• processes 24 trillion operations per second,
• has 8 teraflops of computing power.
These are 10 times faster than its predecessor, Drive PX (presented a year before) that is used already by more than 50 car manufacturers globally. What is the most significant in this astonishing thing is that it, having the powers of 150 MacBook Pros, fits in the small-sized box – far better than the current computing capabilities that occupy the full trunk of the sedan car to make the same what this small cutie already does.
At the demonstration, Jen-Hsun Huang had the screen split into two parts: at one, there was a video of what a human driver sees during the test driving, and on the other part, it was the same picture ‘seen’ by a computer through the use of LIDARs, sensors, cameras, and scanners. The amazing capability of distinguishing the obstacles, bypassing them, and planning the best route was not the only magnificence in the process, coming close to human actions. The salt was in the self-learning technology that allows making the self-educational process much faster with Drive PX 2 computing chip than any other existing technology could provide. Thus, Audi’s representative told that with this chip, it took him to complete a training program in 4 hours that he did 2 years with the other competing technology! It enhances the self-learning process in 30, 40 or even more times.
NVIDIA was going to give this technology to Volvo first but they landed the second – Tesla turned out to be the first company that equipped own new vehicles with these chips from October 2016.
The deep learning
The chip Drive PX 2 will be linked to the neuro-link located somewhere in the computer clouds that will be available to self-learn. Before representing Drive PX 2, there were many issues connected to understanding the real road situations like teams of construction workers on the road, fast-changing environment like closed lanes, fallen trees, heavy rain, bulges of the broken asphalt due to the rocks falling down from the mountains and so on. With the introduction of this technology, the rapid self-learning has become available. All the data collected from every car connected to the cloud computing system will be used by its super-brain to educate oneself on every possible road situation and to find the optimal solution. When it has it, it gets back to the car’s computer brain and a car acts further taking into account all newly-processed situations. Thanks to capturing the data from tens of thousands of vehicles and processing millions of miles of real situations, the cloud system will be able to self-educate fully within the limited quantity of years. That seemed totally impossible only some years or even months ago before this technology occurred.
With the help of this advancement, the driving process will become much safer, and it will be able to prevent all the deaths on roads (or to decrease their level way too dramatically). After texting being on the steering wheel, drunk driving is the most killing reason among the teenagers, so at least these could be prevented. As of the time, 93% of all road accidents are human-induced. What if they will be totally zeroed? It has become possible.
The computers now prevail in many routine tasks over people. For example, they recognize 96% of all the pictures correctly – which is even better than the level of a medium person engaged in the same task. The same they win in games that only human brains could solve just recently. Some may even have the far-going conclusions – that people may become obsolete. We don’t think it will be the case. For now. At least not before 2020 when everyone will hit the road with fully driverless technology and big ‘cloud brains’ educating just in seconds.