Driverless taxis, which have disrupted industries in various US and Chinese cities, are now on their way to London.

As a cyclist, Londoner, and journalist who has closely observed AI developments, I find myself somewhat anxious. Yet, considering the frequent encounters with careless human drivers in London, part of me feels cautiously hopeful.

Ultimately, the question arises: is it preferable to navigate the roads among tired, distracted, and irate humans, or to coexist with potentially erratic AI?

The UK government has affirmed plans for companies like Uber to launch pilot programs featuring self-driving “taxi and bus-like” services in 2026. Following that, in the latter half of 2027, automated vehicle legislation is expected to take effect, establishing a formal legal framework for the industry. Crucially, this law places accountability for accidents on the automakers rather than local residents.

Officials advocate that unmanned vehicles could enhance road safety, given that human error is responsible for 88% of all traffic accidents. The statistics are staggering: London highways reported 130 fatalities last year, which included 66 pedestrians and 10 cyclists. Globally, 1.2 million individuals die in traffic incidents annually.

As someone who cycles often in London, I have firsthand experience with the challenges posed by reckless driving. I’ve witnessed drivers engaging in a variety of distractions, from eating breakfast to watching movies. I have been rear-ended at red lights at least four times. While it is commonly said that AI lacks creativity, humans have certainly mastered the art of poor driving.

In contrast, AI isn’t swayed by distractions such as text messages, alcohol, or fatigue. With numerous sensors, machines lack blind spots and always check their surroundings before making a turn.

Admittedly, there have been alarming reports of autonomous vehicles failing to stop and causing harm to pedestrians. These incidents garner significant media attention. However, considering the numerous fatalities attributed to human drivers, the statistics of road deaths paint a less sensational picture. In the UK, more than four people die daily in traffic accidents.

The safety concerns surrounding autonomous vehicles are complex. While I believe that every road fatality is unacceptable, there exists a compelling argument that if AI can travel the same distance with fewer casualties, it shouldn’t be demonized in the pursuit of progress.

Research indicates that driverless cars often outperform human-driven vehicles in terms of safety, although this advantage may not consistently hold in urban environments, particularly under poor lighting or during complex maneuvers.

These vehicles depend on technology companies to ensure their safety, raising questions about potential conflicts of interest relating to profit versus safety. We have already seen concerning suggestions to equip pedestrians with electronic sensors to enhance their visibility to these machines.

When it comes to cyclists, can tech companies ensure they maintain a 1.5-meter buffer when a robocar passes, or will they simply prioritize not hitting cyclists? The latter might streamline urban travel times, but could pose risks to vulnerable cyclists. Furthermore, to what extent will autonomous vehicles pause to allow pedestrians to fully cross the street, or will they encourage hurried crossings? These parameters can be influenced, and there are inevitably tensions between safety and travel efficiency.

Even if a company aims to act benevolently, AI systems are inherently unpredictable. Just as chatbots can suggest erroneous ingredients, self-driving cars cannot guarantee they won’t misinterpret a pedestrian as a shadow. It’s an unsettling truth.

Personally, I harbor reservations about AI operating vehicles in my vicinity, just as I do about human drivers. However, while human capabilities can improve with time and effort, AI has the potential for rapid advancement. The roll-out of automated taxis in London could provide invaluable data that enhances the safety of our roads. Ultimately, if given a choice, I would prefer an AI driver.

Nevertheless, the stark reality remains: a few tons of steel on four wheels—combined with high-tech systems—will never constitute a wholly safe or efficient urban transport solution. Self-driving taxis may mirror today’s human-operated models, ultimately not resolving London’s transport challenges.

Electric bikes and dedicated cycle lanes are environmentally friendly and often more efficient for city travel, while buses can accommodate multiple passengers, utilizing the space of two SUVs. However, such solutions may not yield substantial profits for big tech companies, will they?

Matt Week

What I’m reading

How music works by Talking Heads frontman David Byrne.

What I’m seeing

Horror movies I’ll bring her back (It’s true, through the hands that cover their eyes at the moment).

What I’m working on

Next spring, I plan to plant various cuttings in my garden to fill empty spaces.

Matt Sparkes is a technology reporter for New Scientist