Help of humans is essential to teach autonomous cars to drive, says Arie Halpern

The battle to make autonomous cars go forward has had the attention of both automakers and technology companies. Several automobile companies have been investing heavily in this technology to bring to the market self-driven cars which do not need human intervention, as soon as possible. The product is expected to reach the market between 2020 and 2025. The investment in this segment per company varies. Just to have an idea, Ford put US$1 billion in its model, the Argo, and Intel has purchased the company Mobileye for US$ 15 billion. In addition to investing a great deal of money, companies are also putting a lot of effort on the search for development processes that can assure the functionality of the new technology. “The participation of ordinary people, in addition to the engineers and programmers, has shown to be essential on teaching the autonomous cars to drive, says Arie Halpern, economist, and entrepreneur with a focus on innovation and disruptive technologies.

According to a news article on the technology site Wired, a multitude of people has been developing artificial intelligence systems that will be responsible for making cars without drivers that will take us to our destinations safely. Might AI is one of the companies that are helping to make autonomous cars a safe reality. While the programmers supply the cars’ system with data about traffic rules, Mighty AI develops an application in which ordinary people are used in helping the cars to identify objects and beings through photographic images and differentiate a person from an animal, a baby stroller from a trash can, for example. This is an essential work for traffic safety, comments Arie Halpern. “The data are used to create a pattern that AI will be able to recognize and replicate over time”.

Another application is the Chffr whose operation is a little different. Rather than selecting photographic images as the Mighty AI does, just install the Chffr on your mobile and attach it to the windshield of the car. The application alone will collect the data needed while the motorist drives the vehicle. Meanwhile, the startup has decided to go a different path: the deep-learning. Arie Halpern explains that all these efforts are geared toward recognition situations that can seem simple for a human being, but not to a machine. “The car knows that it must stop for a pedestrian, but if this person is on a bicycle or with a different outfit, AI may not recognize it correctly,” he says. Therefore it is essential to have human help in the development of autonomous cars.

Another way to teach the vehicles to drive well is the video games, especially the Grand Theft Auto, which ironically is famous for the fact that its players have fun running over people and objects when driving around town. Scientists at Darmstadt University, Germany, and at Intel have developed methods to gather data from the video game and teach the vehicles to recognize different obstacles which may be found on the way. This is another creative way to collect data for the machines’ learning process, through the interaction with humans.

Deixe uma resposta

O seu endereço de e-mail não será publicado. Campos obrigatórios são marcados com *