For years now, Tesla CEO Elon Musk has garnered both praise and distrust -as well as a significant amount of money- with his innovative ideas, flamboyant style, and provocative Tweets. Starting with a fancy electric sports car called The Roadster in 2009 -followed by numerous other electric vehicles-it seems by now that no tech artifact makes us dream of a brighter future that hasn’t been explored by Elon Musk’s companies, which include Neuralink and SpaceX. Neural implants, brain-machine interfaces, rockets, spacecrafts, nanosatellites, rovers designed to slide through the valleys of Mars, or humanoid robots. Among these, the promise of developing fully safe self-driving cars has remained a constant in Musk’s public statements for a while now.
Early in January, the South African businessman declared that he would be “shocked” if by the end of this year Tesla didn’t achieve full self-driving safer than human driving. Now, this is not the first time we have heard this sort of promise – actually, Musk said more or less the same thing during an earnings call about a year ago. He was later contradicted by Tesla’s own autopilot engineer, who said that this statement didn’t match “engineering reality”. And there have been similar declarations as early as 2015. But of course, despite reasonable anxiety about seeing self-driving cars as part of our everyday lives, all these statements should be put in their proper context. After all, it is a skillful businessman who’s making them, not a scientist or an engineer discussing technical issues among peers.
How far has autonomous driving AI been developed
But in any case, all this raises the question about how far autonomous driving software has been developed at this point, and especially how far we really are from the massive adoption of self-driving vehicles. Actually, the first thing to notice about Tesla’s Full Self-Driving Software (FSD) is that it doesn’t make a car fully self-driving. It is not short of abilities, though. It can recognize stop signs and traffic lights, enter and exit highways, change lanes, and even park. This neural network emulates the functions of the human brain, using images to analyze its surroundings and make relatively complex driving decisions. But it is misleading to call it autonomous since it requires full driver attention at all times. A recent beta version of FSD, though it has shown some new rather impressive abilities, has also disappointed the public with baffling decisions and mistakes, being incapable of performing some basic driving tasks.
Naturally, Tesla is not the only company making progress in autonomous driving AI. Most of you have surely noticed a particular species of CAPTCHA that we have been solving for a while now. These CAPTCHAs ask us to identify which frames contain images of objects related somehow to traffic, such as traffic lights, buses, taxis, etc. Well, just as for years we’ve trained software to read hundreds of thousands of scanned copies of books by solving CAPTCHAs, Google is now leveraging them to train their own autonomous driving software. Waymo, a subsidiary of Google’s parent company Alphabet, is already commercializing self-driving taxis in Arizona. However, Waymo autonomous cars have only been tested within the geographical limits of certain markets, and they also seem to have a long way to go in terms of convincing everybody of their safety.
There is yet another significant barrier to the commercialization and use of self-driving vehicles at a grand scale to become a reality: regulators. Images of self-driving cars hitting parked vehicles or bollards, mistaking signs, or being unable to avoid sizable concrete barriers ahead are causing an intensification of regulatory scrutiny, raising serious questions about how safe it would be to flood the streets with autonomous driving vehicles given the current state of things.
Just a little patience
It is clear that the widespread adoption of self-driving cars still has a lot of challenges ahead, both technically and legally. Yet, it is also rather likely that safe self-driving vehicles will be part of our lives sooner rather than later. Probably long before the first human settlers begin their colonization of Mars, we’ll be making long-distance road trips in self-driving cars, dozing, daydreaming, or watching movies in the backseat as a fully developed autonomous driving software takes us safely across one of the thousands of highways that we have already built in our planet.
We just may have to wait a little more than expected…
If you are interested in reading more about the latest trends in technology, keep reading our blog.