Why self-driving cars on African roads may remain a fairy-tale

The idea of a self-driving car, which first gained widespread public attention at GM's Futurama exhibit at the 1939 World’s Fair, where the automaker envisioned "abundant sunshine, fresh air [and] fine green parkways" upon which cars would drive themselves, has now transitioned from Hollywood fantasy to reality.

In San Jose, California, a start-up called Voyage is trialing a taxi service within a gated retirement village, using autonomous cars to ferry elderly passengers around, navigating other cars, pedestrians, golf carts, animals, roundabouts and many other obstacles.

Meanwhile Volvo, the giant Swedish automobile manufacturer, claims 100 real-world customers will be using its self-driving cars on public roads by end of this year - the world’s first large-scale autonomous drive project. 

But for drivers across majority of African countries, that day seems a long way off. There are two formidable obstacles: 

Dysfunctional road infrastructure

At the root of these challenges is the ramshackle road network. Majority of roads across the continent are in poor state: They lack clear markings; road signage is inconsistent or non-existent; and traffic lights are defective or have been pinched by petty thieves. In some instances, a well-marked, two-way tarred road can suddenly squeeze down into a narrow, winding and dusty track.

These conditions are not reflective of the industrious African populace, but rather symptomatic of a gamut of other issues, chiefly endemic corruption by public officials.

According to Tom Vanderbilt, the author of Traffic: Why We Drive the Way We Do (and What It Says About Us), “Traffic behavior is more or less directly related to levels of government corruption.” 

In the light of these challenges, it’s hard to imagine how a self-driving car will navigate the intricacies of Meskel Square, an intersection in Addis Ababa with no traffic lights, no roundabout, no traffic cops, and hundreds of drivers deftly evading collisions of all sorts.

Meskel Square, Addis Ababa, Ethiopia. Credit: https://addisababaonline.com

Granted, Africa is made up of 54 sovereign countries, but the issue of substandard infrastructure is deep and pervasive. According to the World Bank transport infrastructure benchmarking index, more than half (13) of the top 25 countries with poorest roads conditions are African.

Vividly marked roads, reliable street lights and up-to-date roads signs, among several factors, are an integral aspect of the autonomous vehicle. So much so that Elon Musk, the charismatic Tesla CEO , recently called the mundane issue of faded lane markings in USA roads “crazy,” complaining they confused his semi-autonomous cars. When I reflected on the state of roads in Harare, I thought Mr. Musk might have been whining. 

In other developed countries, according to Reuters, greater standardisation of road signs and markings makes it easier for robot cars to navigate.

In Africa, however, unless the governments redirect public funds towards infrastructure development, the hope of commissioning self-driving cars soon might be far-fetched.

The Machine Learning Problem

The second, and equally important challenge relates to machine learning, the disruptive technology that underlie the autonomous car. Manufactures are using machine learning algorithms to read data from sensors, radars, cameras and detailed maps to train and improve autonomous cars’ responsiveness to varied road and traffic conditions.

The effectiveness of an AI algorithm, however, is as good as the conditions and assumptions under which it was trained. Most of these experiments are being performed under controlled environments, such as gated communities or select states in western countries with ideal roads conditions.

These simulations are not reflective of the chaotic realities that characterise many African roads and cities. But even under ‘ideal’ conditions, mishaps still occur. In March 2017, Uber suspended its fleet after one of its self-driving in Arizona fell sideways after it was shoved by a turning Honda.

How then will an autonomous vehicle trained to handle traffic conditions in Silicon Valley, not Silicon Savannah, fair alongside the constantly honking, cursing, and aggressive matatus, as well as navigate Nairobi’s rush hour grinds, petty cops, and a multitude of other ordeals?

Nairobi matatus – Credit: https://www.monitor.co.ug

Algorithms have indeed become adept at mimicking tasks that require thinking, but tend to stumble when called to automate tasks humans perform without thinking – especially those that require instinct, emotional intelligence, or creativity.

In the light of this challenge, it’s difficult to predict how a self-driving car will behave when it approaches a red traffic light midnight in Honeydew, Johannesburg, an area notorious with car jackings. Most Johannesburg drivers intuitively avoid stopping at known blackspots, sometimes opting to pay fines than risk a hijack. I would do the same.

Hijack warning: Credit – Dailymail.co.uk

This is intuition at its core, the innate human ability to constantly evaluate millions of bits of information, detect subtle conflicts within patterns, and alert the mind when something feels awry.

But autonomous vehicles are programmed differently, they respond according to predefined patterns, not instinct.

This dilemma raises yet another important question: Will the robotic car flout traffic laws, speed through a Johannesburg red traffic light to protect the passenger, or screech to a halt, exposing its master to different kinds of harm?

Flawed autonomous vehicles will be dumped to Africa.

Manufacturing autonomous vehicles is a high stakes game, imperfect code can expose passengers and other drivers to sigfnicant harm. Early this year, a flawed algorithm forced Tesla S vehicles to zig-zag wildly across the road, cause unexpected swerves, fish-tails and a host of other miscalculations.

If history is anything to go by, I fear unscrupulous dealers will use Africa as a dumping ground for self-driving cars that fail to meet safety specifications in their own countries, exposing civilians to harm.

These appalling practices are not new. In 2016, Public Eye, a Swiss based NGO, issued a scathing report in which it lamented a rampant practice by some European car exporters, in which they mix clean fuel with dirty fuel streams and other toxic substances - such as the cancer-causing benzene - before exporting them to Africa. This wicked practice, performed to achieve fuel with “African specifications”, result in extensive air contamination, estimated to be 100 times worse than places like Europe.

In the absence of enforceable, tighter vehicle import regulations, I foresee these malevolent dealers extending the scope of their activities to ship vehicles with out-of-date, flawed or hackable software to Africa.

Looking forward

This is not a counsel of despair. Africa holds great promise. As Brahima Sangafowa Coulibaly of the Brookings institute wrote, the Africa rising narrative still holds true. Five of the world’s 10 fastest growing economies, are located on the continent; about two-third of the region’s economies are expanding faster than the global economy this year.

But unless the endemic corruption is weeded out, and regulations tightened, the prospect of fully autonomous cars ferrying passengers across African cities may remain a myth.

My friend, who runs a thriving driving school in Harare, should also relax – his little business may not be disrupted any time soon.

Introducing fully autonomous vehicles under current conditions may be beckoning for disaster - or, as Elon Musk worries about super intelligent systems, summoning a demon.