[text] Peter Cornwell [e]
»Surveillance of Assailants«, 2000/2001
by Peter J. Cornwell
Towards a Typology of Surveillance
Surveillance of Assailants is an installation work, which creates a simulacrum of the apparatus of urban surveillance systems using artificial intelligence and high performance parallel computing techniques. By analysing video in real time, the work synthesises a virtual surveillance camera through 3D modelling and incorporates live reflections from the installation space. This virtual camera responds directly to the behaviour of viewers.
The result is the palpable presence of an independent software agency in the gallery, which continuously builds a knowledge of viewers and their actions: whether they are watching or looking elsewhere. The virtual camera plays a game of »hide and seek«; feigning disinterest when people stare at it. Alternately, it mercilessly pursues an individual until he or she is out of sight, while ignoring everybody else. Sometimes appearing humorous, the work also parallels potentially sinister technological developments, which are being exploited for surveillance purposes, increasingly beyond the control of any one state or organisation.
Phenomenologically the work produces an image that clearly and overtly scrutinises those who encounter it - that of a solid camera moving in response to viewers' gestures and carrying their reflection from the real world. However, this is achieved by gathering visual information, analysing it and determining how to respond, and then generating moving imagery; all so fast as to appear instantaneous. The techniques and computing power employed by Surveillance of Assailants could also be used to recognise individuals from a database of facial features, enabling the cross-referencing of information about us that is held in government and corporate data structures distributed around the world.
To reveal more clearly the growing primacy of information in social power structures and the extent to which data about individuals has escaped human control, it is necessary to look, as Thomas Levin identifies: »[beyond] more traditional imaging and tracking technologies to the largely invisible but infinitely more powerful practices of what is referred to as 'dataveillance' - that today constitutes the extensive arsenal of social control«. Surveillance machinery has conventionally pre-supposed a clear beneficiary of the information gained. A conspicuous early example of electronic surveillance was the development of early warning radar during WWII, and today the renewed appetite for very costly space-borne systems to detect and potentially intercept missiles in flight attests to the perceived deterrent value of the risk of detection. However, the surveillance machinery that acquired a peculiar glamour during the cold war and much military spying technology in general, is covert. Concealed cameras, in the same way as movement sensors and microphones - the bugging that pre-occupied films like The Conversation and the Watergate investigation of the same period - fail what we might call a »Panoptic test«. They do not establish a disciplinary power structure through continual awareness of the possibility of discovery and ensuing self-regulation of behaviour, as does Bentham's spatial design. Rather, discovery is not anticipated, as was the case following the Tiananmen Square demonstrations when footage from street cameras that were previously considered innocuous was used to deadly effect by the Chinese government.
Whether covert or otherwise, the correlation of surveillance data and beneficiary is no longer so clear. The intersection of ubiquitous data communications and technologies gathering every possible kind of information imaginable into a common digital representation, has created a global trade in information about each of us as individuals. Some repositories, such as banking, insurance, law enforcement, medical, property, taxation and vehicle information are directly accessible only to governments and institutions. Others, built from daily transactions on credit cards, loyalty cards and travel are openly traded; each pigeon-holing us ever more accurately as targets for everything from window frame sales to television license investigation. Even when national legislation attempts to protect information in these databases they are scarcely hermetic. Information about us leaks continually into new data structures: our conversation with the doctor is temporarily secure, but the medication prescribed, the date and the location are immediately loaded on to the pharmacy computers. More information still is accrued by implication. The bar codes on our groceries contribute to the profile of the district in which we live, betraying us even when we resort to cash purchases and enabling the creation of meta-data from other information known about us as individuals.
It is the internet that has etched this individual information fingerprint most deeply. Secure digital financial instruments have made internet retail commonplace and data-mining and autonomous search robots - freely roaming programs charged to search without costly human intervention for as long as necessary - have made the trade in information truly global, beyond the control of governments or corporations [the internet provided the only live and uncensored coverage of events in Tiananmen Square to the rest of the world]. The involuntary traces that we leave everywhere on the internet are also more difficult to recognise, and almost impossible to recall. Many internet sites that we visit, especially those of technology companies, secretly interrogate our computers; analysing our habits, reporting the serial numbers of products that we have loaded and leaving information for later exploitation. Our information fingerprint is now a persistent and monotonically expanding trace of our lives, distributed among countless computers across the globe and in space and independent of any specific surveillance agenda. We can no longer determine all of the information that is recorded about us and it will continue to accrue long after our lives are completely forgotten by distant descendants.
However, in contrast to transaction-based data gathering, surveillance imagery is a much more potent source of information about the individual. Combined with the ability to search and cross-reference globally distributed databases, automated image analysis is poised to revolutionise information privacy. The current explosion in numbers of camera installations in public spaces for the »safety of the individual« is only a faint intimation of the forthcoming integration of intelligent imagers with every possible appliance, structure and vehicle.
Closed circuit television cameras have been replaced by the semiconductor imager - the camera on a chip - that is already part computer. Fabricated in hundreds of millions of units from cheap raw materials through the economies of the electronics industry, these computer cameras have become ubiquitous, emulating older analogue equipment where necessary to gain wider market share. However, instead of a television picture waveform, the native format of semiconductor imagers is data, of potentially very high resolution, and they consume very little power. Gone are the colour errors and limited resolution of analogue electronics and expensive conversion equipment for computers. Already it is possible to purchase a computer camera which has only a network address and which is never intended to produce a video tape or printed image, for less than $300.
Once in digital form, image information can be manipulated mathematically; analysed using an armory of software and duplicated without error, almost instantly, on the other side of the world. For example, instead of sending an image of a speeding car, the license plate can be »read« from the image and just those characters, plus the speed, time and location forwarded for automatic issue of a fine. The symbolic information about the car occupies less than one thousandth of the space of the original image. Offences can be transmitted and processed immediately; but every car passing every camera can be relentlessly recorded, each event remaining accessible for future analysis by unknown agencies.
The potential of automatic derivation of higher order information from live imagery is highly significant, having both far-reaching cultural implications and marking a major departure in the functioning of surveillance. The huge numbers of cameras already installed in public spaces, retail outlets, restaurants and innumerable other locations, exceed our capacity to store timecoded imagery exhaustively. However, intelligent cameras which determine for themselves what to record, or better, derive symbolic information such as an ID from facial analysis, allow surveillance information not only to be kept in perpetuity, but also to be filed, correlated and continually analysed, all without human intervention . Freeing visual surveillance from the need for operators to analyse all of the imagery captured, enables far greater numbers of cameras, at ever decreasing costs, to be managed effectively.
It is difficult to resist comparison with the science fiction writing of Bob Shaw, who in The Light of Other Days postulates a culture in which the state has scattered billions of tiny fragments of a material called »slow glass«, which permeate every space of the planet  . Shaw's slow glass, like ordinary glass, delays the passage of light compared to its speed in vacuum. However, light might emerge from slow glass days or even years later, depending upon how it is made, enabling imagery to be replayed from the scene of a crime simply by gathering up glass fragments that »witnessed« the event. No action nor even fleeting facial expression is finally lost for many years. Rather, its image, from several perspectives, continually emerges from one slow glass particle or another.
Our own society, littered with intelligent cameras will have much in common with Shaw's, with the notable distinction that symbolic information from each moment, if not the image itself, will persist for long periods - possibly for the duration of our culture itself. This contrasts notably with Wolfgang Welsch's notion of the fugitive "lightness" of imagery derived from information  . The symbolic information scavenged by intelligent cameras and flowing into data structures worldwide is potentially very long-lived. Unlike the data cards in our digital cameras, symbolic information in databases is never discarded: rather, ever more sophisticated techniques are developed to store and access it whenever the capacity of our systems is exceeded.
Technically, formidable problems remain in developing image analysis software that could be employed routinely to identify individuals from facial characteristics with complete reliability, although commercial software already achieves this in a controlled environment. Cheap intelligent cameras are already being used for the identification of gestures either for use in more natural human-computer interfaces or to determine when to register events or record images. Although we are aware of the multiplicity of cameras around us and of the conspicuous control room screens and operators in environments such as public transport, we do not know if we are merely observed or whether the imagery is recorded; or if the recordings are discarded periodically. Here the logic of the Panopticon is extended temporarily by the persistence of information and the ensuing possibility of its future use against us.
We are less aware of the trade in information about us which is derived from transactions such as credit card purchases. Nagging doubts might pursue us when a personally addressed invitation to take out a loan arrives while we have cash flow problems, but the trade in such information, its currency and continual analysis, is not so prominent as to influence the behaviour of many people. Less obvious still, is the increasingly automated agency by which this data is manipulated. It is not clear who has access to which elements of our information fingerprint at a given time. Corporations, states, institutions and individuals all hold and exchange information about us, but so do independent software programs - fugitive artificial intelligences instructed by potential beneficiaries of information, but which search networks and databases hindered only by other programs - data defences - until they fulfill their goals. Few viruses have yet appeared which gather and manipulate information about individuals, but it is possible that such agencies might be created which specifically seek to conceal the beneficiary of their activity or indeed, themselves become completely autonomous. The field of genetic algorithms which actually synthesise new software without human intervention is beyond this discussion, but it seems that autonomous software agencies will increasingly become the only means of determining the location and accessibility of personal information. Once again science fiction has already visited this territory - here William Gibson, who describes precisely such an eventuality in Neuromancer  .
With the advent of ubiquitous intelligent cameras we have departed completely from the logic of Bentham's Panopticon. Instead of a spatial design ensuring the possibility of observation, we are inextricably woven into the fabric of a virtual information space. Surveillance no longer operates according to a clearly defined logic. Rather, the complexity of the virtual space and the speed with which information is created and manipulated within it is now forever beyond our direct comprehension.
1 Jaques Derrida, Of Grammatology, reprint edition, trans. Gayatri Chakravorty Spivak, Johns Hopkins University Press, January 1998. Some far-sighted observations about information technology arise in Derrida's Of Grammatology which, concerned with the science of writing, was published in the same years as the first geographic links between computers were developed. Derrida identifies the writing of »theoretical mathematics« as being »where the practice of scientific language »challenges intrinsically and with increasing profundity the ideal of phonetic writing and all its implicit metaphysics ...« and speculates that »the theory of cybernetics is by itself to oust all metaphysical concepts...«. He proceeds to observe that »development of the practical methods of information retrieval [which] extends the possibilities of the 'message' vastly, to the point where it is no longer the 'written' translation of a language, the transporting of a signified which could remain spoken in its integrity...« ^
2 Bob Shaw, Other Days, Other Eyes, Victor Gollancz Ltd, 1972. ^
3 Wolfgang Welsch, Undoing Aesthetics, trans. Andrew Inkpin, Sage publications, 1997. In Artificial Paradises Welsch discusses the phenomenology of electronic worlds and in particular addresses the ontology of computer-readable information. ^
4 William Gibson, Neuromancer, Ace Books, 1994. ^