Hubert Dreyfus? Stuart Dreyfus? Gary Marcus? Dave Akin? Ernest Davis? Aesop?
Question for Quote Investigator: Recent advances in artificial intelligence (AI) have been remarkable, but detractors contend that current approaches are inadequate and progress will soon reach a plateau.
Critics of AI research have used the following vivid analogy: You cannot reach the moon by climbing a tall tree or a ladder. This type of criticism has been attributed to the philosopher Hubert Dreyfus and the cognitive scientist Gary Marcus, but I do not know the precise phrasing, and I do not have a citation. Would you please help?
Reply from Quote Investigator: In 1965 Hubert Dreyfus published a sharply critical report titled “Alchemy and Artificial Intelligence” for the RAND Corporation, a prominent think tank. Dreyfus asserted that human-level intelligence required properties such as “fringe consciousness” and “ambiguity tolerance” which could not be implemented with digital computers.
Hence, Dreyfus insisted that AI researchers using digital computers would fail in their attempt to build systems displaying human-level intelligence. Dreyfus used two striking analogies to illustrate the pointlessness of these efforts. Boldface added to excepts by QI:1
An alchemist would surely have considered it rather pessimistic and petty to insist that, since the creation of quicksilver, he had produced many beautifully colored solutions but not a speck of gold; he would probably have considered such a critic extremely unfair. Similarly, the person who is hypnotized by the moon and is inching up those last branches toward the top of the tree would consider it reactionary of someone to shake the tree and yell, “Come down!”
Below are additional selected citations in chronological order.
The task of reaching the moon has been used in analogies for centuries. In 1790 a book about Aesop’s fables included a discussion which condemned boasting:2
Saying and doing are different things; talking is not performing.
If words alone could do, a ladder had been found out before this time, which would have reached the moon; so that from thence a nearer prospect had been taken of the stars. Noise can only affect the ear: prattle will never butter parsnips.
In 1896 a teacher published a piece in “Kindergarten News” which described the beliefs of young children:3
Of course, the very little ones thought the boy could get the moon if he climbed the ladder, for it appeared to be just above the castle. But the older ones knew he could not, and Paul said, “Well, what about that cow that jumped over the moon then?”
In 1905 an article in the quarterly “The Pedagogical Seminary” reported on the thoughts of a young person:4
On Sept. 22, 1904, R. saw the moon and a star out of the window. Then the following remarks ensued: Said R., “Why can’t we have a moon in our house? And a little twinkle star? If you had a long, long ladder to reach ’way up in the sky, you could get a moon and a twinkle star, could n’t you?”
In 1958 the book “Interplanetary Travel” by A. Sternfeld depicted a fanciful ladder from the Earth to the Moon:5
Man is capable of climbing the loftiest mountains. But would he have strength enough to reach the Moon if there existed an Earth-Moon ladder?
Numerous experiments have demonstrated that to reach a height of 1,550 metres takes a full workday. Now if the average distance between the Earth and the Moon were divided by what a “lunar climber” can do in one day we would find that to reach the Moon would take something like 680 years, if the conditions of the first day remained unchanged.
In 1965 Hubert Dreyfus published the report “Alchemy and Artificial Intelligence” which included the analogy under examination as mentioned near the beginning of this article.
In 1966 an advertisement in the journal “Editor & Publisher” announced the availability of a newspaper feature story on the topic of AI:6
Is artificial intelligence research a form of 20th century alchemy? Can you really reach the moon by climbing trees? Voice of the Robot: a technology & science feature with pizzazz!
In 1969 congressman Allard K. Lowenstein criticized the development of anti-ballistic missile systems. He mentioned a whimsical proposal to reach the moon via beanstalks:7
Some men once thought they could reach the moon by planting beanstalks. Now eventually we did reach the moon, but we did not do it by deploying large numbers of beanstalks until some of them grew to the moon. Research provided a better way to reach the moon than climbing beanstalks, and in fact it now seems clear that not much was lost by not having spent much on beanstalks.
In 1971 R. M. Balzer published a RAND report which referred to Dreyfus’s 1965 RAND report although the precise quotation given did not actually appear in Dreyfus’s report:8
Dreyfus’ famous question of artificial intelligence, “Can we reach the moon by climbing up the branches on a tree?” might validly be directed at the major effort in programming tool enhancement, the continued development of more powerful and general-purpose languages built upon the same data elements and organizational techniques.
In 1972 Hubert Dreyfus published the book “What Computers Can’t Do: A Critique of Artificial Reason”, and he revisited the analogy:9
Feigenbaum and Feldman claim that tangible progress is indeed being made, and they define progress very carefully as “displacement toward the ultimate goal.” According to this definition, the first man to climb a tree could claim tangible progress toward reaching the moon.
Rather than climbing blindly, it is better to look where one is going. It is time to study in detail the specific problems confronting work in artificial intelligence and the underlying difficulties that they reveal.
In 1986 Hubert Dreyfus co-authored a book with his brother Stuart titled “Mind Over Machine: The Power of Human Intuition and Expertise in the Era of the Computer”. Stuart was a professor of Industrial Engineering and Operations Research at University of California, Berkeley. Interestingly, the co-authors credited Stuart with the analogy:10
As Stuart put it: “There’s no continuum. Current claims and hopes for progress in models for making computers intelligent are like the belief that someone climbing a tree is making progress toward reaching the moon.”
Aerospace engineer Dave Akin has collected a set of guidelines under the title “Akin’s Laws of Spacecraft Design”. The Internet Archive Wayback Machine contains a 2003 snapshot of the pertinent webpage which lists the following item:11
31. (Mo’s Law of Evolutionary Development) You can’t get to the moon by climbing successively taller trees.
In 2012 cognitive scientist Gary Marcus published a piece titled ‘Is “Deep Learning” a Revolution in Artificial Intelligence?’. Marcus referred to the work of digital-neural-net researcher Geoff Hinton and wrote the following:12
To paraphrase an old parable, Hinton has built a better ladder; but a better ladder doesn’t necessarily get you to the moon.
In 2019 Gary Marcus and computer scientist Ernest Davis published “Rebooting AI: Building Artificial Intelligence We Can Trust” which contained the following passage:13
By this we don’t mean that deep learning systems can’t do things that appear intelligent, but we do mean that deep learning on its own lacks the flexibility and adaptive nature of real intelligence. In the immortal words of Law 31 of Akin’s Laws of Spacecraft Design, “You can’t get to the moon by climbing successively taller trees.”
In summary, Hubert Dreyfus and Stuart Dreyfus deserve credit for applying this analogy to the domain of artificial intelligence research. The first published instance appeared in the 1965 report “Alchemy and Artificial Intelligence” by Hubert Dreyfus. However, in 1986 Hubert and Stuart co-authored “Mind Over Machine”, and the duo credited Stuart with initially formulating the analogy.
Acknowledgement: Great thanks to the anonymous person whose inquiry led QI to formulate this question and perform this exploration. Also, thanks to a August 2019 discussion thread on X-twitter whose participants included Manuel Baltieri, Melanie Mitchell, and Gary Marcus. All errors are the responsibility of QI.
Image Notes: Picture of a full moon between tree branches from Kym MacKinnon at Unsplash. The image has been cropped and resized.
- 1965 December, RAND Report P-3244, Alchemy and Artificial Intelligence by Hubert L. Dreyfus, Quote Page 86, The RAND Corporation, Santa Monica, California. (Accessed rand.org on April 19, 2024; Verified with scans) link ↩︎
- 1790, Aesop’s Fables: With His Life, Morals and Remarks, Fable 31: The Lying Mole, Start Page 51, Quote Page 52, Printed and sold by Alex. Adam, Glasgow, Scotland. (Google Books Full View) link ↩︎
- 1896 May, Kindergarten News, Volume 6, Number 5, The New Mother Play Book, Start Page 190, Quote Page 191, Milton Bradley Company, Springfield, Massachusetts.(Google Books Full View) link ↩︎
- 1905 December, The Pedagogical Seminary: A Quarterly, Volume 12, Number 4, Studies of a Child III by Alexander F. and Isabel C. Chamberlain, Start Page 427, Quote Page 432, Florence Chandler, Worcester, Massachusetts. (Google Books Full View) link ↩︎
- 1958, Interplanetary Travel by A. Sternfeld, Second Revised Edition, Translated from the Russian by G. Yankovsky, Chapter 6: Space Flight, Section 1: A Trip To the Moon, Quote Page 108, Foreign Languages Publishing House, Moscow, Russia. (Verified with scans) ↩︎
- 1966 September 17, Editor & Publisher, Volume 99, Number 38, Classified Section, Newspaper Services, Features Available, Quote Page 64, Column 4, Editor and Publisher Company, New York. (Verified with scans) ↩︎
- 1969 October 2, Congressional Record, House, 91st Congress, First Session, Authorizing Appropriations For Military Procurement Research and Development 1970 and Reserve Strength, Mr. Allard K. Lowenstein of New York, Quote Page 28131, Column 1, United States Government Printing Office, Washington D.C. (Google Books Full View) link ↩︎
- 1971 August, RAND Report R-622-ARPA, On the Future of Computer Program Specification and Organization by R. M. Balzer, Section: Introduction, Quote Page 1, The RAND Corporation, Santa Monica, California. (Internet Archive; Verified with scans) ↩︎
- 1972, What Computers Can’t Do: A Critique of Artificial Reason by Hubert L. Dreyfus, Part I: Ten Years of Research in Artificial Intelligence (1957-1967), Section 1: Phase I (1957-1962) Cognitive Simulation, Quote Page 12, Harper & Row, New York. (Verified with scans) ↩︎
- 1986, Mind Over Machine: The Power of Human Intuition and Expertise in the Era of the Computer by Hubert L. Dreyfus and Stuart E. Dreyfus with Tom Athanasiou, Prologue: “The Heart Has Its Reasons That Reason Does Not Know”, Quote Page 10, The Free Press: A Division of Macmillan, New York. (Verified with scans) ↩︎
- Internet Archive Wayback Machine, Date: November 1, 2003, Webpage title: Akin’s Laws of Spacecraft Design, Webpage creator: Dave Akin. (Accessed web.archive.org on April 20, 2024) link ↩︎
- Website: The New Yorker, Article title: Is “Deep Learning” a Revolution in Artificial Intelligence?, Article author: Gary Marcus, Date on website: November 25, 2012, Publisher: Condé Nast, New York. (Accessed newyorker.com on April 20, 2024) link ↩︎
- 2019, Rebooting AI: Building Artificial Intelligence We Can Trust by Gary Marcus and Ernest Davis, Chapter 3: Deep Learning, and Beyond, Quote Page 66, Vintage Books: A Division of Penguin Random House, New York. (Verified with scans) ↩︎