1939: Albert Einstein signs the Einstein–Szilárd letter

On 2 August 1939, at a summer cottage on Long Island, Albert Einstein signed a two-page letter addressed to President Franklin D. Roosevelt. The text had been drafted chiefly by the physicist Leo Szilárd, with help from Eugene Wigner. It warned that recent discoveries in nuclear physics made it conceivable to produce a chain reaction in uranium and, from that, “extremely powerful bombs of a new type.” Einstein’s immense public prestige gave the warning force; the letter was delivered to the White House that autumn.

The immediate result was the creation of what became the Manhattan Project. Beyond its military goal, the project gathered an unprecedented concentration of mathematical and engineering talent and generated an urgent demand for rapid, large-scale numerical calculation—work that existing mechanical desk calculators could not handle.

One of the mathematicians drawn into these problems was John von Neumann. Confronted with the non-linear hydrodynamics of implosion, he became convinced that only electronic, stored-program machines could supply the necessary speed and flexibility. In the spring of 1945 he set down the logical design of such a machine in the celebrated “First Draft of a Report on the EDVAC.” That document articulated the architecture—program and data held in the same addressable memory—that still underlies nearly every computer in existence.

The machines that followed (the IAS computer, UNIVAC, and their commercial descendants) furnished the hardware on which artificial-intelligence research was born. Early list-processing systems, theorem provers and chess programs written by Newell, Simon, McCarthy and others all presupposed the stored-program concept whose wartime acceleration can be traced back to the computational pressures unleashed by the 1939 letter.

Today the same architectural lineage powers the GPU clusters that train large language models and other contemporary AI systems. The path from a quiet signature on Long Island to the data centres of the present runs through nuclear research, numerical necessity and von Neumann’s clarifying insight—an indirect but unbroken thread in the intellectual history of artificial intelligence.