On the 10 February 2016 the US National Highway Traffic Safety Administration agreed to classify Google's artificial intelligence system as the legal ‘driver’ of its automated vehicles (AVs). With this regulatory recognition, artificial intelligence and navigation systems’ future role on our streets seems assured. But I believe we don’t yet understand either how to set or govern the rules of this new driving era.
To date, safety has been the main concern. How will the AVs interact with pedestrians, cyclists and human driven vehicles? Most reported incidents involving a Google AV and another vehicle have so far been rear-end shunts, where a human driven car has driven into the back of an AV. Gauging other vehicles’ intentions is hard, and the AVs caution is understandable, but as any experienced driver will tell you, a dithery, traffic-shy driver doesn’t necessarily make for safer roads.
On any typical street there are hundreds of simultaneous intentions and interactions taking place at once. Adding AVs further complicates this situation. For example, how will pedestrians’ and cyclists’ behaviour change when they realise that stepping out in front of an AV will cause it to stop? And if AVs always play it safe, does this give human drivers an opportunity to take advantage of them? Worse, who would buy an automated vehicle that always loses out to person-driven cars?
Perhaps AVs will fare better on AV-only roads or in specific areas of a city centre. But even assuming this more constrained role for AVs we still have to decide how they will interact with each other.
A touted advantage of AVs is that they can travel tightly packed together, increasing road capacity and maximising passenger journeys at predictable speeds. But if an AV needs to change lanes and cannot find a gap in the traffic it will wait patiently. What if it takes ages for a gap to appear? Who decides how long it can hold up traffic before this becomes intolerable to passengers in other AVs?
Stationary AVs, trapped in their lanes, could actually cause traffic to worsen. Add in the slightly unpredictable behaviour of human passengers waiting to be picked up from the side of the road and you can imagine scenarios where vehicle stand-offs could last indefinitely.
The industry is trying to address these issues with vehicle to vehicle communication (V2V) systems and by building shared rules of engagement into the artificial intelligence of each AV. But who sets the priorities in the above scenarios? Will the rules be regulated by the state so that every brand’s AVs behave the same way? Or will different manufacturers be allowed to apply different behaviour parameters to their vehicles, perhaps meaning that a Google car will always give way to an Audi AV?
If we allow the commercial logic, or the price mechanism to intrude into the rules for AV artificial intelligence we run the risk of inventing a new level of inequality, where only the rich arrive on time. I believe a public debate is needed before the driver-less technology becomes widespread on our streets.