Key People in Artificial Intelligence: John McCarthy

McCarthy’s Upbringing and Career

John McCarthy was born in Boston, Massachusetts to a multinational family in 1927. His family apparently moved around several times due to the Great Depression until his father nailed down steady factory work in Los Angeles. His aspirations to work in an area related to mathematics seemed to begin during his high school years when he apparently taught himself college-level mathematics through borrowing college textbooks from California Institute of Technology. This was also what led him to apply to and begin studying Mathematics at Caltech in 1944. In 1948, he graduated with his B.S. in Mathematics.

During his time working towards his B.S., McCarthy attended a lecture by none other than John Von Neumann, which is claimed to have inspired his initial and future work into what is now Artificial Intelligence. In 1951, he worked his way to a Ph.D. in mathematics from Princeton University and began his teaching career at the university level. After having been a professor at several universities, including short-stint at Stanford, McCarthy was hired as a researcher at MIT, in 1956.

In 1962, McCarthy began working at Stanford for the second time, where he stayed for good until he retired in 2000.

LISP and McCarthy

During the time that he was moving around between university gigs, McCarthy developed the Lisp programming language. Beyond being based on Lambda Calculus, which is apparently a form of Calculus that truly caters to the widespread usage of functions as well as to the development of Turing machines. If you haven’t heard of Turing Machines before, they’re simply the first example of an artificial neural network. It’s also claimed by institutions such as the University of Cambridge that Turing Machines can simulate the performance of any algorithm. Even so, these algorithms seem to have to be simply structured. Essentially, Turing Machines have historically worked with only 1s and 0s and the machine can read as well as edit any digit that appears.

This is somewhat of a deliberate generalization with the aim of sticking to the subject of McCarthy and his Lisp programming language. The connection between Turing Machines and the aim of McCarthy’s Lisp language seems to be that Turing Machines eventually evolved into the AI systems or Artificial Neural Networks that we know today.

Therefore, it seems that even before the Artificial Intelligence industry truly began to develop, McCarthy began to deliberate create one of the programming languages that would become essential to its growth. In fact, Lisp is still widely used in the AI industry, today. In future posts, we’ll go into benefits and drawbacks of Lisp for AI in order to try to better understand one of McCarthy’s biggest contributions to the space.

McCarthy as one of the Fathers of AI

Perhaps even more important than his creation of Lisp, is the Dartmouth Workshop of 1956.

According to John McCarthy’s personal material from Stanford on the subject, the four organizers of the workshop were himself, Marvin Minsky, Nathaniel Rochester and Claude Shannon. Essentially, the organization of said workshop was catalyzed by the work of all four in the AI field before it was the AI field from 1948-1956. All four worked on projects that could be considered foundational to Artificial Intelligence today, with each other, prior to organizing the workshop.

After seven years of alternating among different projects such as a project involving the inversion of Turing functions which McCarthy said was not an appropriate approach to AI. In the summer of 1955, everything came to a head when McCarthy and Rochester worked together at Rochester’s lab at IBM. At the same time that they were there, Minskey and Shannon joined them and they all came up with the original idea for the Dartmouth Workshop.

The overall conference seems to have been inspired by their dissatisfaction with their work related to developing any sort of Artificial Intelligence, from 1948-1955. The group’s first idea was that they and anyone whom they invited would spend two months at Dartmouth working on projects related to AI, together. From this, the goal was to make substantial advances in the nascent space.

How did it all turn out?- The Birth of the AI Industry

In the end, the results of the workshop seem to have turned into something more akin to a small conference due to its size and as McCarthy makes clear, it became about the work of different figures, beyond himself and his three closest colleagues. Key takeaways include Herbert Gelertner’s presentation of his Fortran List Processing programming language, which he developed with Nathaniel Rochester to power a primitive sort of AI system that solved certain Geometry theorems in a novel way.

McCarthy states that this inspired him to create the Lisp programming language for AI systems because he began to see limitations in the FLP programming language, related to processes like recursion and others. To clarify, recursion is actually the computer science practice of breaking a problem down to the simplest way of explaining it and solving it. With AI, this is particularly helpful in teaching the algorithms that run AI systems to work more efficiently. According to Utah University, the benefits of using recursion in algorithms include the aforementioned practice of using the simplest solution to each problem that that AI needs to solve as well as that the system apparently remembers every iteration of the problem as it works towards solving it.

To be clear, recursive algorithms reportedly simplify problems over and over again until they are correctly solved, while at the same time remembering all of the data that the system has already processed in working towards doing so. In the interests of avoiding a lengthy digression, if you’re interested in knowing more about recursion, start with the link below and stay tuned for our future posts on the subject.

In going back to the conference, however, the development of Lisp by McCarthy was not the only substantial achievement. Mainly, according to McCarthy as well as other sources, the chief success of the 1956 Dartmouth workshop was that after it, AI began to be considered as a legitimate branch of science for the first time.

McCarthy continued developing projects for the AI space until his death in 2011. Right out of the gate from 1956, John McCarthy began developing systems that seem to have inspired the continued efforts related to the creation of human-like AI. This included systems that effectively produced computer versions of all of the human senses as well as the human brain’s ability to reason. During the creation of these systems, he reportedly also stated that any form of learning could be mimicked by machines, which appears to show that he foretold the creation of the field of Deep Learning, in the future. In addition to this, he apparently pioneered the idea of connecting computers to a network to save time and money, which can be said to have led to developments like Blockchain Networks.

McCarthy didn’t stop with AI, and according to a report by The Independent, he even basically also foretold the coming of E-commerce in a paper that he published in the 1970s.

All in all, it seems quite clear why John McCarthy is often called the father of AI and if you’re interested in learning more, look no further than the links below and get ready for our future posts.