Little by little, Tesla continues to improve and build on the advances the company has already made in electric vehicles, but their latest acquisition of DeepScale, a Silicon Valley start-up that uses low-wattage processors to power more accurate computer vision, will hopefully aid them in making leaps and bounds towards turning their already high-tech vehicles into robotaxis.
At this point in their technological advances, Tesla vehicles are not yet considered to be fully autonomous, or Level 4 as the SAE would designate. Level 4 cars are able to handle every aspect of driving under particular conditions without the need for a human.
But Tesla vehicles have only made it to level 2 so far, which means that the autopilot feature is more advanced than most other vehicles.
CEO Elon Musk’s vision is that, eventually, all of Tesla’s vehicles will have such advanced driver assistance capabilities that they will be fully automated. He also said earlier this year that Tesla would be launching an autonomous ridesharing network by 2020.
This most recent acquisition of DeepScale, reported first by CNBC but confirmed by TechCrunch with two unnamed sources, is just a stepping stone towards this goal being realized. Tesla has gone through quite a few departures this year, so DeepScale should bring the talent and creativity that Tesla needs.
DeepScale has discovered a way to improved perception system accuracy through efficient deep neural networks on small, low-cost, automotive-grade sensors and processors.
Utilizing sensors, mapping, planning and control systems to interpret and classify data in real time, these perception systems are necessary to the operation of autonomous vehicles, helping them to understand the world around them.
Since Tesla uses low-wattage and low-cost sensors and processors, the company can provide vehicles with driver assistance and autonomous driving at any price point.
In $3 million seed and $15.6 million Series A rounds, they have already raised over $18 million from investors such as Autotech VC, Bessemer, Greylock, and Trucks VC.
DeepScale’s co-founder Forrest Landola updated his LinkedIn account and posted an announcement on Twitter, saying, “I joined the @Tesla #Autopilot team this week. I am looking forward to working with some of the brightest minds in #deeplearning and #autonomousdriving.”
Tesla has developed a custom chip that’s now in all new Model 3, X, and S vehicles; it is designed towards full self-driving capabilities, and according to Musk, Tesla vehicles already possess the essential hardware to make this possible.
“All you need to do is improve the software,” Musk said at Tesla’s Autonomy Day in April.
Musk has held onto this claim despite the cynicism and doubt of others in the industry, and the software has steadily improved.
In fact, just earlier this month, Tesla released a software update that adds new features to its cars, including Smart Summon, a parking feature that enables owners to use their Tesla app to summon a car from its parking spot.