CEO Elon Musk and other Tesla executives told investors on Monday that they would be driving more than a million self-driving Teslas on the roads next year.
Optimistic rhetoric contrasts with the rest of the autonomous vehicle Industry is increasingly warning how long and hard a road will be to get such cars on the road. Tesla has already made progress with its autopilot software, which has been used in some vehicles since 2015 and handles smaller driving tasks. Tesla sees the autopilot as a step-by-step move towards fully self-driving vehicles. The software can be updated with an over-the-air update.
Tesla is also characterized by the approach of a safe self-propelled car.
The company does not rely on high-resolution maps to run its vehicles. There are no geofterierten borders, which limit the possibilities for the use of the cars and it is used no sensor, which consider almost all other as indispensable. The sensor, called LIDAR, is known for its ability to tell a vehicle exactly how far nearby objects are, such as a pedestrian, car, or intersection.
Instead, Tesla will rely on cameras, radar and ultrasound sensors to understand the environment of a car.
"LIDAR is a joke thing, and anyone who relies on LIDAR is doomed to fail," said Musk, who described LIDAR as costly and unnecessary.
Typically, LIDAR bulges off the roof of a self-driving car, reminiscent of a spinning bucket of fried chicken. Early LIDAR sensors cost $ 75,000, but costs are falling as industry evolves and sensors are mass produced.
The approach of Tesla is controversial and has criticized the competition in public.
Tesla argues that when people can drive with two eyes, they can drive machines with cameras.
Andrej Karpathy Tesla, senior director of artificial intelligence, described LIDAR Monday as an abbreviation that gives motorists a false sense of progress.
"Is this person distracted and on her phone? Are you following her trail?" Said Karpathy. "These answers can only be found in the vision."
He said Tesla is training an artificial intelligence system to understand what is happening on the streets, using camera images from the Tesla fleet. He described how Tesla had taken a picture of a bicycle mounted on the back of a vehicle and used it to look for similar images in the Tesla fleet. Then Tesla found a series of photos of bicycles on the back of cars and used them to train his AI to distinguish between a bicycle driven on a street and a bicycle attached to the back of a car. Understanding such differences is crucial before cars can drive without human drivers.
Experts say the technology is up to date.
"This is remarkable, technologically it's unbelievable," said Bryan Reimer, a researcher at MIT AgeLab and deputy director of the New England University Transportation Center at MIT. "I do not think anyone else in the industry is capable of doing so in near real-time."
Reimer also warned of possible restrictions that slow Tesla's success.
"Everyone else is going the other way," said Reimer. "Eventually, we'll know if Musk is right or wrong, but it could be decades."