Another month, another scrappy foretelling of the future by Elon Musk. He is mostly known for his electrifying visions of a Mars colony and electric cars as the answer to our gas-guzzling transportation problems. He is a man noted for making bold predictions that seem impossible to fulfill, and then delivering on his promises even if he does so behind schedule.
He has just predicted that he will produce a car that can drive from Los Angeles to New York City without human intervention by the end of 2017. If he succeeds, he will be building autonomous cars two years or more ahead of every big car manufacturer out there.
The nearest possible competitor is China’s Baidu, who are aiming for 2019, followed by Ford, Google, and GM, all aiming for around 2021. Musk’s claim is even more controversial than it seems at first glance, as he is talking about a fully autonomous vehicle while competitor’s cars will be staying in limited areas and will not drive across countries.
Can he make it happen? Let’s look at his track record.
Fully autonomous means self-driving
In the Tesla environment “Full Self-Driving Capability” means that you will be able to get in your car and just tell it where to go. Although Tesla has not (yet) released software that does more than keep vehicles in their lanes, and at safe following distances, an edited video of a Tesla Model X SUV driving autonomously in Palo Alto has been released. Researchers from various universities agree that the video seems impressive, but they aren’t ready to concede that the video provides a proof of concept.
Data and deep learning
Cars that can manage any situation need a lot of learning. Google is considered to be the data Leviathan, with its fleet having covered over two million miles in and around a few cities. We don’t exactly know what data Tesla is collecting, but when we consider that Tesla cars in autopilot mode have already covered more than 222 million miles around the world, we understand that Tesla must be getting a lot of data.
There can be no doubt that Tesla has both bigger volumes of data and much more different data than its competitors – perhaps enough to leapfrog them.
Imaging technology and learning
Musk’s avoidance of LIDAR sensor technology is compounding the difficulty factor of his quest. Google and others use LIDAR which are rapid laser pulses that produce a highly accurate representation of the surroundings. Musk is avoiding reliance on a mechanism that involves moving parts and believes that, unlike radar, LIDAR has trouble in snow, fog, and rain. Instead of the Google “KFC bucket” on top of self-driving cars which costs $80,000, Tesla is concentrating on a forward-facing radar system in tandem with eight cameras.
Researchers are skeptical that Tesla will achieve 100% safety without LIDAR, but Tesla is using the deep learning abilities of Nvidia’s supercomputer, the Drive PX2, to teach cars how to use radar.
While upstart Tesla is already selling cars that are capable of full autonomy, their website makes it clear that: “Self-Driving functionality is dependent upon extensive software validation and regulatory approval, which may vary widely by jurisdiction.” We may have fully autonomous vehicles before we’ve come to terms with the legal aspects of having no one in the driver seat.
Disruptive technology disrupts. Tesla may be changing the landscape from the ground up by continuing to forge ahead in the law’s gray area and adopting an “asking for forgiveness, rather than permission” approach. While Tesla’s cars become gradually autonomous through incremental software roll-outs, we need laws to state at which point a car becomes independent, and we must know how to regulate its behavior.
Is the world ready?
Car crashes cause 1.2 million deaths every year, yet Musk believes this technology to be at least twice as safe as human drivers. Some researchers are concerned that this zeal could cause setbacks for the industry by prematurely introducing technologies that are not quite ready, causing fatalities that trigger push-back from society and regulatory authorities.
Tesla’s self-driving capable cars are already on the market and buyers will soon show a bias. With one reported death in Florida and possible death in China, buyers have failed to air serious concern. The target market for Tesla’s Model 3 is the average consumer and buyers will soon let us know whether they trust computers, and Elon Musk, enough.