Young Sydney Scientist Invents a Robot Brain







CSIRO Problem In Computer Network




The Sydney Morning Herald, 1 December, 1951 November, 1951



By Barbara Richards

The president of an instrument making company in the United States challenged an "electronic brain" the other day to a 20-game chess tournament with 1,000 dollars at stake. But the "brain" could not accept because it was too busy with the mathematical problems of defence to take time off for "recreation".

A young Sydney physicist, Mr. T. Pearcey, has done better than that. In a room in the Radiophysics Division of the C.S.I.R.O., he has actually played a game of chess (and won it) with an electronic "brain" of his own design.

Furthermore, he has begun a game of draughts with the brain while at was in the middle of a high speed problem on supersonic flight. The changeover from flight to draughts took only a minute or two, and at the conclusion of the game, the brain went on with the problem in supersonics.

Among the everyday, done-without-human-help feats of the robot Brain may be listed:

Adding up 333,000 digits every second.

Working out the chances of two coins falling heads, tails  or one of each, 200 times a second

Remembering accurately thousands of numbers between one and a million, all at the one time.

This new man-made brain, better than anything of its kind in Australia, and equal to only two or three others in the world, works entirely by electrical impulses.

It is not a pretty piece of gleaming apparatus. Rather, it is huge and almost cumbersome. You will see a solid slab of equipment about 20 feet long, 8 feet high and 18 inches thick. But, like its human counterpart, its mechanism is delicate, and balanced to perfection.

WHEN you meet the brain, the first impression you get is of its "nerve" — at least twenty square feet of extremely intricate wiring, tied up with knob-like terminals spread out on one side of the slab.

Once, I accidentally let my hand stray near the wires.

"Better not touch them", a voice said. I drew back suddenly.

"Give you a shock. I suppose" I asked.

The reply was quite serious. "It probably would, but you'd disrupt its thought"!

Farther along, 1,500 electronic radio valves wink at you from neat little rows, while to one side, set apart, is the brain's "memory" — a box containing 64 five-foot long steel tubes filled with mercury, each connected by wire to the valves. These tubes can store 1,000 numbers, which the machine can use at a rate of 1.000 times a second.

Over to one side, a control board with knobs and three small windows where green dots flash shows you what's going on inside (when you know how to work out just what is going on inside).

From one corner of the room comes a click-clucking noise. It is a white coated mathematician giving the brain its orders by feeding cards through a machine. Close beside it, the brain is printing its answers.

THIS is one at the most impressive features of the brain: That, capable as it is of solving extraordinarily difficult problems in an astonishingly quick time, it can be told what to do by a punch card machine, known in almost every office, and print its answers on a teleprinter.

Although the machine "thinks" in silence, it uses a lot of power — enough to light up more than 200 ordinary household lamps; and although it has its own air-conditioning system, the room is always uncomfortably hot.

The brain is quite new.  Three years ago, it was just a lot at worried lines on Mr. Pearcey's forehead.

When he had finished the design, the construction was completed by Mr. M. Beard and several other mathematicians.

SO FAR, Mr. Pearcey has been working out how to use the machine. To tell it what to do, he has been compiling programmes of instructions. These are recorded on to punched cards so that the machine can “read”' them.

They will be something like these: solving problems of supersonic flight, the designing  of wings for jets, and the performance and flight of rockets. It may also save engineer's time, like those on the Snowy Mountains scheme, by helping the design of dams. Or in chemistry, by helping to find out how complicated chemical compounds, such as the proteins we eat are constructed. It will also help the C.S.I.R.O in its rain-making research by calculating what goes on in clouds.

To make things simple for the brain, all numbers and instructions for its use are converted to a chain of 1's and 0's by a simple mathematical device.

A thousand of these numbers, each 20 digits long, can be kept at the one time in the memory.

The numbers, already converted to 1's and 0's are kept constantly flowing through the mercury in 64 columns, nose to tail, in the form of electrical impulses, where they can be taken out and use at the will of the operator.

The machine must be given each instruction separately. Some are quite short, like telling it to "alert" a particular section of its anatomy to send or receive a new number; or longer, like those which require the calculation of a square root or for printing out a number onto paper.

THE operator feeds the punched cards through the punch card reader in their proper order, converting the numbers and instructions into electrical impulses. Red lights flash in the valves, green dots appear and disappear on the little screens, the machine is working hard — in silence.

Impulses flash at an incredible rate from one part of the brain to another, sending numbers to and fro, and flashing action signals to prepare the various parts for action as they are required in the course of calculation.

In its almost human capacity, the brain needs to be told only about one quarter of what it has to do. The rest of its actions correspond to those made while the machine is "thinking" what to do next, one-thousandth of a second later!


Robots And Chess

Sir — With regard to the article on the "electronic brain" ("Herald", December 1), allow me to point out that it has never played a game of chess.

Dr T. Pearcey, the designer, has stated he sees no reason why it should not he adapted to play chess, but thinks it would be easily beaten.

There has been come research into the possibility of "chess robots" in America, but the only serious claim put forward (by Dr. C. E. Shannon) is that a computer could be adapted so as to play a "tolerably good" game. But computers have been constructed which will take the winning side in a few simple end-games — such at King and Rook versus King — and invariably win.

C J. S. PURDY. Greenwich