Driverless Cars Face Roadblocks as AI Struggles with Real-World Nuances
The rapid advancement of autonomous vehicle technology is facing a critical juncture, with recent incidents highlighting significant challenges in the artificial intelligence’s ability to accurately interpret complex, real-world scenarios. As trials of self-driving taxis are set to launch in London this April, and testing has already been underway in various British locales including York, concerns surrounding the safety and reliability of these computer-controlled vehicles remain a prominent topic of discussion.
Professor John McDermid, a software expert and advisor to the government on self-driving vehicles, based at the University of York, recently shared insights into the persistent difficulties faced by these advanced systems. He explained that even sophisticated AI can be easily confused, leading to potentially hazardous situations.
One particularly striking example involved a driverless car mistaking a life-size advertisement on the side of a bus for a group of actual pedestrians. The advertisement, for the film The Man from U.N.C.L.E., featured several actors depicted at their actual size. To the AI, however, these static images were interpreted as living individuals in the roadway. This misinterpretation triggered an emergency stop by the vehicle, a manoeuvre that could pose a significant risk to traffic following closely behind.
Professor McDermid elaborated on this incident, stating, “One of the automated vehicle companies I work with had a situation where their vehicle did a sudden emergency stop because it’s all pedestrians in the road, except they weren’t. It was a life-size advert on the side of a bus, but to an AI, it was human beings. That seems very obvious [to us], but actually, to the AI, it’s not.”
Navigating Unpredictable Pedestrian Behaviour
Beyond static advertisements, driverless cars in trials conducted in York have also demonstrated confusion when encountering the often unpredictable movements of pedestrians. This includes instances where individuals may attempt to cross the road even when the traffic lights have turned green for vehicles, or when the “no crossing” signal is still illuminated.
In contrast, the United States, a pioneer in self-drive car development, has strict regulations against “jaywalking,” with potentially severe penalties for pedestrians who cross roads unlawfully. This legal framework may encourage pedestrians in American cities to adhere more closely to traffic signals. However, in the UK, pedestrians often hold a perceived primacy on the roads, and the new generation of autonomous vehicles appear to be struggling to fully grasp this cultural norm.
Professor McDermid described the confusion observed in York: “It’s seen that there’s a traffic light, so identified the hazard, because the light is red. It changes to green, the vehicle is about to move off. But this is York, so the tourists – although the lights change to green – still walk across the road. Computer vision doesn’t understand what it doesn’t have models for in the world. It doesn’t know what a roundabout is.” This highlights a fundamental limitation: AI’s reliance on pre-programmed models and its inability to intuitively understand novel or culturally specific behaviours.
International Concerns and Public Apprehension
The challenges are not confined to the UK. American self-driving car company Waymo is slated to commence its driverless taxi trials in London from Easter. The company has ambitious plans, including integration with Uber to eventually offer robotaxi services to the general public. However, Waymo’s own operations in San Francisco have previously raised concerns. Approximately two years ago, school crossing guards, often referred to as “lollipop ladies,” reported numerous near-miss incidents involving Waymo’s autonomous vehicles.
A survey of 30 such crossing attendants revealed that roughly a quarter had experienced a “close call” with an autonomous vehicle, with some even having to dart out of the way to avoid a collision. Theresa Dorn, a veteran crossing guard, recounted three near-misses with driverless cars in a single year. In one alarming incident, a parent had to swiftly intervene to save a child from a potential accident. Dorn expressed her bewilderment, stating, “The parent grabbed the child, looked at the car – and there was nobody driving it. Why do they have these driverless cars? I think somebody should be driving them.”
Regulatory Standards and Public Safety
In the United Kingdom, current government guidance mandates that “self-driving vehicles should be held to the same high standard of behaviour as that expected of human drivers.” However, a significant portion of the public surveyed indicates a desire for even higher standards, reflecting a palpable fear of an increase in road fatalities. Currently, around 1,600 people lose their lives on UK roads annually.
Professor McDermid has voiced a strong caution against allowing pedestrians to become a “moral crumple zone” for the burgeoning robot car industry. This statement underscores the ethical imperative to ensure that the introduction of autonomous vehicles does not inadvertently place vulnerable road users at greater risk, necessitating a robust and proactive approach to safety and regulation.
