
In 2012 computer scientist Moshe Vardi attributed an instance to John McCarthy:15
At the same time, AI’s accomplishments tended to be underappreciated. “As soon as it works, no one calls it AI anymore,” complained McCarthy. Yet it is recent worries about AI that indicate, I believe, how far AI as come.
In 2017 Garry Kasparov published “Deep Thinking Where Machine Intelligence Ends and Human Creativity Begins” which included the following passage which contained a variant of the saying attributed to Tesler:16
Chess is the perfect example of Larry Tesler’s “AI effect,” which says that “intelligence is whatever machines haven’t done yet.” As soon as we figure out a way to get a computer to do something intelligent, like play world championship chess, we decide it’s not truly intelligent.
In 2020 Melanie Mitchell published “Artificial Intelligence: A Guide for Thinking Humans”, and she attributed an instance to John McCarthy. The supporting footnote pointed to the 2012 article by Moshe Vardi:17
Nowadays, computer-chess programs are used by human players as a kind of training aid, in the way a baseball player might practice using a pitching machine. Is this a result of our evolving notion of intelligence, which advances in AI help to clarify? Or is it another example of John McCarthy’s maxim: “As soon as it works, no one calls it AI anymore”?
In summary, Bertram Raphael expressed the central idea of this family of sayings by 1971. Larry Tesler received credit for a concise instance in this family in 1979. A humorous instance with an anonymous creator appeared in “Fortune” magazine in May 1982. Different versions proliferated in subsequent years. John McCarthy received credit for an instance after his death in 2011. This article presents a snapshot of current research. Future discoveries may clarify this history.
Image Notes: Metaphorical illustration of a brain-like circuit from Steve Johnson at Unsplash. The image has been cropped and resized.
Acknowledgement: This exploration was inspired by three very interesting books which each credited John McCarthy with this saying: “Superintelligence: Paths, Dangers, Strategies” (2014) by Nick Bostrom; “Artificial Intelligence: A Guide for Thinking Humans” (2019) by Melanie Mitchell; and “The Coming Wave: Technology, Power, and the Twenty-first Century’s Greatest Dilemma” (2023) by Mustafa Suleyman. The three books each cited Moshe Vardi’s 2012 article for support. This led QI to attempt to find earlier evidence.
- 1971, Artificial Intelligence and Heuristic Programming, Edited by N. V. Findler and Bernard Meltzer, (Articles based on lectures given at the First Advanced Study Institute on Artificial Intelligence and Heuristic Programming held in August 1970), Formation and Execution of Plans by Machine by D. Michie, Start Page 1010, Quote Page 101, University Press Edinburgh, Edinburgh, Scotland. (Verified with scans) ↩︎
- 1979, Machines Who Think: A Personal Inquiry into the History and Prospects of Artificial Intelligence by Pamela McCorduck, Chapter 8: Us and Them, Quote Page 175, W. H. Freeman and Company, San Francisco, California. (Verified with scans) ↩︎
- 1979, Gödel, Escher, Bach: An Eternal Golden Braid by Douglas R. Hofstadter, Chapter 18: Artificial Intelligence: Retrospects, Quote Page 601, Basic Books, New York. (Verified with scans) ↩︎
- 1982 May 31, Fortune, Practical Uses for a “Useless” Science by Tom Alexander, Start Page 138, Quote Page 139, Time Inc., New York. (Verified with scans) ↩︎
- 1982 September, Computing Reviews, Volume 23, Number 9, Section: General Literature, Review by B. Flanagan in Poissy, France of Peter Laurie’s “The Micro Revolution: Living with Computers”, Quote Page 435, Column 1, Association for Computing Machinery, New York. (Verified with scans) ↩︎
- 1984, Artificial Intelligence Applications for Business: Proceedings of the NYU Symposium, May 1983, Edited by Walter Reitman of BBN Laboratories, Chapter 16: The Artificial Intelligence Tool Box by Beau Sheil (Xerox Special Information Systems. Palo Alto, California), Start Page 287, Quote Page 288, Ablex Publishing Corporation, Norwood, New Jersey. (Verified with scans) ↩︎
- 1984 Copyright, The Computer Chronicles by H. D. Lechner (SRI International), Chapter 24: Artificial Intelligence, Quote Page 354, Wadsworth Publishing Company / Continuing Education, Belmont, California. (Verified with scans) ↩︎
- 1984 February 21, PC Magazine, Volume 3, Number 3, What Is Artificial Intelligence? by Jared Taylor, Start Page 172, Quote Page 172, Column 1, Ziff-Davis Publishing Company, New York. (Verified with scans) ↩︎
- 1984 September, Electronic Learning, Volume 4, Issue 1, Artificial Intelligence: A Tutorial for Educators by Diane McGrath (Adjunct Professor of Education at the University of Illinois), The New Directions of AI Research, Quote Page 40, Column 2, Scholastic Inc., New York. (Verified with scans) ↩︎
- 1985, Databases: How To Manage Information on Your Micro by Peter Laurie, Chapter 7: The Intelligent Database, Quote Page 126, Chapman and Hall / Methuen, London. (Verified with scans) ↩︎
- 1985 April, Byte, Volume 10, Number 4, The Quest To Understand Thinking by Roger Schank and Larry Hunter, Start Page 143, Quote Page 155, Column 1, McGraw-Hill, New York. (Verified with scans) ↩︎
- 1985 June, The Atlantic, Volume 255, Number 6, Reports and Comment: The “Star Wars” Defense Won’t Compute by Jonathan Jacky, Start Page 18, Quote Page 22, The Atlantic Monthly Company, Boston, Massachusetts. (Verified with scans) ↩︎
- 1988 (1987 Copyright), Putting Artificial Intelligence To Work: Evaluating & Implementing Business Applications by Seymour Schoen and Wendell G. Sykes, Chapter 4: Artificial Intelligence Concepts, Quote Page 48, John Wiley & Sons, New York. (Verified with scans) ↩︎
- Website: Blog of Communications of the ACM, Article title: John McCarthy (Obituary), Article author: Bertrand Meyer, Date on website: October 28, 2011, Website description: Blog of research organization for computer science. (Accessed cacm.acm.org on Jun 20, 2024) link ↩︎
- 2012 January, Communications of the ACM, Volume 55, Number 1, Editor’s Letter: Artificial Intelligence: Past and Future by Moshe Y. Vardi, Quote Page 5, Association for Computing Machinery, New York. (ACM Archive at acm.org) ↩︎
- 2017, Deep Thinking Where Machine Intelligence Ends and Human Creativity Begins by Garry Kasparov with Mig Greengard, Chapter: Conclusion – Onward and Upward, Quote Page 251 and 252, John Murray, London. (Verified with scans) ↩︎
- 2020 (2019 Copyright), Artificial Intelligence: A Guide for Thinking Humans by Melanie Mitchell, Chapter 9: Game On, Quote Page 157, Picador, New York. (Verified with scans) ↩︎