ΣΕΜΦΕ forum

John McCarthy — Father of AI and Lisp — Dies at 84

When IBM’s Deep Blue supercomputer won its famous chess rematch with then world champion Garry Kasparov in May 1997, the victory was hailed far and wide as a triumph of artificial intelligence. But John McCarthy — the man who coined the term and pioneered the field of AI research — didn’t see it that way.

As far back as the mid-60s, chess was called the “Drosophila of artificial intelligence” — a reference to the fruit flies biologists used to uncover the secrets of genetics — and McCarthy believed his successors in AI research had taken the analogy too far.

“Computer chess has developed much as genetics might have if the geneticists had concentrated their efforts starting in 1910 on breeding racing Drosophila,” McCarthy wrote following Deep Blue’s win. “We would have some science, but mainly we would have very fast fruit flies.”

According Daphne Koller — a professor in the Stanford AI Lab who still carries the torch for McCarthy’s orthodox vision of artificial intelligence — it’s a quote that sums up both McCarthy and his work. “The word that bests describes him is ‘uncompromising’,” she tells Wired. “He believed in artificial intelligence in terms of building an artifact that could actually replicate human level intelligence, and because of this, we was very unhappy with a lot AI today, which provides some very useful applications but focuses on machine learning.

“He wanted AI to pass the Turing test.”

John McCarthy died on Monday at the age of 84, according to Stanford University, where he served on the faculty for almost four decades. In organizing the Dartmouth Summer Research Conference on Artificial Intelligence in 1956, McCarthy not only added a term to the popular lexicon, he founded an entirely new area of research alongside fellow pioneers Marvin Minsky, Nathaniel Rochester, and Claude Shannon. In the years to come, he would go on invent LISP — one of the world’s most influential programming languages — and he played a major role in the development of time-sharing systems.

“Without time-sharing, you wouldn’t have the modern internet,” says Lester Earnest, who worked with McCarthy at the Massachusetts Institute of Technology (MIT) in the late fifties and later at the Stanford Artificial Intelligence Lab (SAIL), the research operation McCarthy helped found in 1962.

But for all his influence over today’s world, McCarthy envisioned something much greater. Says Google’s Sebastian Thrun, who revived SAIL in 2003 before joining Google to build the company’s self-driving cars: “When it came to artificial intelligence, he was a philosopher.”

Sharing Time

Lester Earnest first encountered McCarthy at MIT while working on the government’s Semi-Automatic Ground Environment (SAGE) defense system — an early computer network that allowed multiple users to access the system at the same time — and according to Earnest, SAGE inspired McCarthy’s work on time-sharing. “He was first to come up with an idea of how to do time-sharing in a general purpose way, as opposed to special purpose,” Earnest says.

A McCarthy paper gave rise to the Compatible Time-Sharing System (CTSS) — which was demonstrated at MIT in 1961 — and a similar system McCarthy helped build at Bolt, Beranek and Newman, a private consultancy that would later play a big part in the founding of the internet.

But for McCarthy, time-sharing took a backseat to AI. At MIT, he founded an “AI group” alongside fellow Dartmouth artificial intelligence pioneer Marvin Minsky, and this eventually spawned the university’s Project Mac, a DARPA (Defense Advanced Research Projects Agency)-funded effort that would help lead the field for years.

While at MIT, McCarthy developed LISP, which became the standard programming language of the artificial intelligence community, but would also permeate the computing world at large. Today, it’s the second oldest high-level programming language still in use — after Fortran. But unlike Fortran, it’s still feeding new languages.

“[McCarthy] really encapsulated what computation meant,” says Peter Norvig, the director of research at Google, pointing to modern languages like JavaScript and Python as Lisp’s successors. “To some extent, that had been done before. People like Turing had a mathematical way of defining computing. But he was the first one to really put the essence of computing into a simple programming language, and that had a big effect on a lot of people.”

SAIL and beyond

As Project Mac evolved, McCarthy left for Stanford, where he founded SAIL, a longtime rival to the MIT effort. Lester Earnest would later join McCarthy at the lab, and he sees it as a place that helped spawn so much of the modern tech world — a notion that’s seconded by Google’s Sebastian Thrun, who points to everything from the robotics work done at SAIL to the user interface and programming work.

Earnest says the lab’s influence is exemplified by Alan Kay, who studied at the lab before moving onto the Palo Alto Research Center (PARC), where he invented object-oriented programming with his SmallTalk language, citing McCarthy and Lisp as a major influence. Kay would later call Lisp the “greatest single programming language ever designed.”

In the 80s and 90s, the lab fell out of favor. “AI went through a winter period,” says Google’s Sebastian Thrun. “There was a mismatch between the promises that were made and reality. People realized we couldn’t duplicate human intelligence.” But in 2003, Thrun revived SAIL and — in at least some ways — the McCarthy spirit.

Like Daphne Koller, Thrun says that today’s AI work didn’t always agree with McCarthy’s sensibilities. “In my mind, Google is all about artificial intelligence,” he says. “But McCarthy was troubled by this. He would come to me regularly and tell me I was doing the wrong thing.”

Thrun would respectfully disagree. “I know how to do my thing right, not your thing,” he would tell McCarthy. But Thrun acknowledges that one day, McCarthy will win the day. “What he stood for matters,” Thrun says.

“The point will come when we do understand human reasoning.”

In 1955, the computer scientist John McCarthy, who has died aged 84, coined the term "artificial intelligence", or AI. His pioneering work in AI – which he defined as "the science and engineering of making intelligent machines" – and robotics included the development of the programming language Lisp in 1958. This was the second such high-level language, after Fortran, and was based on the idea of computing using symbolic expressions rather than numbers.

McCarthy was also the first to propose a "time-sharing" model of computing. In 1961, he suggested that if his approach were adopted, "computing may some day be organised as a public utility, just as the telephone system is a public utility," and that that utility could become the basis of a significant new industry. This is the way that "cloud computing" is being sold today.

However, when obliged to choose between the time-sharing work at the Massachusetts Institute of Technology (MIT) and AI, he chose AI. He said: "The ultimate effort is to make computer programs that can solve problems and achieve goals in the world as well as humans. However, many people involved in particular research areas are much less ambitious."

In AI, McCarthy tackled such problems as whether it was "legitimate to ascribe certain beliefs, knowledge, free will, intentions, consciousness, abilities or wants to a machine or computer program". This led to a heated argument about whether thermostats could be said to have beliefs. This is not an idle debate: in the US, Nest has just announced a self-learning thermostat.

In other papers such as Free Will – Even for Robots, and Deterministic Free Will, McCarthy explored ideas of robot decision-making. He wrote a science fiction story, The Robot and the Baby, to "partly illustrate my opinions about what household robots should be like". His robot's reasoning is displayed in a Lisp-like manner as R781 decides to simulate love for Travis, the human baby. The story includes lines such as "(Required (Not (Cause Robot781) (Believes Travis (Person Robot781))))".

McCarthy was also involved with computer chess, as one of the ways of exploring computer decision-making, though he quickly became disillusioned. "Unfortunately, the competitive and commercial aspects of making computers play chess have taken precedence over using chess as a scientific domain," he said. "It is as if the geneticists after 1910 had organised fruit fly races and concentrated their efforts on breeding fruit flies that could win these races."

McCarthy was born in Boston to an Irish father and a Lithuanian mother, John Patrick and Ida Glatt McCarthy. They lost their house in the depression, and the family moved via New York and Cleveland to Los Angeles, where his father worked as an organiser for the progressive Amalgamated Clothing Workers union. In his teens, McCarthy taught himself calculus from textbooks used at the nearby California Institute of Technology. At the age of 16, as a Caltech student, he was assigned to a graduate class.

In September 1948, McCarthy found his life's work. He attended the Hixon Symposium on Cerebral Mechanisms in Behaviour, a Caltech conference that included papers on automata, the brain and intelligence, and Why the Mind Is in the Head. According to a celebration of his work published in AI Magazine, "From that time on, his chief interests related to the development of machines that could think like people."

McCarthy graduated from Caltech in 1948, then gained his PhD in mathematics at Princeton in 1951. At Princeton he became friends with another student, Marvin Minsky, who shared his passion for AI. They collaborated on numerous projects over the next decade, which included co-founding the AI Lab at MIT. The initial request was for "a room, and a keypunch, and two programmers," which came along with six graduate students whom MIT did not know what to do with. The lab was a huge success, though the interests of the founders diverged. As AI Magazine noted: "McCarthy became increasingly committed to the logicist approach to AI; Minsky came to believe that it was wrong-headed and infeasible."

In 1962, McCarthy switched coasts, moving to California and founding SAIL, the Stanford AI Laboratory, at Stanford University. He continued his work there as an emeritus professor even after his official retirement in 2000. He was still working on a new computer language called Elephant, based, he said, "on two slogans. One is that an elephant never forgets, and the other is: 'I meant what I said and I said what I meant. An elephant's faithful 100%.'" His numerous honours included a Turing award (1971), Japan's Kyoto prize (1988), and America's National Medal of Science (1990).

McCarthy married three times. His second wife, Vera Watson, an IBM computer programmer and researcher, made a solo ascent of Aconcagua in South America, and was killed in a climbing accident in 1977, when part of an all-women's assault on Annapurna in the Himalayas. He is survived by his third wife, Carolyn Talcott, and their son Timothy; by his daughters Susan and Sarah, from his first marriage, to Martha Coyote, which ended in divorce; by two grandchildren; and by his brother, Pattrick.

• John McCarthy, computer scientist, born 4 September 1927; died 24 October 2011