History paper

Thompson, a senior executive of J. Lyons and Company Limited, persuaded the company to give £3,000 to The University of Cambridge in exchange for advice about building its own computer and, consequently, in 1951, the LEO (Lyons Electronic Office) computer was running its first business application (Land, 1998; Aris, 2000). Thompson had visited the USA and recognised that computers could prove useful in business data processing (Land, 2000). Thus he was well placed to consider the cost effectiveness of using computers in business, stating that if there were sufficient work to keep the computer running then 'there should be no doubt' about cost savings.

However, it was essential that the tasks for the computer were organised carefully so that computer run-time was minimised. Errors needed to be planned as they would inevitably occur and programmers could be trained so that the coded program could be quickly prepared. Problems occurred with form design and checks were needed to ensure that only the correct program could be used for a particular task (Thompson, 1958). Maintenance of the hardware also posed a problem with valve deterioration taking a prominent role but Thompson (1958) recommended that fault diagnosis should closely follow that of medical diagnosis, where symptom knowledge is used to eliminate faults until the root cause is deduced. Thompson listed the most frequent difficulties, the most problematic being 'finding out precisely what the job [for the computer] is required to do' (Thompson, 1958).

In large organisations no one person knows all the ramifications of a procedure involving many departments and seldom is there a person in a position to make comprehensive decisions. In addition, requirements may be prejudiced by past (irrelevant) experiences as well as conservatism of staff at all levels. Rationalising the many facets of an involved procedure was a concept the understanding for which many members of an organisation lacked mental capacity (Thompson, 1958). It is tempting to suggest that in the last 40 years this aspect of business computing has scarcely improved in spite of the work of Jackson, Beer and Wood-Harper amongst others. However, Thompson (1958) concluded that staff at Lyons were working well and all seemed happy in the work associated with the computer installation.

Future diagnosis would suggest that this approach was 'inward looking' (Aris, 2000) but the LEO project achieved a number of 'firsts' including flow-charts for specification and the concepts of systems integration (Aris, 2000).

These two examples alone show that project management and requirements specifications were issues even with earliest computers. In addition, form design, both as input and printed output, could be problematical as could the reliability of the hardware. Even so, computers were being used for both statistical and non-statistical forecasting (Douglas, 1958b), for example, Unilever used an Elliott 405 (Hickman, 1959) for stock recording and control (Douglas, 1958c) and management decisions based on operational research techniques such as linear programming (Douglas, 1958d).

Indeed, production control was a very popular application area for computers with EMI Electronics Ltd (Hickman, 1959) and International Computers and Tabulators Ltd (Bryen, 1959) being amongst the earliest. The Royal Dutch/Shell Group (Galer, 1959) had been using a Ferranti Mark I for some four years to model the most economically viable scenario using linear programming techniques. Thus, from about 1956, a number of UK companies had invested in a variety of different computers which were used either for payroll/personnel type activities or were used for gaining competitive advantage through tighter stock control and decision management (mainly operational research type modelling).

These various application areas of data processing by computer have continued to the present day but perhaps not surprisingly, accounts of systems such as we have seen above diminished in the Journal so consequently by 1963 (Volume 6) the number of papers describing computerised data processing case studies had decreased considerably although areas such as Government National Insurance contributions (Drummond, 1963) and the modelling of life assurance scenarios (Giles, 1964) were still featured.

Already by the mid-1960s, data transmission had become topical as businesses tried to optimise use of the computer by sharing processing time. In March 1963, a one-day Symposium was held on Data Transmission and the need for standards so that integration was possible. The papers reveal that data transmission using the GPO telephone and telegraph facilities was already being used by such companies as Vauxhall Motors in the UK (Russell, 1963), The Atomic Energy Establishment (Chapman, 1963) and The Royal Air Force (Davey, 1963).

At this Symposium, Smith (1963) spoke on behalf of the BCS Data Transmission Committee stating that the 'combination of data transmission and data processing systems gives rise to an entirely new type of system, frequently referred to as Real Time Systems or Operating Systems'. It is important to distinguish the latter term from that used by programmers as Smith was referring to an 'operating system' as one which is closely allied to management's function of running the organisation in that it supported data access by managers, rather than the internal program that manages the computer.

Point to point systems and high speed links were the normal components of data transfer between different buildings, but in some cases, data were transferred by means of an off-line system that used the punch-cards as input. An on-line system utilised a real-time channel which allowed data to be channelled directly from the data (buffer) store into the main working memory of the computer (Smith, 1963). In 1963, eight organisations in the UK used the public telephone network for electronic data transmission through 14 links. Indeed, Dale (1963) advocated much more use of the GPO telephone system but made the point that costs were prohibitive for small amounts of data.

Similarly, Hitchcock (1963) argued that at that time, a light van or even a bicycle was an adequate, and a much cheaper, form of data transmission! However, by 1966, data transmission and real-time programming had arrived in the type of applications associated with it today; for example, work at The Defence and Research Agency (then The Royal Radar Establishment) at Malvern in Worcestershire used these facilities to program an interchange of data for radar surveillance (Phillips, 1967) and National Physical Laboratory (through DW Davies) was calling for a national message-switching network (Campbell-Kelly, 1988). Of historical importance, these developments in the UK seem to mirror those in the USA (O'Neill, 1995).

By 1968, electronic data processing in most commercial applications such as payroll, inventory control and forecasting using statistical and operational research type models was commonplace. Electronic data transmission was being used though both public and private systems (Dale, 1963) and real-time programming associated with data transmission was emerging from its embryo state. It is worth repeating that over the 10 years considered here, the number of papers published in the Journal devoted to business applications of any sort diminished considerably. The rising profile of numerical analysis as an academic subject is seen clearly in the increase of papers devoted to mathematical matters related to computing, either through programming or algorithmic modelling.

Scientific and technical research

The Journal has always given space to original results and developments in the mathematical foundations of computer science. In addition, it has supported the publication of papers using the computer to model mathematically scenarios found in engineering. Thus, the first volume devoted pages to the modelling of wind tunnel design (Hollingdale and Barritt, 1958a and 1958b) and harmonic analysis (Beck, 1958). Simulation was utilised by industry from these early days of computer usage; helicopter flightpaths and takeoff and landing constraints were modelled (Harrison, 1959) as were melting shop operations (Neate and Dacey, 1959)

which utilised flow diagrams and Monte Carlo methods (Hurd, 1985) in its implementation, the latter, in particular, requiring heavy computational demands (MacKenzie, 1991). This necessitated random number generation, an area also studied by the numerical analysts (Clark and Holz, 1961). Production control was also simulated for Courtaulds (Musk, 1959), Shell (Galer, 1959) and to simulate a more economical control in a light engineering factory (Bryen, 1959). An Elliott computer was used to model a hydro-electric scheme at Lochaber in Scotland (King and Peel, 1960) while a similar computer was used at Berkeley Power Station to simulate the complete power station for training (Dempsey, 1960). This seems to be the first paper using simulation for training as against modelling a system's activities but it is typical of this decade that the paper gives no evaluation of the system's worth as a training tool, only detailing the equations (in combination) and circuitry for their solution.

Aircraft testing was also carried out using an HEC 2M computer (Taylor, 1961) which proved useful for the drudgery of smoothing calculations.

In March 1961, the BCS held a conference in London on data transmission, document handling and character recognition. Space does not allow for a detailed discussion of the papers presented but papers covered the problems of using public switched circuits of telegraph and telephone types (Williams, 1961), requirements (Long, 1961) and error detection and correction (Wright, 1961) of such systems. By the early 1960s, computers were being interconnected in order to automate the hot strip mill of a steelworks (Massey, 1962) as well as to enable satellite communications (Pearson, 1962). Other papers of this Conference included automatic character (pattern) recognition (Grimsdale and Bullington, 1961;

Clowes and Parkes, 1961) and character quality through scanning (Merry and Norrie, 1961). Some of the illustrations of binary mappings suggest that progress has slowed significantly in more recent years.

Medical applications of computers were amongst the first (Wilkes, 1958) but papers in this area seem to have been few as far as submission to the Journal was concerned. In five years just one appeared, on simulating the liver (Watt and Young, 1962).

Chrystallography had become dependent upon the computer for analysis (Matthewman, 1967) and dynamic programming was being used to model the construction of an overhead electric power transmission line so that costs were minimised. This again used dynamic programming techniques and the results compared favourably with expectation (Ranyard and Wren, 1967).

Volume 7 of the Journal contained just one simulation study (Ridgion et al., 1964), that of a Cowper, or hot blast, Stove used in blast furnaces, while Volume 8 also modelled just one technological scenario, that of underground storage of compressed air for gas turbines (Langham, 1965). By 1966, papers on simulation derived from programming a mathematical model had declined, to the extent that none appeared in Volume 9 (1966-1967) although there was a paper using networks and nodes (today part of combinatorics and graph theory) to help design electricity supply networks. The paper utilised tree searching algorithms in order to optimise the design.

A more unusual application was that of a Pegasus computer to compose music (Gill, 1963). This was completed for a BBC Television Programme on 30 August 1962 and involved using the Pegasus to compose music in the style of Schoenberg. Gill's conclusion was honest in that he was 'relatively unmoved' by the music. However, he made a futuristic comment: 'Perhaps ... we shall see musical composition taking the form of a co-operative venture between the human composer and the computer, with the computer supplying a number of plausible passages along lines suggested by the composer...' (Gill, 1963). No doubt Gill would have been impressed by the state of music technology today. The interest in synthesised speech and music was furthered by two papers by Woodward (1966) and

Lavington and Rosenthal (1966). The former noted that as long ago as 1954, S. N. Higgins had produced a common chord from a TREAC computer. However, musical sound was produced only as a step to speech synthesis-for example, the word 'fee' has higher frequency than 'food'. But the computer modelled the sounds by time, not frequency as a musical instrument does, and so there was a basic specification of computer (a minimum instruction time of 20 microseconds) below which sound synthesis was not possible through simple periodicity generation. This could be overcome by using a variety of techniques (Woodward, 1966). Several projects analysing wave forms were being undertaken at this time (1966) including digital analysis of speech waveforms (Lavington and Rosenthal, 1966). These were stored on tape and confusion was noted between similar sounds such as 'F' and 'Th'.

In fact, Manchester University had devoted research energy to speech synthesis in the early and mid-1960s with Lavington, Mathers and Rosenthal all obtaining postgraduate research degrees in this area.

Simulation of a computer system so that it may be evaluated was becoming popular by 1967 as there was concern as to whether the computer systems performed as they should (Huesmann and Goldberg, 1967). Thus while computers were being used for innovative applications such as speech, they continued to support scientific research through out the decade.

Beginnings of artificial intelligence

There are a number of interesting papers throughout the decade that show that computers were being used for games such as 'Dama', which is played on a chess-board (Findler, 1960) and for applications such the change-ringing of Church bells (Papworth, 1960). While these applications appear to be recreationally based, the aspects of machine learning considered in these papers actually reflect the beginnings of artificial intelligence (AI)-a phrase coined by McCarthy in 1956 (Pratt, 1987). However, in later volumes of the Journal in this period, papers about these applications become negligible.

Moreover, papers relating to more philosophical approaches appeared as the decade progressed, including a series of three papers by Musk (1967a, b, c) and a plea from the then President of The BCS, Sir Edward Playfair, that members should be learning from the human sciences, and, in particular, from neurology (Playfair, 1964). However, as early as 1952, Grace Hopper (1952) had sought ways to replace the human brain with a computer. Perhaps influenced by the rise of cognitive science (Brink and Haden, 1989) and the work of John von Neumann (Sejnowski, 1989),

this was reiterated by de Ferranti (1966) who discussed the possibility of designing a computer processor that mimicked the structure of the human brain, likening a neuron [sic] network to an electronic gate and program compilers to the organising aspects of the human brain. Again, this small group of papers reflected the response of the Journal to the work of Simon and Newell associated with machine learning and the seeds of AI (Pratt, 1987).

Next page

• Site Map
• Privacy Policy
• Frequently Asked Questions