Elon Musk: Self-driving is hard

Elon Musk fa mea culpa admits that self-driving is harder than he bargained for. Many Tesla users who were impatient with the poor progress of the full autopilot (the package can be purchased separately and costs 7,500 euros) began to mock the American businessman on Twitter. In response to someone who scoffed at the promise of the imminent arrival of the Fdf V9 Beta, Musk tweeted the promise of the long-awaited release, acknowledging that autonomous driving is harder than expected because it requires sophisticated artificial intelligence capable of communicating with the real world (hereafter tweeted).

Fsd V9 الانتظار Waiting Output

–

In recent days, Tesla has started releasing the much-awaited Beta 9 version for some Beta 9 users for fully autonomous driving. One of the major innovations in this version is the Tesla Vision, a new computerized vision system entirely based on cameras without the need for a radar sensor. Musk has a long history of failed predictions about “real” Tesla self-driving. Initially, the version was launched in 2018, then in 2019, and only recently Musk has changed the communication log which is no longer talking about self-driving, but rather a “full-featured” system that still depends on the driver’s attention, but can lead to real autonomy The data shows that it is safer than men’s driving. This complete system that Elon Musk is referring to is precisely the Full Autonomous Driving (Fsd) pilot program that started in October 2020 open to a limited number of users (about 3 thousand).

Tesla’s approach

–

Tesla’s approach to self-driving is very different from that of any other car company. The American company relies on a fleet of more than one million vehicles that generates a lot of data that is exploited by machine learning neural networks, which in turn generates a system pure vision. However, it is very difficult to predict the system optimization curve with these properties. Among other things, compared to the competition, Tesla relies only on cameras and, in fact, on computer vision. A completely different approach compared to Waymo which instead uses an array of devices consisting of different radars, lidars, and different sensors.