Permanent Staff

Jump To Main Content

Jump Over Banner


Jump Over Left Menu

Professor I J Good

Jack Good was the first Atlas/Trinity Fellow. Before the 1939-45 war, he had completed a doctorate at Cambridge under the direction of G H Hardy. During the war (1941 onwards) he worked at the Government Code and Cypher School at Bletchley Park with Hugh Alexander, Max Newman, Alan Turing and many others. He arrived on the 27th May, 1941 which was the day that the Bismark was sunk. Although the Luftwaffe and Army Enigma codes had been broken, the Naval Enigma was still to be broken. He worked in Hut 8 where the Bombes built by the British Tabulating Company (later to become ICT and take over Ferranti) were being used for code breaking. At the time, a new set of machines called the Robinsons (after (Heath Robinson) were being introduced. Jack Good worked at Bletchley throughout the war.

After the war, Jack joined Max Newman at Manchester University in the period when the Ferranti Mark I was being designed. Tom Kilburn arrived later and was the main engineer for the Baby machine.

In 1948, he moved to the Government Communications Headquarters (GCHQ) and in 1959 to the Admiralty Research Laboratory. A mischevious character, he even managed to publish an article by I. J. Good and K. Caj Doog, his name spelt backwards!

He joined the Atlas Computer Laboratory in 1964 as the first Atlas/Trinity College, Oxford Fellow. Some of his publications during this period were:

In 1967, he became University Distinguished Professor at Virginia Polytechnic Institute and State University (Virginia Tech). He was a prolific author with over 900 papers to his credit. He gained the IEEE Computer Society's Computer Pioneer Award in 1998 for his work in code breaking technology during World War II. His work led to the development of the first working special purpose, electronic computers and the first computer to be controlled by an internally stored program.

Jack Good

Jack Good at Virginia Polytechnic

A good synopsis of Jack Good has been written by John Lee.

Excerpts: Acceptance Speech for the 1998 Computer Pioneers Award from the IEEE

During the war, I was a cryptanalyst working at Bletchley Park in England at a salary of about $1000 a year. Cryptanalysis is the analysis of the opponent's cryptographic system. Two of the main cryptographic machines used by the Germans were the Enigma and a machine, the SZ42, which we called Fish. (There was a third one called Sturgeon.) the Enigma had three (later four) wheels each wired from one face to the other, whereas Fish had twelve wheels with pins on their circumference, each denoting a binary digit, zero or one. the wheel patterns varied from one communication link to anotther and from one period of time to another period. In the Naval Enigma, the three (or four) wired wheels were selected from a set of library of eight wheels, so there were 8 × 7 × 6 = 336 possible wheel orders on a given day.

I joined Bletchley park in May 1941 and was put in the Naval Enigma section called Hut 8. The reading of the Naval enigma was probably essential for winning the Battle of the Atlantic, the battle against U-boats, the German submarines. The heads of Hut 8 were the famous Alan Turing and later the British chess champion Hugh Alexander. Hugh was extremely intelligent but he never believed that a computer would ever play a good game of chess.

The keys for the Naval Enigma were changed every second day. One of the cryptanalytic methods was the method called cribbing, familiar to all cryptanalysis, of guessing stretches of language. I was part of a team of about ten people who worked on another method. It was a statistical method invented mainly by Turing and called Banburismus because the stationery was printed in the town of Banbury. The method was, I believe, the first example of sequential analysis, at least the first notable example, and preceded Abraham Wald by two years.

. . .

When I arrived at Bletchley, Banburismus was already in use. I had an extremely simple idea that cut the work by about 50%. It was the replacement of scores such as 3.6 decibans (stored as 36 centibans) by 7 half-decibans. Nearly all the entries, of which there were a few thousand, could then be expressed by a single digit. I calculated that the expected loss of weight of evidence from this change would be small. Some months later I had another statistical idea for quickly identifying which of nine so-called bigram tables was in use on a given day. The day's work on Banburismus couldn't be started until the bigram table was identified. This method of identification was needed because an earlier method due to Turing had become obsolete.

. . .

Banburismus ceased to be of value by about April 1943 after two years of use. i was then transferred to the Newmanry, the section headed by M H A Newman, FRS, where machine methods were used for the daily work on Fish. After the war, Newman built a very strong Department of Mathematics at Manchester. The Newmanry was never called the Fishery perhaps because there was another section headed by Major Tester where hand methods were used against the Fish. Among the people in the Testery were Roy Jenkins, later a Chancellor of the Exchequer, and Peter Benenson, prominent in Amnesty International.

. . .

The first cryptanalytic assistant to Max Newman was Donald Michie. I was the second one and eventually the team had at least sixteen members. When I arrived in the Newmanry the fairly small crypanalytic machine being used, for setting the wheels for the messages, was called Heath Robinson. it made use of two teleprinter tapes, one containing an enciphered message and the other containing information about the key or wheel patterns. both tapes ran simultaneously on pulleys at high speed. often the tapes would break or mistakes were made in the preparation of the tapes, or the machine would break down. Sometimes one could diagnose the machine fault by the sounds it made or by the smell when it was trying to catch fire. Because of all these troubles, the future of the Newmanry was in the balance.

I produced significance tests for the runs and also insisted on much more careful checking. I adopted the saying If it's not checked, it's wrong. Donald has recalled that we collected relevant statistics of the texts of partially broken German messages, perhpas partly from work done in the testery. Consequently, Heath Robinson had a few successes, usually with big delays, but just enough to save the section, and to justify the building of a much better and much bigger machine, the Colossus. One of the advantages of Colossus was that the data on the key tapes were represented electronically inside the machine. this avoided most of the tape preparation and therefore saved a great deal of time. This was a suggestion of tom Flowers, who became the head engineer. he should have been knighted. he knew that thermionic valves or at least the gas-filled ones called thyratrons, are reliable if they are left on all the time A valve would either die early or it would have a long life. mark I had about 1500 valves. it was the first large-scale electronic computer, but not general purpose.

. . .

Newman appointed me as the equivalent of an associate Professor of Mathematics at Manchester university, with part-time interest in electronic computing. he hired David Rees also, but David said he soon lost interest in the project. Later F C Williams, Tom Kilburn and other engineers were appointed. I cooperated well with Kilburn but only part-time. The first machine built was called the baby machine. It incorporated the Williams Tube, an adaption of a cathode-ray tube. it stored 256 binary digits at first and this was increased to 1024. Newman asked the engineers to stop trying to improve the Williams Tube and actually build a computer. It was completed on June 21, 1948. It was the first internally stored-program electronic computer and was a prototype for the large-scale Manchester Computer whose slang name was MADAM or MADM, the Manchester Automatic Digital Machine. I mad a proposal, at Kilburn's request, for the basic mathematical instructions for the Baby machine, roughly programmed in terms of the micro-operations of the machine.

Because of the Cold war, I resigned from Manchester and rejoined the Civil service about twelve weeks before the Baby machine was completed.

. . .

I have kept my rather untidy notes that I made in my period in Manchester. In these notes, one of the ideas in February 1947 was what I called Machine-building in quotes. It anticipated Microprogramming which was independently proposed a few years later, and in more detail, by M V Wilkes. But my proposal was intended for the user of a computer not for the engineer. that was the reason for the quotes. The idea was to allow the user to design his own basic instructions for some large project in terms of the primitive micro-operations of the computer. This was a natural idea for anybody who had worked with Colossus. I have selected 17 further items from those Manchester notes written in 1947 and 1948, namely:

  1. An emphasis on the value of having several accumulators not just one. Today this is standard even in hand-held computers such as the Hewlett-Packard 15C.
  2. A suggestion for distinct subroutines for a given job, depending on whether it is more important to save time or to save space.
  3. A proposal that, for the purpose of checking programs, a single binary digit could indicate whether a word represents a number or an instruction.
  4. Examples of some programs that could be run on the Baby, and how.
  5. A rough classification of the components of the machine and a proposed block diagram (June 1947).
  6. A suggestion that the length of a word should be 40 binary digits instead of 32 so that two basic instructions could be packed into one word. this was adapted for the Baby and also for MADAM. In those days computers had very small memories so the packing was important.
  7. How a puzzle called the Tower of Hanoi could be run on the Baby
  8. Calculations showing the need for all the mercury delay-line storage units to be of the same temperature within one degree Centigrade (being used on the computers at NPL and Cambridge).
  9. I made the suggestion of returning to instruction number x+1 automatically after performing a routine following instruction x.
  10. A request for flexibility in design with a facility for adding new basic instructions after the computer is complete.
  11. Proposed a form of variable-length multiplication to compromise between accuracy and running time.
  12. Proposed random rounding off for finding the effects of rounding without laborious mathematical analysis.
  13. On August 8, 1948, after I left Manchester, in a letter to Newman I distinguished between centralized and decentralized computers. In a centralized machine, the result of each calculation passes through a particular register or accumulator. My block diagram for the Baby machine was for a centralized machine, whereas Kilburn's was decentralized. I said the brain is a centralized machine except when it uses only unconscious processes.
  14. In the same letter, I drew Newman's attention to the advantages of magnetic drums over Williams tubes and mercury delay lines. Magnetic drums were adopted by Williams and Kilburn for MADAM, some time after June 21, probably independently of my letter to Newman.
  15. In letters to Turing, on September 16 and October 3, 1948, I mentioned the idea of resonance circuits in the brain; especially as a method for noticing analogies. . . . In the postscript I discussed chess-playing machines, which he and I had discussed in 1941, and gave a reasonable definition of a forced variation. I took for granted the need to distinguish between quiescent and non-quiescent positions. Shannon's paper on chess appeared in 1950.
  16. In a letter to F C Williams in July 1951 I said A facetious question is whether it is intended to display chess positions on the monitoring tubes. Of course today it is no longer at all facetious.
  17. On 1948 February 27, Williams and Kilburn asked me whether one was likely to return often in a short time (a) to a single word, (b) to a single stretch of words (of length of say 32 words), (c) to adjacent words. The engineers didn't like such things. I replied that for (b) and (c) the difficulty could be very largely removed by numbering the words so that consecutive words were in consecutive tubes instead of in consecutive rows of the same tube. Regarding (a), they didn't like say 32 occurrences in 128 words - this would seldom happen except for short iterative processes required several times. I suggested that this difficulty could be removed by having an additional elementary instruction, wait five periods, to be used whenever there is a danger of (a) occurring.

Obituary, The Times

Brilliant mathematician and Bletchley Park codebreaker who laid the foundations of modern statistics

The mathematician Jack Good played a key role among the code breaking team at Bletchley Park during the Second World War. He went on to help to build one of the first computers, was the father of a branch of modern statistics and contributed to the development of artificial intelligence.

Isidore later anglicised his name to Irving John Good but he was always known as Jack. Good was slow to learn to read, but partly as a result of being bed-bound with diphtheria at the age of 9 - when he began to discover mathematics for himself - his extraordinary intelligence became clear to his teachers.

From Haberdashers' Aske's School he won a mathematics scholarship to Jesus College, Cambridge, where he graduated with a first in 1938. He won the Smith's Prize in 1940 and did postgraduate work under A. S. Besicovitch and G. H. Hardy; he was awarded his PhD in 1941 for a thesis on topological dimension. Meanwhile, Good had also established himself as a chess player of county standard.

He was recruited in 1941 by Hugh Alexander, the reigning British chess champion, to work in the Government Code and Cipher School (GCCS known to its staff as the Golf Club and Chess Society) in Bletchley Park in Buckinghamshire.

There he was put to work in Hut 8, under Alexander and Alan Turing (1912-54), on deciphering German military and naval radio traffic. This was encoded on Enigma machines, thought by the Germans to be undecipherable. To attack the Enigma codes, special machines called Bombes were developed. With the war in the Atlantic and the threat posed by U-boats, naval Enigma became a top priority.

So too did work on German teleprinter-enciphering machine systems, a project known in Bletchley Park as Fish. Turing provided the mathematical leadership here, using statistical techniques that he developed for himself; Good became Turing's statistical assistant.

The two men remained close for the rest of Turing's short life. They played chess together - Good was much the stronger player - and discussed the possibilities of mechanisation of chess playing.

In return, Turing taught Good the Japanese board game Go, which is even harder than chess and which Good later helped to popularise.

In 1943 Good transferred to the Newmanry - the group working under the mathematician Max Newman. Here he was involved in the development of machines - precursors of the modern computer, using vast arrays of vacuum tubes and paper tape - such as the "Heath Robinson" and the Colossus.

fter the war GCCS became the Government Communications Headquarters, or GCHQ. Good followed Turing and Newman to the University of Manchester in 1947, where he was involved in work on the first computer (in the modern sense, of having an internally stored program), the Manchester Mark 1.

In 1948 Good went to GCHQ, where his work remained classified. But he continued to develop the statistical ideas he had worked on with Turing during the war. The result was his first book, Probability and the Weighing of Evidence (1950), in which the author appears on the title page as "I. J. Good, MA, PhD, former lecturer in mathematics at the University of Manchester" (GCHQ was never mentioned publicly in those days - the site it occupied was not even marked on Ordnance Survey maps). He left GCHQ for the Admiralty Research Laboratory in 1959.

Other books included The Estimation of Probablilities: An Essay of Modern Bayesian Methods (1965) and Good Thinking: The Foundations of Probability and its Applications (1983).

In 1964 Good returned to academia, to Trinity College, Oxford, where he continued his interest in computing in the Atlas Computer Laboratory, as well as his work on statistics. Though apparently well suited to Oxford life - where he enjoyed chess, and threw parties noted for the attractive women who attended - Good said he found Oxford "a little stiff", and left in 1967 to become a professor of statistics at Virginia Tech, where he spent the rest of his career, retiring in 1994.

Soon after arriving in the US he visited Hollywood to advise the film director Stanley Kubrick on scientific matters relating to his film 2001: A Space Odyssey (1968). He was elected Fellow of the American Academy of Arts and Sciences in 1995 and in 1998 he was presented with the Computer Pioneer Award of the Institute of Electrical and Electronics Engineers.

To statisticians, Good is one of the founding fathers of Bayesian statistics, an approach to the discipline based on work of Thomas Bayes in 1764. In it one forms a view of the phenomenon under study, quantifying one's uncertainty in terms of a probability distribution (the prior distribution). One then draws a sample, obtaining data, and uses the data and Bayes's theorem to update this prior uncertainty to give a new distribution, the posterior distribution. This approach - the Bayesian paradigm, as it is now called - was little used before Good's work but was given an important boost by his 1950 book and his extensive subsequent writings, and is firmly established today. Good's other interests included artificial intelligence - in particular, training computers to play chess and philosophy.

Good's importance outside academia rests on his having been a key figure in the mathematical team at Bletchley Park. This enabled the Allies to read much German naval and military radio traffic for a large part of the war. The resulting intelligence - the Ultra secret - played a key part in winning the Battle of the Atlantic. Churchill confessed later that the U-boat threat in the Atlantic was the thing that frightened him most during the war.

The Ultra secret was so crucial that great efforts, and sacrifices, were made to conceal its very existence. This even continued after the war, when Churchill ordered the destruction of the machines used at Bletchley. For several decades the Official Secrets Act was successfully used to keep everything to do with Bletchley Park out of the public domain. This secrecy eventually ended when key figures became willing to speak and write about their experiences in public. Good was one of the first to do so (he memorably remarked that it was a good job that the authorities had not known that Turing was homosexual, as had they done so he would have been banned from secret work as a security risk, and the Allies would have lost the war). Bletchley Park is now open as a museum, and its role in the history of the Second World War and the development of the computer is now acknowledged.

Professor Good was unmarried.

Professor I. J. Good, mathematician and wartime codebreaker, was bom on December 9, 1916. He died on April 5, 2009, aged 92