Elon Musk is convinced that within a year the electric cars produced by his company Tesla will basically be able to drive themselves, without the people on board having to worry about the road. Musk, who is known for making effective announcements and indicating deadlines that he then fails to meet, explained this during a sort of seminar organized at Tesla's headquarters in Palo Alto, California on April 22, to expose the new plans. of the society on autonomous driving, one of the biggest challenges for auto manufacturers along with the shift to electric vehicles.
During the event, Tesla announced that it has completed the construction of a new microchip, specially designed to handle the large number of calculations and operations that self-driving cars have to carry out. The microchip has already been installed for a few weeks on the new cars produced by Tesla to replace the previous processor, which was made by the company Nvidia (one of the largest and most famous manufacturers of computer components in the world).
The new microchip, Musk said, is up to 21 times more powerful than the one supplied by Nvidia and installed on Tesla's up until a few weeks ago. It was designed by Tesla, it is materially produced by Samsung and has the task of analyzing all the data coming from the various sensors of the car, which detect the surrounding environment and other information to allow the on-board computer to drive the car without a direct intervention by those behind the wheel.
The chip uses an artificial intelligence system that mimics the functioning of the brain (“neural network”), with the ability to perform 36 trillion operations per second. Two identical neural networks are active on the processor that do the same thing, and Tesla says he has a spare microchip installed just in case the first one stops working properly: since the microchip is driving the car, it does not seem such a stupid idea.
Musk said the new microchip is an intermediate step and that Tesla is already working on a new version, which will be three times more powerful. No date has been given for when it will be ready, but the company says its development is about halfway through.
Tesla isn't the only company working on self-driving cars. Alphabet, the holding company that controls Google, collaborates with several auto companies through its subsidiary Waymo. Ford, one of the largest auto manufacturers in the United States, has its own division that is focusing on self-driving systems. In addition, several other Silicon Valley companies are working on perfecting these systems. Among the largest and most important is Uber, which has already started experimenting with self-driving taxis in some US cities.
Musk hinted that Tesla aims to compete with Uber by offering a service to request a self-driving taxi via an app, tell the system the destination and be guided to the destination. The first trials will begin in 2020 and will involve owners of a Tesla. Those who wish can make their electric car available to others, receiving compensation for the ride offered. A car you own spends much of its time parked, so the idea is to take advantage of that downtime, making it profitable for the vehicle owner.
According to Musk's calculations, a Tesla put into taxi service could bring in up to $ 30,000 a year to its owner. In this way, every customer of the company could quickly get back the expense of at least 35 thousand dollars to buy a Model 3, the cheapest model available today.
Tesla Vision uses path prediction to accurately predict how a road will extend, even when it can't see around the corner pic.twitter.com/09qPkpqwSC
– Tesla (@Tesla) April 23, 2019
Musk also said Tesla will shortly offer “level 5” autonomous driving systems. To avoid contradictions and misunderstandings between car manufacturers and customers, an international standard was defined in 2014 that identifies 6 different levels for automatic driving. They range from level 0, where the car has no autonomy, to level 5, where the car is completely autonomous and is able to handle every situation as (if not better than) a human being, without the need for duty. never intervene at the controls.
Experts and analysts have received with skepticism Musk's announcement that he will reach level 5 in such a short time frame with his Tesla. Currently, the most advanced systems for autonomous driving are between levels 3 and 4, with various technical problems that have not yet made it possible to achieve full automation of the vehicle. It is therefore not clear how the Teslas can achieve it so quickly, even if they have shown to make enormous progress in recent years.
The most advanced systems for autonomous driving involve the use of a LIDAR, a system that, by throwing a beam of light (not in the visible spectrum) in every direction, and capturing how it comes back, manages to create a very accurate three-dimensional map. of everything the car has around it. This real-time 3D map is compared with other 3D maps built earlier in the same spot, in the world to identify the differences and distinguish the objects that are always there (a traffic light, a wall, a tree) from the things that change their position. (pedestrians, bicycles, other cars). The detection is very accurate and allows the car to manage every movement, avoiding obstacles and following a path with the help of the GPS.
This 3D reconstruction shows the immense amount of depth information a Tesla can collect from just a few seconds of video from the vehicle's 8 cameras pic.twitter.com/w2x6pkM2Eb
– Tesla (@Tesla) April 23, 2019
Tesla does not use LIDAR, but a series of cameras and sensors placed in various points of the car. The images are then sent to the processor which, through artificial intelligence systems, decides how to move the car and how to avoid fixed and moving obstacles. Manufacturers using LIDAR say Tesla's solution is less secure and reliable, while Tesla argues otherwise, noting that its cars' cameras and sensors record and share a huge amount of data, which the same company can then use to perfect the algorithms that make autopilot work.
It's hard to say if Tesla's system is more reliable than LIDAR, it sure has enabled the company to progress faster and at lower cost, albeit in ethically questionable ways. The experimentation of automatic driving systems in recent years has in fact occurred by exploiting the same owners of the Tesla, which according to the most critics have acted as guinea pigs. Autopilot has almost always proved to be efficient and in many cases it has made it possible to avoid accidents, which would hardly have been avoided by a human being. But, in a limited number of cases, autopilot was instead the cause of the accidents, raising further concerns about Tesla's approach.
Tesla currently sells an option for its on-board system called “Full Self Driving” (FSD) at a price of $ 5,000 if purchased at the same time as the car, otherwise as an extra for $ 7,000 if purchased later. . The system is currently limited, as Tesla says it is awaiting certifications and authorizations to make its cars fully autonomous. According to Musk, once specific laws are passed for these types of vehicles, Tesla will just need to send an update to the on-board computers of its cars to make them completely autonomous.
In reality, no specific laws should be needed to activate the system, even if in the United States, as in many other countries, autonomous driving continues to be a poorly regulated gray area. There is a common belief that it takes at least a decade before fully automated driving is achieved, so it is unclear how Tesla can achieve the same result in such a short space of time. Musk is known for making very optimistic statements about the timing to realize his visions, hence the numerous speculations around Tesla's actions in recent years and doubts about the economic viability of his projects.
