Computer Science

The 1932 appointment of mathematician Warren Weaver as director for the Natural Sciences marked the expansion of Rockefeller Foundation (RF) endeavors to the newly emerging field of computer science. The RF had long invested in physics and chemistry through the work of the International Education Board (IEB) in the 1920s and its own support for National Research Council fellowships. In the 1930s, as Weaver developed targeted programs of research, advances in mechanical calculating devices showed significant promise for machine solutions to complex problems that Weaver and his staff quickly recognized.

“A Roomful of Brains”[1]

In the 1930s, the Foundation underwrote Vannevar Bush’s massive analog computer at the Massachusetts Institute of Technology (MIT), which was finally dedicated in 1942 as the Rockefeller Differential Analyzer. One of the largest and last of the mechanical analog computers, it weighed 100 tons, used 2,000 vacuum tubes and 150 motors. Dubbed “analog” because it carried out an analogy of a real physical process, the machine was touted as revolutionary in mechanized calculus. It operated in real time and, unlike later digital computers, did not require a program to translate high language to binary machine language.

A typical record as produced by the output recorder of the M.I.T. Differential Analyzer

A typical record as produced by the output recorder of the M.I.T. Differential Analyzer

During World War II, the differential analyzer was used for computations of missile trajectories. But the war also accelerated the obsolescence of the differential analyzer. War work--in which Bush, Weaver, and scores of researchers trained on the differential analyzer participated at various levels--pushed digital technologies forward quickly. The RF responded to these developments in two ways. It continued to seek uses for the analog computer, the most ambitious of which was a project that computed market variables in the economies of developing countries. It also funded an experimental mathematics group at MIT to develop digital computing. The group was soon underwritten by the deeper pockets of the United States Navy’s Whirlwind Program.

Merging Biology and Mathematics

In the 1940s, RF funding helped mathematician Norbert Wiener develop the new field of cybernetics. Weiner published his landmark book, Cybernetics, in 1948, after the Foundation’s grant to his joint project with cardiologist Arturo Rosenbleuth to study feedback loops in the human body. His work with Rosenbleuth led Wiener to broader speculations about similarities between the communications and control systems in machines and those in natural and biological systems.

Natural Sciences director Weaver was also involved with a breakthrough book of the same era. In 1949, he co-authored the landmark text entitled The Mathematical Theory of Communication with Claude Shannon. The book remains essential to the understanding of information theory. Weaver’s introductory chapters are widely credited with making Shannon’s higher mathematical theories accessible to a more general audience.

Machine Learning

In the summer of 1956, the RF funded a milestone: the Dartmouth Conference on Artificial Intelligence. The five-week gathering engaged a group of leading researchers who were a veritable “who’s who” of computing innovators, including:  John McCarthy, founder of the Stanford Artificial Intelligence Lab; Marvin Minsky, who developed the idea of neuron nets and founded the MIT Computer Science and Artificial Intelligence Lab; Claude E. Shannon, pioneer of Information Theory; and Nathan Rochester, developer of the IBM 701, the first general purpose, mass-produced computer.

MIT computation center, 1957

MIT computation center, 1957

The field was so new that mathematician John McCarthy had to invent a new term to help explain the concept of machine learning. In fact, the first known use of the term, “artificial intelligence,” was in the proposal that McCarthy (with co-authors Shannon, Minsky, and Rochester) submitted to the RF. At the time, the implications of the field were understood only by a handful of researchers. Even Weaver, an insider to applied mathematics, tempered his support for the proposal with restraint, although he authorized RF Director of Biological and Medical Research Robert S. Morison to act as he saw fit.

McCarthy requested $13,500 for a two-month conference, but Morison offered $7,500 for five weeks. As Morison explained,

I hope you won't feel we are being overcautious but the general feeling here is that this new field of mathematical models for thought, though very challenging for the long run, is still difficult to grasp very clearly. This suggests a modest gamble for exploring a new approach, but there is a great deal of hesitancy about risking any very substantial amount at this stage.[2]

The significance of this new field, while murky to outsiders, was immediately evident to practitioners. IBM, Bell Laboratories, and RAND all lent support for their key researchers to attend.

Today, the conference remains widely acknowledged as the founding moment of artificial intelligence and has influenced research and development in mathematics, engineering, linguistics, computer science and psychology ever since. Its wide range of topics included complexity theory, language simulation, neuron nets, abstraction of content from sensory inputs, and the relationship of randomness to creative thinking.

Beyond subject matter, the Dartmouth conference signaled a major paradigm shift in how mathematical work would happen in the computer age. Essential to the conference was bringing researchers with disparate specialties together, freeing them from the strictures of teaching and publishing within their own subfields, and offering time and opportunity for collaboration and exchange of ideas and information.

The Dartmouth conference proved to be the high water mark of RF funding in the field of computer science. With military and corporate research increasingly driving the computer industry in the coming years, RF funding became less and less necessary. But the seeds the Foundation planted proved to be crucial to many later advances in the field.