Elon Musk Says Tesla Developing Neural Network Training Computer for Full Self-Driving

With a complete overhaul of the Tesla autopilot software, full self-driving capabilities will be possible as the AI will be able to interpret traffic signals, turns and stops.

The concept of self-driving cars has existed at least since the 1920s but the vehicles haven't really fulfilled or performed on par with the promised potential. Combine that with rules and regulations, a complete autopilot mode has remained a distant dream. But Tesla founder Elon Musk says the company is developing a Neural Network training computer named 'Dojo' to completely overhaul how self-driving is now perceived.

This has come after the Tesla Chief Executive Officer revealed in Q2 2020 Earnings that the company was working on advancing the already established artificial intelligence (AI) setup that would improve the autopilot or full-self driving capabilities of the electric car. He said in his tweet that the overhaul was required to "process truly vast amounts of video data." At present, some Tesla models have semi-automation that requires driver's intervention.

Elon Musk
Elon Musk expects to launch fully self-driving vehicles by the end of 2020 Wikimedia Commons

Rewriting AI Autopilot Software

Following his tweet, a follower, Tim Dodd, who also runs a YouTube channel Everyday Astronaut, asked Tesla's Chief Executive Officer if there would be any change in autopilot, Musk answered in affirmative. He added that he was already driving an alpha build of the car from home to work without any intervention.

"The FSD improvement will come as a quantum leap because it's a fundamental architectural rewrite, not an incremental tweak. I drive the bleeding edge alpha build in my car personally. Almost at zero interventions between home & work," Musk tweeted on Friday, August 14, adding that a limited public release was scheduled in six to 10 weeks.

Twitter post

Robotaxi Fleet

Last year, Musk also announced that the company had plans to deploy a million robotaxis or self-driving vehicles by end of 2020. With the development of the AI, that may now be possible if not in 2020 then in 2021. The development also includes a complete rewriting of Tesla's autopilot software. It will help the AI better understand and process images. The AI will also make a transition from 2D to 4D to interpret objects well that include traffic lights, turns, stops and sudden changes in acceleration. Musk earlier said that Tesla was close to Level 5 FSD cars.

"The actual major milestone that's happening right now is really a transition of the autonomy system of the cars, from thinking about things in — like two-and-a-half feet. It's like think — things like isolated pictures and doing image recognition on pictures that are harshly correlated in time but not very well and transitioning them to kind of a 4D, which is video essentially," said Musk, who is confident of launching fully automated cars by the end of the year (2020).

Tesla cars
Tesla cars fall in Level II of autonomous driving with partial self-driving capabilities when FSD is engaged Wikimedia Commons

If Tesla manages that in the software rewrite, Dojo can be deployed in automated taxis, provided it receives regulatory clearance.

Related topics : Elon musk Artificial intelligence