New weather "brain" to give better forecasts





h EDP order






CSIRO Problem In Computer Network




The World News (Sydney), 27 December, 1952



Machine answers 800,000 problems in an hour.


FROM OUR NEW YORK BUREAU

A complicated electronic brain tested recently at Princeton University, New York (USA) brought nearer solution the problem of why it often rains in buckets when a weatherman says it should be fine and warm.

The machine is the latest and fastest digital computer yet put into operation.

This new machine took six  years to design under the direction of Dr. John von Neumann at the Institute of Advanced Study at Princeton. It was first sponsored by the Army Ordnance Corps, and has continued under sponsorship of the army, the navy, the air force and the Atomic Energy Commission .

The team that developed the computer included Julian Bigelow and James Pomerene, who worked out the engineering with assistants. The mathematical and logical design was created by von Neumann and Herman H. Goldstine. One of the machine's recent tasks has been several 12-hour weather studies covering the US. These took about an hour each and involved 800.000 multiplications. The institute scientists stress that they are basically theorists. What really interests them are such things as testing out the mathematical theory of the 19th Century's E. E. Kummer, a job that took the machine six hours and involved 20-million multiplications, or the solution of "cubic diophantine equations"

Problems like these, though important, are quite incomprehensible, even to highly educated laymen.

The Princeton computer is small for the work it does, only 8ft x 8ft x2ft, a little more than half the size of the one recently completed at the Los Alamos Atomic laboratory. It has 2340 vacuum tubes, enough for 750 television sets, and a mile of wiring. It is not really a thinking machine but a highly glorified and fantastically speedy adding machine that works by electronics instead of by gears. It will produce 2000 multiplications; 1200 divisions or 100,000 additions per second.

Data carefully planned by mathematicians beforehand is fed into it on a perforated tape. The machine contains four principal divisions - one that does the calculating, a control organ that carries out the instructions given the machine, an input-output organ, and most startling of all, a memory organ which stores both the figures and the instructions for use when needed.

The memory division is a system of 40 cathode tubes, similar to those used in television or radar. To the untutored observer, the most interesting feature is a small "radar" tube flickering with tiny, fast-changing green dots of light which give a picture of the various 40 tubes in the memory division.

There it is possible to watch the building up of the problem as the captain of a ship watches the changing shore line in his radar set. "When we were doing the weather problem," said one of the Princeton scientists, "we could actually watch the building up of the various pressure areas on the tube, almost like watching a changing weather picture on radar. Of course, it takes skill to know what you are looking at".

When the machine did its weather problems it was fed thousands of figures on air pressure, wind velocity and other data from all over the US. Then, as scientists put it, it is a "problem in fluid flow."

If there is enough data that can be put together fast enough it is possible, for example, to figure what rain clouds will be moved where by the wind. To bring this down to local forecasting, however, would be even more complicated. The Office of Naval Research and the US Air Corps plan a large program of research into theoretical meteorology at Princeton.