"I'm extremely confident that level 5 [self-driving cars] or essentially complete autonomy will happen, and I think it will happen very quickly," Tesla CEO Elon Musk remarked in a video message to the World Artificial Intelligence Conference on July 9. "I remain confident that we will have the basic functionality for level 5 autonomy complete this year."
Why is it such a big deal that Musk claims level 5 autonomy is within Tesla's reach? Level 5 autonomy means cars can fully self-drive without human assistance. In other words, vehicles that have achieved level 5 autonomy can do all the driving in all circumstances, removing the need for even a steering wheel and driver's seat. Sounds amazing, right?
As is usually the case in Elon's world, his claim has generated substantial controversy. Achieving level 5 autonomy is not just difficult -- it is, according to experts in A.I. and robotics, next to impossible.
So who is right--Elon Musk or a boatload of independent experts? Is Elon's flexing just that, a controversial PR stunt meant to bring even more attention to the eccentric billionaire?
Article continues after video.
FEATURED VIDEOCompass Founder Robert Reffkin on How to Bounce Back From Failure
What Does a Psychologist Have to Say?
From a behavioral science and psychology perspective, we know that the human mind is incredibly sophisticated. One of the reasons why achieving level 5 autonomy is so difficult is that human drivers can use intuition to make rapid yet rational decisions when encountering new situations.
"Distracted and fumbling toddler running away from mother up ahead. Slow down and give space."
"Sketchy-looking masked man pointing a gun behind. Speed up ... and duck!"
We have these rapid responses even though it's a totally novel experience. We simply know the best course of action because the computer in our head has endowed us with the gift of intuition.
But the algorithms that underlie self-driving cars need to be pre-trained for every possible situation they could encounter, distracted toddlers and all. There have been several instances of Tesla cars on Autopilot encountering unusual situations (such as on overturned vehicles on the road) and crashing. In the cases when the algorithm has not been pre-trained for a situation, there is no human intuition to fall back on when forced to make a rapid decision.
On the other hand, the human brain has been trained, so to speak, by millions of years of selective pressures via evolutionary change. As impressive as our computers are today, they pale in comparison to the remarkable adaptive processes of the human brain.
According to Daniel Kahneman, a Nobel prize in economics for his work on human judgment and decision-making, there are three ways that humans develop their intuition:
We create predictable situations. The environment must be sufficiently regular and orderly to allow individuals to generalize past mistakes to future situations.
We practice. At a basic level, the opportunity to hone an intuition exposes an individual to a greater number of potential situations, presenting the opportunity to learn.
We get feedback. For practice to be worthwhile, individuals have to know whether they're right or wrong. Feedback that is unambiguous and instant is most conducive to developing an intuition.
Let's look at each in turn.
Predictability is not an issue for self-driving cars, because the rules of the road are generally universal.
What about practice and feedback? While self-driving cars receive unambiguous and instant feedback just as humans do when a car crashes, it's an open question whether self-driving cars will ever have enough practice in the right domain to develop human-like intuition. While humans can leverage experiences in everyday life to build intuition--ranging in domains as diverse as walking on the street to enjoying a meal at a restaurant, through playing recreational sports--algorithms and self-driving cars are limited to experiences on the road.
This means that humans have a much more diverse set of experiences to build intuition to understand new situations. Self-driving cars, on the other hand, are limited to a single domain: driving.
Whether Musk's claim that Tesla is on the brink of solving one of the hardest technical problems that's ever existed may come down to whether algorithms have had the same set of experiences to practice being human as humans experience throughout the course of daily life.
Are Musk and the Tesla team up to the challenge? I guess we'll have to wait and see. As someone who hates driving long distances, he's sure got my vote.
Comments
Post a Comment