AISB Symposium on the Turing Test

Citation

Andrew, A.M. (2009), "AISB Symposium on the Turing Test", Kybernetes, Vol. 38 No. 1/2. https://doi.org/10.1108/k.2009.06738aac.003

AISB Symposium on the Turing Test

Article Type: Conference reports From: Kybernetes, Volume 38, Issue 1/2

A symposium designed to present a formal academic critique of issues around what has become known as the Turing Test was sponsored by the Society for the Study of Artificial Intelligence and Simulation of Behaviour (www.aisb.org.uk) and held on Sunday 12 October 2008 in the Palmer Building of the University of Reading. The organiser was Dr Mark Bishop of Goldsmiths College of the University of London, formerly a member of the Cybernetics Department at Reading, after graduating there.

The Turing Test, or Imitation Game, was proposed in the famous paper of Turing (1963) as a criterion of machine intelligence. The idea was that a machine should be accepted as intelligent if a person, communicating with it through some digital interface (so as to eliminate clues from appearance, tone of voice, etc.) is unable to tell whether the interaction is with a machine or another person.

The test is the basis of an annual “Loebner Prize” competition (www.loebner.net/Prizef/loebner-prize.html) in which submitted computer programs, referred to by the unflattering term “chatterbots” are tested in the Turing situation. To give the chatterbots a chance of success, the time of interaction is limited to five minutes and there is some restriction of subject-matter. The criterion of success is based on an early prediction that the test, in this form, would be passed by the year 2000, with success judged as the interrogator being right in less than 70 per cent of trials, and consequently wrong in at least 30 per cent. Full details are given on: www.eecs.harvard.edu/shieber/Biblio/Papers/loebner-rev-html/loebner-rev-html.html

The venue and date for the AISB Symposium were chosen to coincide with the 2008 Loebner Prize competition in Reading University. Professor Kevin Warwick was in fact slightly late in arriving because he had appeared to talk about the test on that morning’s BBC Breakfast Television. He arrived with the announcement that the 30 per cent target had not been achieved, though 20 per cent had been, and later in the day he was able to up this to 25 per cent. Later still there was the presentation of a Loebner bronze medal and a cheque to the student who had authored the most successful chatterbot (It would have been a gold medal if 30 per cent had been reached.)

The opening presentation of the Symposium was by Baroness Susan Greenfield who has a chair in Pharmacology in Oxford and specializes in neuroscience with particular attention to Parkinson’s disease, and is author of several books and active in popular presentation of science. Her topic was: “The brain, consciousness and the Turing Test”. Her dynamic treatment of this ranged widely, with mention of the walking robot developed in Japan and named Qrio, but mainly dealing with relevant data from neurophysiology. Consciousness appears to be a distributed phenomenon, a view that was supported by showing NMR scans of brain activity of volunteers at various stages of anaesthesia, where the fading of activity was over a large area. The importance of chemical neurotransmitters was emphasized, with the observation that a neuron might help determine firing of another with which it has no synaptic connection.

Baroness Greenfield also referred to the influence of conscious experience on neural development, with pictures comparing the amounts of branching of neurons in the brains of rats raised in environments that were, respectively, rich and poor in variety of experience. It was also mentioned that the brains of London taxi drivers are unusually well developed in the hippocampus, attributable to the requirement of committing the map of London streets to memory. She conceded, however, that the “hard problem” of explaining the relationship of mind and consciousness remains unsolved.

The other speakers in the morning were Professor Selmer Bringsjord of Rensselaer Polytechnic Institute, USA, whose title was: “God, Souls and Turing”, and Dr Michael Wheeler of Stirling University, on: “Plastic machines: behavioural diversity and the Turing Test”. Both of them referred to objections to the possibility of machine intelligence that were acknowledged and countered by Turing. Professor Bringsjord defended the view of Descartes, that mind and matter are different things and that a machine can never truly think or be conscious. Dr Wheeler referred to the objection from complexity, or the argument that a machine can never have the diversity of behaviour of a human. He sympathized with the view, not to the extent of defending “never” but certainly “only with great difficulty”, and argued that the difficulty stems from much more besides a need for vast information storage capacity.

The first two presentations of the afternoon session were by Dr Andrew Hodges of Oxford with the title: “Alan Turing’s Imitation Game” and Professor Luciano Floridi, of the Universities of Hertfordshire and Oxford on: “The Turing Test as an instance of The Method of Abstraction”. The former of these was an extremely detailed and sympathetic account of the life and works of Turing, by this author of a definitive biography. The second was an analysis of the test situation from a particular viewpoint.

Following the afternoon coffee break, and handing-over of the Loebner bronze medal, there were presentations by Professor Maggie Boden OBE of Sussex University on: “The Turing Test and artistic creativity” and by Professor Owen Holland of the University of Essex on: “The argument from (machine) consciousness”. Professor Boden is an eminent figure in AI and author of a classic text (Boden, 1977). In her talk, she acknowledged that computers are used by artists as tools (particularly with Adobe Photoshop) but she concentrated on art forms that were:

• •

  not possible without a computer; and

• •

  accepted for display alongside manually-produced art, and she was able to describe a surprising number of examples satisfying both criteria.

The first requirement is most readily satisfied by art that interacts with its audience and a particularly successful example was an interactive projection of a starfish in the Millennium Dome, which several commentators considered to be the most impressive item there.

Reference was made to music as well as to visual art, and to a program that can compose music in the style of a given composer, such that experts accept the generated output as genuine. Reference was also made to achievements by the late Professor H. Longuett-Higgins relating to music performance, where for example the program implemented an instruction such as a crescendo or decrescendo sign in a manner appropriate to the musical context, indistinguishable from the response of a human performer.

Professor Holland devoted part of his time to early history of cybernetics in Britain, and the involvement of Turing with the Ratio Club, somewhat overlapping his contribution to the 2006 Annual Conference of the UK Cybernetics Society, reported in Kybernetes, 36/1, 2007, pp. 118-21. He then described current work aimed at achieving machine consciousness, and in particular a robot that includes an internal model of itself and its immediate environment, such that planned actions can be executed first by the internal model. The operation was illustrated by video clips.

The intention was to close with a Panel Discussion involving all the speakers and this was embarked on but had to be cut short because the conditions of use of the theatre required it to be vacated by 6’O clock. This was not a serious loss since the proceedings had already given plenty to go away and think about. It was a stimulating event and there are plans to publish the papers in a special issue of Kybernetes.

Alex M. Andrew