Image credit: Sergio Monti Photography / Shutterstock.com
Renowned for its high-tech future orientated innovation and accelerating the world towards a more sustainable use of energy, automotive and energy storage company Tesla has made another step towards realizing their vision of fully automated vehicles on the road.
The company has recently acquired DeepScale, a silicon valley start-up that has been enhancing AI models by providing advanced low-wattage processors to enable more accurate computer vision. DeepScale’s processors would allow vehicles to carefully perceive the world around them by serving as a pivotal component in a network of mapping, planning and control systems with state-of-the-art sensors.
“DeepScale is a great fit for Tesla because the company specializes in compressing neural nets to work in vehicles and hooking them into perception systems with multiple data types,” stated Chris Nicholson, the CEO and founder of Skymind.
That's what Tesla needs to make progress in autonomous driving.
Established back in 2015, DeepScale is made up of an engineering team that have gained over 30,000 citations and references for their work as well as 25% of the team holding a PhD. They have led pioneering research in machine learning and deep neural networks such as SqueezeNet, a network specifically tasked to supplement computer vision by producing a CNN architecture with fewer parameters.
Co-founder and CEO Forrest Iandola recently announced joining the Tesla team via twitter stating, “I joined the @Tesla #Autopilot team this week. I am looking forward to working with some of the brightest minds in #deeplearning and #autonomousdriving.”
The coming together of the pedigrees of DeepScale and Tesla could mark a significant shift in autonomous driving technology as it could signify a move away from the use of LiDAR technology in this field. This is due in part to the fact that DeepScale’s technology and its creation of SqueezeNet consumes less energy due to smaller architecture which aligns with Tesla’s drive towards creating new energy solutions.
Tesla seems certain that they don't need LiDAR for effective computer vision, but there are lots of other types of sensors you could see on their vehicles in the future, and sometimes just placing a second camera facing another angle can improve the AI model.
Arjan Wijnveen, CEO, CVEDIA
What’s more is that lower power consumption means a lowering in the costs of the technology which would serve as an additional blow to expensive LiDAR systems.
Shifting the paradigm before it has been fully set would be a bold move but Tesla and its most prominent CEO, Elon Musk, are no stranger when it comes to making bold moves or statements. Back in April, Musk claimed the company would produce cars that would be completely driverless as early as the end of this year with robotaxis following shortly after, he said, “We will have more than one million robotaxis on the road… A year from now, we’ll have over a million cars with full self-driving, software… everything.”
While there are those that rebuke Musk’s claims Tesla continue to move forward in improving the software and hardware integrated into their vehicles. Current Model X, S, and 3 vehicles all have built in autopilot and SmartSummon systems and even Netflix, Spotify, or Karaoke applications but the company is quick to maintain the human driver be lucid when operating any these functions.
Yet, it’s not difficult to do a quick internet search to find videos of people chasing empty Tesla vehicles through parking lots or sleeping drivers on the freeway. This has led to some industry insiders and consumers being skeptical of the safety and practicability of Tesla’s claims and their driverless technology but a spokesperson for the company maintained that many of these videos are “dangerous pranks or hoaxes.”
Still, if Tesla’s latest acquisition of DeepScale means a step on the gas in the move towards fully automated vehicles on the roads and the launch of robotaxis there is plenty of regulation to get through. Therefore, vast and rapid improvements of the deep neural networks and the algorithms that steer them need to be made. Tesla already has a massive database when it comes to the R&D of driverless vehicles so the addition of the DeepScale engineering team will allow the company to go from strength to strength.
Not to mention the fact that Tesla has also acquired around five other companies including SolarCity and Maxwell Technologies which places them firmly in the driver’s seat in becoming the firm that makes an automated vision a reality.