Hostname: page-component-586b7cd67f-l7hp2 Total loading time: 0 Render date: 2024-11-23T20:39:52.707Z Has data issue: false hasContentIssue false
  • English
  • Français

What is science for? The Lighthill report on artificial intelligence reinterpreted

Published online by Cambridge University Press:  10 July 2020

JON AGAR
JON AGAR*
Affiliation:
Department of Science and Technology Studies, University College London. Email: [email protected].
Get access Rights & Permissions [Opens in a new window]

Abstract

This paper uses a case study of a 1970s controversy in artificial-intelligence (AI) research to explore how scientists understand the relationships between research and practical applications. It is part of a project that seeks to map such relationships in order to enable better policy recommendations to be grounded empirically through historical evidence. In 1972 the mathematician James Lighthill submitted a report, published in 1973, on the state of artificial-intelligence research under way in the United Kingdom. The criticisms made in the report have been held to be a major cause behind the dramatic slowing down (subsequently called an ‘AI winter’) of such research. This paper has two aims, one narrow and one broad. The narrow aim is to inquire into the causes, motivations and content of the Lighthill report. I argue that behind James Lighthill's criticisms of a central part of artificial intelligence was a principle he held throughout his career – that the best research was tightly coupled to practical problem solving. I also show that the Science Research Council provided a preliminary steer to the direction of this apparently independent report. The broader aim of the paper is to map some of the ways that scientists (and in Lighthill's case, a mathematician) have articulated and justified relationships between research and practical, real-world problems, an issue previously identified as central to historical analysis of modern science. The paper therefore offers some deepened historical case studies of the processes identified in Agar's ‘working-worlds’ model.

Type
Research Article
Information
The British Journal for the History of Science , Volume 53 , Issue 3 , September 2020 , pp. 289 - 310
Check for updates
Copyright
Copyright © British Society for the History of Science 2020

Access options

Get access to the full version of this content by using one of the access options below. (Log in options will check for institutional or personal access. Content may require purchase if you do not have access.)

Footnotes

This paper benefited immensely from discussion at the Science, Technology and Innovation Studies (STIS) seminar, Edinburgh, November 2017, as well as with colleagues at UCL. My thanks, also, to UCL Special Collections for access to the Lighthill papers, to the British Library for assistance with the Michie papers, to Amanda Rees, to Trish Hatton and to the anonymous reviewers.

References

1 Agar, Jon, Science in the Twentieth Century and Beyond, Cambridge: Polity, 2012, p. 3Google Scholar.

2 Agar, op. cit. (1), p. 4.

3 James Lighthill, ‘Artificial intelligence: a general survey’, published as part of Science Research Council, Artificial Intelligence: A Paper Symposium, London: SRC, 1973, pp. 1–21.

4 ‘I have never done anything “useful”. No discovery of mine has made, or is likely to make, directly or indirectly, for good or ill, the least difference to the amenity of the world. I have helped to train other mathematicians, but mathematicians of the same kind as myself, and their work has been, so far at any rate as I have helped them to it, as useless as my own. Judged by all practical standards, the value of my mathematical life is nil’. Hardy, G.H., A Mathematician's Apology, Cambridge: Cambridge University Press, 1940, p. 49Google Scholar.

5 Crevier, Daniel, AI: The Tumultuous History of the Search for Artificial Intelligence, New York: Basic Books, 1993, p. 117Google Scholar

6 See also, for example, Nilsson, Nils J., The Quest for Artificial Intelligence: A History of Ideas and Achievements, Cambridge: Cambridge University Press, 2010, p. 204, p. 282Google Scholar. There is a surprising absence in the secondary literature of survey histories of artificial intelligence written by professional historians of science. There is, however, a growing corpus of excellent, more specific studies, for example: Adam, Alison, Artificial Knowing: Gender and the Thinking Machine, London: Routledge, 1998Google Scholar; Edwards, Paul N., The Closed World: Computers and the Politics of Discourse in Cold War America, Cambridge, MA: MIT Press, 1996Google Scholar. Dick, Stephanie, ‘AfterMath: the work of proof in the age of human–machine collaboration’, Isis (2011) 102, pp. 494–505Google ScholarPubMed. Ensmenger, Nathan, ‘Is chess the drosophila of artificial intelligence? A social history of an algorithm’, Social Studies of Science (2012) 42, pp. 5–30CrossRefGoogle ScholarPubMed.

7 Lenat, D.B., ‘The voice of the turtle: whatever happened to AI?’, AI Magazine (2008) 29(2), pp. 11–22, 16Google Scholar.

8 Anthony Tucker, ‘Sir James Lighthill: maths in motion’, The Guardian, 21 July 1998, p. 16. The obituary is riddled with errors, not least consistently calling Lighthill ‘Lightman’.

9 David George Crighton, ‘Sir James Lighthill FRS’, The Independent, 22 July 1998, p. 10. Lighthill's 1952 paper launched the field of ‘aeroacoustics’ (‘sound generated aerodynamically’) and, says Crighton, was remarkable for two reasons: first, Lighthill delayed publication for sixteen months so that he could rewrite it in a form understandable to aeroengine designers, and second, the ‘paper neither contains nor needed as much as a single reference to any prior work’.

10 Lighthill papers, UCL Special Collections (hereafter MJL) D4, untitled document beginning ‘Ladies and Gentleman’, is his resignation speech to RAE staff, and includes a review of achievements and developments.

11 ‘Sir James Lighthill’, Daily Telegraph, 20 July 1998, p. 23.

12 Another obituary says six times. Also from this Telegraph obituary (op. cit. (11)): he was the first person to swim around Sark, in 1973. He used a ‘two-arm, two-leg backstroke, thrusting with the arms and legs alternately’.

13 Crighton, op. cit. (9).

14 MJL K686, Vick to Lighthill, 25 January 1966.

15 Francis Edgar Jones – Frank Jones – had worked during the war at the Telecommunications Research Establishment (TRE), leading the Oboe development team, his managerial competence providing a pragmatic complement to the more disorganized but inventive A.H. Reeves. Unlike many of the TRE high-flyers, he did not return to academic research and instead supported the attempts to build a post-war role for TRE. In 1953, six years before Lighthill himself arrived, Jones joined the Royal Aircraft Establishment, becoming deputy director (equipment). In 1956 he was recruited by the electronics company Mullard Ltd, where he remained until retirement in 1972. Like Lighthill he championed the application of science. For example, with Mullard's John Bunton he persuaded the Royal Society to mint a new prize in 1967, the Mullard Medal, because both ‘felt that a medal for technology and applied science on the same level as the Copley Medal could make a contribution in influencing scientific effort in that direction’. He was an accomplished builder of networks between industry, government and science. While his Royal Society obituarist does not mention Jones's role in the Culham report, he does note Jones's chairing of the more publicized output of the Working Group on Migration of Scientists, the 1967 ‘Brain Drain’ report. MacFarlane, George G. and Hilsum, Cyril, ‘Francis Edgar Jones, 16 January 1914–10 April 1988’, Biographical Memoirs of Fellows of the Royal Society (1988) 35, pp. 179–99, 194, 196Google Scholar.

16 MJL K694, ‘Culham Laboratory Review Panel. Minority report by Dr F.E. Jones and Professor M.J. Lighthill’, undated (1966).

17 MJL K694, ‘Minority report’, p. 1.

18 MJL K694, ‘Minority report’, p. 7.

19 MJL K668, Flowers to Lighthill, 20 September 1971.

20 MJL K668, draft letter from Lighthill, undated.

21 MJL K668, ‘Investigation of artificial intelligence research’. The external list included, among computer scientists, R.A. Brooker (Essex), R.M. Needham (Cambridge), D.C. Cooper (Swansea), P.J. Landin (Queen Mary); and among neurobiology, linguistics and other disciplines, R.L. Gregory (Brian Perception Unit, Bristol), H.C. Longuet-Higgins (Machine Intelligence and Perception, Edinburgh), R.C. Oldfield (MRC Speech and Communications Research Unit, Edinburgh), D. Fry and A.J. Fourcin (Phonetics, UCL), D.M. MacKay (Communications, Keele), S. Brenner (MRC Laboratory of Molecular Biology, Cambridge), J.N. Holmes (Joint Speech Research Unit, General Post Office) and J.J. Kulikowski (UMIST). The rough notes of replies shows that several others were also contacted, including Christopher Strachey (Oxford), Tom Kilburn (Manchester) and D.E. Broadbent (Applied Pyschology Unit, Cambridge). In the report itself, Lighthill lists the following: J. Annett, H.G. Barrow, S. Brenner, D.E. Broadbent, R.A. Brooker, O.P. Buneman, R.M. Burstall, A.D.B. Clarke, M.B. Clowes, A.H. Cook, D.C. Cooper, J.E. Doran, J.F. Duke, E.W. Elcock, I.J. Good, C.C. Green, R.L. Gregory, P.J. Hayes, A.L. Hodgkin, J.N. Holmes, J.A.M. Howe, D.H. Hubel, S. Isard, H. Kay, T. Kilburn, J. Kulikowski, D.N.L. Levy, H.C. Longuet-Higgins, D.M. MacKay, D. Marr, J. McCarthy, B. Meltzer, D. Michie, M. Minsky, D. Mollison, E. Moore, J.S. Moore, R.M. Needham, N.J. Nilsson, C. Oldfield, J.V. Oldfield, I. Pohl, R.J. Popplestone, B. Raphael, J.A. Robinson, C. Strachey, N.S. Sutherland, M.M. Swann, H.P.F. Swinnerton-Dyer, D. Wilshaw, T. Winograd.

22 I discuss some below. In addition, extracts of Christopher Strachey's response, 15 February 1972, is in Add MS 88958/1/383.

23 MJL K669, Lighthill to Michie, 16 February 1972.

24 MJL K669, Lighthill to Cook, 18 February 1972. Lighthill also took the opportunity to congratulate Cook on his imminent move to Cambridge to become Jacksonian Professor. With Cook ending his direct tie with Edinburgh and moving to Lighthill's university, the briefing could have been particularly frank.

25 MJL K670, Lighthill to Flowers, 23 March 1972. A later estimate of expenses shows that Lighthill had worked thirty-nine whole days, for which he charged £12.60 per day, a total expense of £491.40.

26 MJL K670, Lighthill to Flowers, 27 April 1972.

27 MJL K670, Flowers to Lighthill, 20 April 1972.

28 This group would be chaired by Flowers and include Lighthill, Dr D.E. Broadbent, Dr R.M. Needham, Professor G.S. Brindley (Institute of Psychiatry, London), Professor R.A. Brooker (Essex) and Professor N.S. Sutherland.

29 Lighthill, op. cit. (3), p. 3.

30 Lighthill, op. cit. (3), p. 6.

31 Lighthill, op. cit. (3), p. 2.

32 Lighthill, op. cit. (3), p. 8. Lighthill did identify failures in categories A and C. For example, mathematical theorem proving, category A work, ‘is particularly an area where hopes have been disappointed through the power of the combinatorial explosion in rapidly cancelling out any advantages from increase in computer power’.

33 Lighthill, op. cit. (3), p. 7. Lighthill included a postscript that dealt with Terry Winograd's PhD on the SHRDLU program, for which Lighthill had high praise, while noting, ‘One swallow does not make a summer’ (p. 17). It was successful category B work, although the best further developments would come through category C.

34 Lighthill, op. cit. (3), p. 19, my emphases.

35 Lighthill, op. cit. (3), p. 19, my emphasis.

36 Lighthill, op. cit. (3), p. 20. He continued, ‘The structuring and utilisation of scientific data bases is another area where good results depend on detailed study of the data's special characteristics. The one part of that field with which the present author has been closely involved, as Chairman of the Steering Committee for the Experimental Cartography Unit of NERC since its inception, affords a good example of this: the structuring of geographically located data is found to demand quite specialised techniques, closely related to the cartographic character of the output’.

37 N. Stuart Sutherland, ‘Part II Some comments on the Lighthill report and on artificial intelligence’, in SRC, op. cit. (3), pp. 22–31.

38 Roger Needham, ‘Comment by Dr R.M. Needham’, in SRC, op. cit. (3), pp. 32–34. ‘Since I basically agree with Lighthill''s conclusions there is perhaps less to say than in Sutherland's commentation’ (p. 32).

39 Longuet-Higgins, ‘Comment by Professor H.C. Longuet-Higgins FRS’, in SRC, op. cit. (3), pp. 35–37. ‘Sir James places in his category C all the artificial intelligence work which he regards as scientifically promising, and refers to this category as Computer based studies of the central nervous system. In so doing he aligns himself with those of us who hold that the main justification for artificial intelligence is the light it can throw upon human intellectual activity. But his chosen heading, and some of his later remarks, indicate that he attaches more significance to work on the hardware of the brain than to work on its software. This is the only point on which I want to take issue with him’ (p. 36).

40 ‘In short whatever the technological prospects of artificial intelligence, its principal scientific value, in my view, is that it sets new standards of precision and detail in the formulation of models of cognitive processes, these models being open to direct and immediate test’. Longuet-Higgins, op. cit. (39), p. 36.

41 Longuet-Higgins, op. cit. (39), p. 37.

42 Michie, Donald, On Machine Intelligence, Edinburgh: Edinburgh University Press, 1974Google Scholar. Fleck, James, ‘Development and establishment in artificial intelligence’, in Elias, Norbert, Martins, Herminio and Whitley, Richard (eds.), Scientific Establishments and Hierarchies, Sociology of the Sciences (1982) 6, pp. 169–217, 183CrossRefGoogle Scholar.

43 Alex May, ‘Dane Anne Laura Dorinthea McLaren (1927–2007)’, Oxford Dictionary of National Biography, at https://doi.org/10.1093/ref:odnb/98949. Anne McLaren and Donald Michie married in 1952 and divorced in 1959, but remained close friends. Donald Michie resigned from the Communist Party of Great Britain in 1956, while McLaren remained a member.

44 The meeting at Firbush Point, Loch Tay, involved two British (both Edinburgh, Michie's Department and Meltzer's Department of Computational Logic), one Russian (Mathematical Institute of the USSR Academy of Sciences, Novosibirsk) and four American bodies (SRI, MIT, Syracuse and ARPA). Firbush News (June 1972) 2, p. 4, reproduced a letter from Lighthill.

45 British Library Add MS 88958/1/384. Copy of first thirty-five pages of Flowers's submission to the SRC, sent to Michie by ‘J’, who had been given a copy by Freddy Lines (presumably Alfred Lines, who had worked, like Lighthill, at RAE).

46 Donald Michie, ‘Comments on the Lighthill report and the Sutherland reply’, in SRC, op. cit. (3), pp. 38–45, 38. Michie would continue to complain that Lighthill had been unqualified, too hasty, distracted and unwilling to consult AI experts widely enough (especially outside the UK), well into the 1980s. See ‘Artificial Intelligence research’, prepared for submission to the House of Lords Select Committee on Science and Technology and drafted as a letter to Nature, 12 January 1980, British Library Add MS 88958/1/382.

47 British Library Add MS 88958/1/385, Michie, ‘On first looking into Lighthill's “Artificial Intelligence”’, August 1972. Michie's own report on the field, delivered to the SRC, Computing Science in 1964: A Pilot Study of the State of University-Based Research in the United Kingdom, London: Science Research Council, 1965, had represented work as a ‘dumbbell’, with the bells A (‘Advanced Computer Science and Automation) and C (Computer-based CNS Research) linked by a connecting piece – ‘or bridge’ – marked with a question mark and seen as a ‘channel of communication’. Michie placed ‘Machine Intelligence’ in bell A, near the bridge, and ‘Computational Psychology’ in bell B, also near the bridge. The two pictures of the field – Michie's in 1964 and Lighthill's in 1972 – are so similar that Lighthill must have, directly or indirectly, drawn his map of artificial intelligence after Michie's own visualization – a remarkable irony.

48 Michie, op. cit. (46), p. 39.

49 Michie, op. cit. (46), p. 41. Some of the problems to be solved were discussed in Michie, Donald, ‘Machines and the theory of intelligence’, Nature (23 February 1973) 241, pp. 507–512CrossRefGoogle Scholar.

50 Donald Michie, ‘Machine intelligence in the bicycle shed’, New Scientist, 22 February 1973, pp. 422–423.

51 See YouTube https://youtu.be/yReDbeY7ZMU, and also via Edinburgh's current Artificial Intelligence Applications Institute website, at www.aiai.ed.ac.uk/events/lighthill1973. McCarthy's response can be found in the review of the paper symposium, published in Artificial Intelligence (1974) 5, pp. 317–322. ‘If we take the [ABC categorization] seriously’, wrote McCarthy, ‘then most AI researchers lose intellectual contact with Lighthill immediately’.

52 The ‘concentration of talent, the surfeit of publicity, and perhaps more than anything else, the predominance of research over teaching in AI, attracted hostility from other departments weighed down with heavy teaching responsibilities’. Fleck, op. cit. (42), p. 190.

53 British Library Add MS 88958/1/383, Charles A. Rosen (SRI) to Flowers, 1 June 1973. British Library Add MS 88958/1/384, Nils J. Nilsson and ten others (from SRI, Stanford, MIT and University of Southern California) to Flowers, 16 May 1973. Rutovitz to Edwards, December 1973, proposes the example of a planned Jet Propulsion Laboratory Mars ‘crawler’ as a practical case of category B work.

54 ‘Report of Long-Range Computing Research Panel’, Appendix 1 of Policy and Programme Review of Research and Training in Computing Science, London: Science Research Council, 1972.

55 ‘The one postgraduate course (at Edinburgh) places too much emphasis on conventional mathematics and logic and does not capture the essence of the subject as understood in the US’.

56 Fleck, op. cit. (42), pp. 209, 205.

57 Fleck, op. cit. (42), p. 205.

58 Turing, Alan, ‘Computing machinery and intelligence’, Mind (1950) 49, pp. 433–460Google Scholar.

59 Note, however, that Fleck refers in passing to Michie's projects as actually being entrepreneurially commercial: ‘Michie favoured a rather swashbuckling style of directing large team projects oriented to goals which could be linked with industrial applications, and, in fact, was involved in launching a university-based company to market compiler systems and other software for the POP-2 language, which was developed in the department’. Did Lighthill know about, or ignore, or set aside this applicability?

60 MJL D8, Lighthill, ‘Transcript of the speech made by the Director to members of Chemistry Department and Metallurgy & Physics Department at the Assembly Hall on the 1st November 1962’, 1962. He uses the boundary-condition metaphor again in MJL K14, Lighthill, ‘Synopsis of speech given at the Fourth British Theoretical Mechanics Colloquium, Bristol, 1962’, 14 March 1963.

61 MJL K14, Lighthill, ‘Synopsis of speech given at the Fourth British Theoretical Mechanics Colloquium, Bristol, 1962’, 14 March 1963.

62 My emphasis.

63 ‘One essential thing to a successful applied mathematics school is closeness of contact of at least some of its members with areas where the work is successfully applied’. ‘Summing up, the applied mathematician to-day has to span the whole quadrilateral [rigorous mathematical reasoning – full numerical computation – equivalent physical argument – application to practical problem] and to forge clearly, also, all the links I have shown here’.

64 The articles of association are in MJL K14. The first registered address was 29 Gordon Square (Gordon House). Its working office was in Maitland House, Warrior Square, Southend-on-Sea, Essex.

65 MJL K28, Sir William Hodge (Joint Mathematical Council) and Lighthill, draft funding application to Leverhulme Foundation, undated (1964). MJL K30, Lighthill to Pack, 11 February 1965.

66 MJL K29, J.T. Cambridge to Lighthill, 1 January 1965.

67 MJL K36, ‘Remarks by the President on presenting a certificate of election to the Inst. First Honorary Fellow. Sir Geoffrey Taylor’, undated.

68 MJL J35, Lighthill, ‘Bridging the chasm separating examination questions (even “difficult” ones) from real-world problems (even “easy” ones)’, lecture delivered 4 May 1977.

69 His capitalization. Lighthill and others published a book with Penguin, Newer Uses of Mathematics (1977), that gave examples.

70 K685, Kingman to Lighthill, 21 October 1982.

71 K685, Needham and Swinnerton-Dyer, ‘Artificial Intelligence research in the UK’, October 1982.

72 Needham and Swinnerton-Dyer made a number of recommendations, one of which was for the strengthening of Edinburgh, a ‘significant upgrading of its staff and facilities’.

73 K685, Lighthill to Kingman, 17 November 1982, my emphasis.

74 My emphasis. The difficulties were the reason for the departure of Richard Gregory and Christopher Longuet-Higgins, Lighthill asserted. He went on to say that ‘the change of direction and of organisational structure that followed my Report greatly helped AI work at Edinburgh to flourish. You may care to ask John Burnett, the Principal, for his view on this (the view of the former Principal, Michael Swann, was identical). He will tell you how the atmosphere engendered by my Report allowed the University to adopt arrangements with the main work in AI proceeding very satisfactorily while Michie moved to the sidelines (and, this year, he accepted the University's offer of early retirement)’. For the creative destruction of interdisciplinarity, specifically the destruction of the bridge between biological and computer scientists, see also Lighthill, James, ‘“The development of AI has been very much as I forecast”: interview with Sir James Lighthill’, Machine Intelligence News (1986) 2(5), pp. 3–4Google Scholar.

75 My italics, but I have no doubt that this is Lighthill's emphasis. He continued, ‘oriented towards neurobiology but failing to recognise its biological and biochemical complexity. They want AI to dabble in the neurosciences, which are prospering, at the expense of British industry, which is not!’

76 British Library Add MS 88958/1/385, Donald Michie, ‘On first looking into Lighthill's “Artificial Intelligence”’, August 1972.

77 The original source is probably Edward Feigenbaum, ‘Artificial intelligence. Themes in the second decade’, Advanced Research Projects Agency (ARPA) Memo No. 67, 15 August 1968. Feigenbaum, quoting a view from within the SRI group, wrote that ‘the most economic and efficient store of information about the real world is the real world itself’.

78 British Library Add MS 88958/1/385, Yorick Wilks and Bruce Anderson, untitled, 6 December 1982.

79 British Library Add MS 88958/1/385, Michie to Wilks, 13 December 1982, original emphases.

80 Agar, Jon, Science Policy under Thatcher, London: UCL Press, 2019, esp. pp. 88–99CrossRefGoogle Scholar.