The first time we talked about Roborace was last November, and fortunately we can say that the project is not only still standing but is full of news. On the one hand, in the last days of March the generic design of the cars was unveiled, and now comes the news that they will have Nvidia DRIVE PX 2 supercomputers at their core, with the capacity to process 24 trillion AI operations per second.
The current Formula E season is underway, and although participating teams have several technical details to resolve and optimize, the next championship will be special due to the debut of Roborace, a new category whose races will share the same circuits. The biggest aspect to highlight about Roborace is that it has no human drivers, and each car will be controlled autonomously by an artificial intelligence. The goal of the teams is to develop the best AI possible since all cars will be identical, but this requires the intervention of very advanced hardware, and that's when the people at Nvidia come in. CEO Jen-Hsun Huang announced last Tuesday during the GPU Technology conference in Silicon Valley that Nvidia will be the exclusive provider of DRIVE PX 2 supercomputers in the category.
The DRIVE PX 2 was officially presented last January. Its specifications tell us of two Tegra central processors accompanied by two discrete video units, which in combination can process a maximum of 24 trillion operations per second, focused on deep learning tasks. According to Nvidia, this processing power is equivalent to 150 MacBook Pro computers, and allows the DRIVE PX 2 to analyze in real time feeds from twelve cameras simultaneously, in addition to radar, LIDAR systems, and ultrasonic sensors.
The rest of the work will be up to each engineer, each developer, and each technician behind the cars, detecting and correcting the grossest errors before, during, and after each race, and if the DRIVE PX 2 performs as Nvidia says, we have to admit that the category will not lack "muscle", so to speak. The big question is, how will this affect the show in general? The human factor is (by far) what makes a race entertaining. Can artificial intelligences take that position? We will find out in a few months.
