Robocop might not be a reality yet – but governments are using artificial intelligence in everything from law enforcement to medicine, says Joshua Chambers
Arnold Schwarzenegger became an elected politician, but could his alter ego – the Terminator – one day become cabinet secretary?
Artificial intelligence is already showing great potential in government. From policy making to service delivery, countries around the world are experimenting with new technologies, showing what could be possible in the next 30 years.
Let’s start with the smaller schemes. Nestled in the Himalayan Mountains is the Kingdom of Bhutan, with a population of 750,000 people scattered amongst remote settlements on steep hills and deep valleys. Poor roads and bridges make some of the country completely inaccessible to vehicles, meaning that citizens can often go without receiving basic medicine supplies.
A San Francisco startup, Matternet, has partnered with the government to test out unmanned drones for delivering drugs over the mountains. The quadcopters can also be used to deliver samples back to regional hospitals for tests. It is intended that these devices will soon be able to fly themselves, “replacing human labour, and reducing cost and errors in the system altogether,” Matternet’s chief strategy officer, Paola Santana, told me recently.
What about in law enforcement: could Robocop become a reality? Western Australia’s police force has tested a smart police car that automatically scans all the vehicles around it, reporting suspicious details and wanted license plates straight back to the main control centre without driver input. It isn’t much of a leap to see these cars eventually doing this without any human drivers; after all, the state’s lucrative mining industry is increasingly turning to driverless dumper trucks.
Late last year, Microsoft even showed off robot security guards on its campus, with the fully autonomous devices using sensors and cameras to read number plates, monitor situations and take action if necessary. They do have a downside, mind: the egg-shaped androids are susceptible to being toppled over by pranksters.
More advances have been behind the scenes, where computers provide strategic advice to decision makers. Police forces across the world are testing predictive analytics, to ascertain whether someone’s previous criminal record and social media postings can show their likelihood of reoffending.
Aside from policing, some computer systems are now taking decisions that used to be made by public service managers. Hong Kong’s Mass Transit Railway has its maintenance schedule run by a system that oversees 10,000 staff working on 2,600 repair tasks each week. Previously, key tasks and priorities were determined by a meeting of department heads negotiating a weekly plan, but now the schedule is generated automatically.
This is small fry compared to what happens in Singapore. The government there adopted software that came from the Total Information Awareness Office of the United States’ Defence Advanced Research Project Agency (DARPA). Called the Risk Assessment and Horizon Scanning System, it is an enormously advanced policy making tool which picks up weak signals from masses of datasets.
One key dataset is geolocation data taken from mobile phones. Monitoring these movements generates a picture of citizen behaviour right across the city. Unlike monitoring more basic data, like travel card information, it can show a very detailed – anonymised – picture. Urban planners can see crowded areas, popular routes, lunch spots and more, showing where to build new schools, hospitals, cycle lanes and bus routes.
Feeding it consumer spending data and trade information can assist with economic planning. It can pick out key economic trends, warning signs, and give a detailed understanding of the current state of an economy. Indeed, the data it’s given can be anything from labour market flows to agricultural information.
The system is designed to copy humans’ cognitive processes, but without their common biases, such as linear thinking – our desire to think of things in stages. Another key human bias is called ‘hyperbolic discounting’, which makes us far more prepared to deal with a problem in front of us than pay a premium for a future event that may or may not happen.
The American creators of the system believed that it didn’t require any human intervention; that the computer should be fully autonomous. Singapore has decided that the computer must be directed by humans, who ask it questions, then interrogate the data. And this common thread runs through all the examples above: humans are monitoring the drones, sitting in the cars, approving the maintenance schedule and interacting with the planning system.
But the result is still a completely transformed method of public service delivery. Computational aides aren’t much like the Terminator – a fully autonomous decision-making machine – but more like C-3PO: an android which constantly points out the risks of a decision, but remains happy to defer to strong political leadership.