Chinese researchers have equipped a bicycle with a revolutionary chip that runs parallel algorithms of a completely different nature. As a result, a bike with amazing adaptability, able to respond to voice instructions or follow a person.
In 2016, Google unveiled a YouTube video in which we could see a bike roam the streets of Amsterdam by itself, get up when it falls, or take children on a ride. ” We adapted our know-how from the autonomous car to create this bike and we even added new features, ” said a Dutch technician from Google. Becoming viral, the video had attracted more than 6 million visitors in a few days. Except that this famous bike never existed: it was an April fool of the company.
An electric bike with sensors and cameras
The smart bike developed by Shi Luping and his colleagues at Tsinghua University in China is very real. If he does not bring the children back to school yet, he is able to detect surrounding objects, respond to vocal instructions, avoid obstacles and maintain his balance. It is equipped with a gyroscope, microphones, cameras, motors to control the wheels and the speed and a battery. But this bike is actually a demonstration intended to show the true feat of researchers: the chip it contains.
Two opposing approaches to artificial intelligence
For simplicity, experts in artificial intelligence are divided into two camps: most often, chips rely on virtual neural networks, which are in fact successive layers of computer processors calculating a numerical value from an incoming data. These neural networks get excellent results on specific tasks when they have a large amount of incoming data, for example, to recognize images or translate texts. But they are limited when it comes to responding to complex and evolving problems.
The second type of chip, called neuromorphic, seeks to reproduce as faithfully as possible the functioning of a human neuron. In 2017, CNRS researchers have created an artificial synapse that can adjust its resistance under the action of electrical impulses similar to those of neurons.