Tesla Dojo is already training neural networks to achieve fully autonomous driving. According to the announcement, production of the supercomputer will begin in July.<\/p>\n\n\n\n
Making cars autonomous requires a lot of work. While other companies are programming their cars to follow a given route, Tesla is aiming to make its Autopilot really smart and able to drive even in unfamiliar terrain.<\/p>\n\n\n\n
Cars manufactured by the company process data on a large scale using chips of their own production. Tesla vehicles need a powerful computer on board. The company had to develop its own chip capable of processing hundreds of neural tasks and thousands of images per second. In fact, the chips must work in real-time, processing extremely complex information.<\/p>\n\n\n\n
After an initial announcement in 2020, Tesla unveiled its Dojo supercomputing technology in 2021. The purpose of the Dojo is to increase the speed and accuracy of training. It is a neural network (NN) chip to increase the learning rate of NNs on the server side. The faster a company can train NN, the faster it can deliver autonomous driving. The AI will be trained using videos that the company’s vehicles are assembled while they drive and fed back to Tesla.<\/p>\n\n\n\n
Until recently, the public was just waiting for updates on Tesla’s new supercomputer. However, on Wednesday the company created a new Twitter account dedicated entirely to the manufacturer’s AI developments. Tesla shared a graph that contained interesting information.<\/p>\n\n\n\n