by Andrew Dolan
It is often asked where the future of policing lies. Many commentators, beguiled by the emergence of novel and innovative technologies – much of it mislabeled as artificial intelligence (AI) – see future law enforcement as a blend of ‘Robocop’ and ‘Minority Report’ cinematic cliches, which often says more about Hollywood than Silicon Valley.
In order to answer the question, a useful point of departure is to seek clarity on where AI itself is trending. Behind the commercial marketing and technological incubation, there remains a constant for what is likely to stimulate AI development, namely the enhancement of data, the dash for greater computation power and algorithmic improvement.
There undoubtedly has been significant movement in these areas and indeed one might also add the notable development of network platforms. Microprocessor capacity, steady incremental leaps in the power of deep neural networks and stunning enhancements in voice and image recognition are all adding to a more nuanced and networked society, which offers widely diverging potential and avenues for societal exploitation.
Therefore it is within this set of technological developments that future applications of artificial intelligence is likely to evolve and with it, enhanced capabilities for the law enforcement community.
Of course identifying precisely where the new technologies might develop is fraught with imprecision; we are probably incorrect in our assumptions but close enough to identify the contours of where we are likely to end up. It seems likely that we are already at the foothills of significant operational enhancement through AI.
It seems safe therefore to predict that future policing will benefit enormously from AI-enhanced image recognition platforms, ranging from the intimate image capturing from body or helmet-mounted cameras to various forms of earth observation provided by surveillance applications associated with unmanned aerial vehicles or drones.
Steady improvement in image recognition of course depends largely on the volume and quality of images used to develop the appropriate software algorithms. It was also require forms of AI to process, interrogate or exploit multiple data streams, much of it in a time sensitive fashion. Think of the data and algorithmic complexity associated with data fusion comprising data streams from surveillance vehicles, body cameras, drones, CCTV and most likely, domestic applications like doorbell cameras or car cockpit cams. Throw in acoustic signatures associated with high-risk area policing or mini satellites or bots and one can imagine that only some forms of AI might be able to handle this at an economically viable cost.
Another secure prediction involves AI in predictive policing strategies. The exploitation of data is very often at the heart of technological improvements. It is when this technological benefits from significantly greater volumes of data, much of it personal, smarter algorithms – perhaps even developed by other smart machines and greater computation power that one is likely to see enhancements in predictive power being utilized by police forces, including key functions such as investigation, risk assessment, operational deployment, routine and special surveillance of communications and resource management.
Equally likely is the continued use of AI to support more tangential or less time sensitive law enforcement functions such as so-called ‘cold case investigation’, parole adjudication or laboratory analytics. The future police force will need more brain and less brawn.
One final prediction involves robotics. This will not be the advent of ‘terminator-style’ partners but rather an early exploration of various robotic platforms that might offer regular forces with logistic capabilities, enhanced personal protection within vehicles, crowd dispersal functions or – in certain exceptional circumstances – kinetic force. A careful look at military sector AI suggest that some forms of robotic capability is not unlikely and the day of the drone platform with a loitering capability and a payload range including lethal force or ‘swarms’ of ‘nanotechnology’ in the form of bots could enhance police situational awareness in real time and in extreme hostile urban environments.
These predictions I think are ‘safe’ as they are, to an extent. already here in some shape for form. It is also not unrealistic to assume that AI will equally support niche policing support activities ranging from ballistic analysis to the identification of fake imagery, documentation or voice. The police will undoubtedly need to get involved in supporting citizens’ efforts to recover stolen identities.
However, a Cassandra-like word of caution to think about seems appropriate. The significant deployment of AI to support modern policing is not free of risk, both in terms of intended or unintended malicious consequences and societal ethics. Critics of certain forms of policing in China, much of which is associated with a range of AI applications, claim that it negates any boundaries between protective surveillance and control. Individual privacy is an early casualty. Critics nearer to home and more likely to focus on issues of transparency regarding the applications of new technologies, particularly as bias seems to be a constant source of friction for those impacted by such technologies or software. Unfortunately, software designers might be attracted by the lure of having intelligent machines develop algorithms in order to eradicate systemic bias but this is equally problematic. Can we be sure we understand why the machine thinks as it does?The future of policing will be the result of the evolution of technology, not an AI-inspired ‘Big Bang’. What we cannot say for sure is how it might evolve and whether or not we will be satisfied with the result.