As Musk stated in the recent Tesla Autonomy Day, Tesla has a vision to create autonomous robotaxis that act as a competitor to current ride-share apps. From a hardware standpoint, Musk claims that any car with hardware 3.0 (i.e. every newly produced Tesla) will be ready for full self-driving, and in turn, able to be a robotaxi. The only thing left, however, is the software.
This is where Tesla’s most recent acquisition makes sense. As CNBC reported, Tesla has acquired AI startup Deepscale and has confirmed the deal with two unnamed sources, although no financial terms were revealed. Deepscale Co-Founder, Dr. Forrest Iandola took to twitter to confirm the acquisition, writing, “I joined the @Tesla #Autopilot team this week. I am looking forward to working with some of the brightest minds in #deeplearning and #autonomousdriving.”
I joined the @Tesla #Autopilot team this week. I am looking forward to working with some of the brightest minds in #deeplearning and #autonomousdriving.
— Forrest Iandola (@fiandola) October 1, 2019
Deepscale specializes in efficient Deep Neural Networks (DNN) that are used to improve the vehicle’s perception systems. The perception systems, which use sensors, mapping, planning, and control systems to interpret and classify data in real-time are tremendously important if any car wants to be able to perceive the world around it and have any chance of being fully autonomous. Additionally, the network can use low-wattage and low-cost sensors and processors that allow the system to work on vehicles of all price points and not demand high-powered computers for it to work.
Here is an example of Deepscale’s object detection.
The startup has received more than $18 million in funding from Steve Cohen’s venture fund Point72, and Siemens-backed venture fund next4, Autotech VC, Bessemer, Greylock , and Trucks VC.
While the DNN working on low-watt hardware might not be too much of a problem given Tesla’s newest chip, Tesla is in dire need of talent as about 10% of the Autopilot team left over the summer. For full self-driving and robotaxis to become a reality, Teslas must reach an autonomy level of at least 4 which means that it can handle driving conditions in almost all areas and doesn’t need driver input if there is some sort of surprise event. Currently, they are at level 2, which means that the driver must not only be ready to be able to take over during a dynamic event but must also always be monitoring the cars speed/braking and steering.
We’ve had our first real taste of Tesla’s progress in terms of full self-driving with Smart Summon in the version 10 software update. Results seem to have been mixed thus far, but hopefully, Tesla’s acquisition does them some good in regards to making full self-driving a reality.
What do you guys think of the acquisition? Will Deepscale make a meaningful impact? Will robotaxis ever become a thing? Let us know down in the comments below.