PERU´S CRAZY DRIVERS OFFER DATA DELUGE FOR SELF-DRIVING CARS / THE ECONOMIST

The lawless roads

Peru’s crazy drivers offer a data deluge for self-driving cars

A startup sees value in the chaos of the country’s streets Street with auto rickshaws and vehicles in Ayacucho, Peru.

Tuk-tuks with precarious cargo zoom down highways. 

Street vendors, angry protesters and careless llamas weave through vehicles. 

Buses literally overflow with passengers. 

Welcome to Peru, home to some of the world’s most chaotic streets. 

According to Compare the Market, a car insurance aggregator, Peruvian drivers are the world’s worst after Thailand’s. 

The ranking considers factors such as deaths caused by car accidents and time wasted in traffic. 

Typical road journeys in the capital, Lima, take nearly twice as long as they need to as a result of congestion. 

Police prefer asking for bribes to enforcing rules.

One startup sees opportunity in the chaos. 

“It’s a hidden goldmine of data,” says Arturo Deza of Artificio Inc, which is creating a dataset from Peru’s shoddy roads to train self-driving cars. 

Many tech firms train autonomous cars in Europe and the United States. 

But conditions there “are too clean and pristine”, says Mr Deza, a former researcher at MIT and Harvard University. 

Such cars still struggle to identify traffic cones, and would fare hopelessly amid roving livestock. 

Artificio plans to collect 10m hours of driving data from Peru and other Latin American countries by getting transport companies to install cameras on their fleets of vehicles. 

It would then license its dataset to tech firms.

Other startups also see value in the unruly roads of emerging-market countries. 

Swaayatt Robots and Minus Zero, two Indian tech firms, have tested their algorithms on the streets of Bengaluru, Bhopal and Jalandhar. 

Minus Zero recently partnered with Ashok Leyland, one of India’s largest commercial vehicle-makers, to develop self-driving lorries. 

In June Swaayatt Robots said that it had secured $4m as part of a larger fund-raising round.

Far more money and data are needed to scale up such projects. 

Artificio, for example, has gathered a scattering of videos from three Peruvian cities in a pilot phase. 

Even 10m hours of driving data is puny compared with what companies like Tesla own. 

But if the likes of Tesla want to sell in developing countries, they may need to shift gear.