When I got the email to gauge my interest in flying to Germany to witness an artificially intelligent robot meet a self-driving car, I nearly wet myself. Never having driven in a self-driving car, I was already interested – but what got me to pull out my passport was the underlying theme of autonomy.
We’ve been dreaming of robots and flying cars for decades, but it’s only been in the past few years that these dreams have become a reality. Beneath all the rhetoric, there are a few concepts that are really driving us forward: machine learning, deep learning, and artificial intelligence. But before we dive into how the objects around us are becoming aware of their surroundings and themselves, I’m going to start on the surface with what it was like to drive in an autonomous car.
Nicole Meets Jack
Putting my life in the hands of Jack, Audi’s multi-million dollar prototype was simple. Press two buttons simultaneously with your thumbs, some bells chime and a line of LEDs stretch across the front window change from red to yellow to a soothing aqua blue. That’s the signal that Jack is now in charge.
It’s easy to humanize a car. The front end looks like a face, and Hollywood has shown us endless movies of cars coming to life. But the reality of Jack’s personality or sense of identity is more than just a physical design aesthetic, his identity is embodied in his driving style.
The kind of driver he is tells you about his personality. Is he the guy that sits in the middle lane driving 10 below the speed limit? Is he someone who tailgates and weaves in and out of traffic? He’s neither and I can assure you that what Jack lacks in charming dialogue he makes up for with a bit of swagger on the road.
With my hands off the wheel, and not able to use my phone (you can’t have anything between you and the airbag for safety reasons), I really had nothing to do but sit there and observe Jack. I’d call him a gentleman; his courtship with the road is a very courteous one.
Jack is an active participant with his environment. The first time he went to overtake a slower truck, he let me know he was going to change lanes on the display, he signaled and smoothly moved over. I was surprised a few times when he accelerated with such assertiveness that I was pressed back into the seat, but if I was in charge I would have done the same thing.
As far as learning environments go the A9 Autobahn outside of Audi’s HQ in Ingolstadt offers a wider set of driving conditions because it doesn’t have a speed limit. Trucks and high-end sports cars share the road. He would change lanes to pass slow trucks and get out of the way when fast moving cars would suddenly appear behind him. It’s the little things in his driving style where you’ll find the evolution, things like leaving more room and not driving in the center of the lane when passing a truck.
Where is Jack on his journey to being a fully Autonomous Vehicle?
Audi’s piloted A7, Jack, is a Level 3 autonomous vehicle, which if I’m going to draw parallels to when I learned to drive is kind of like having a learner’s permit. You’ve got an idea of what’s going on, but you still need supervision in certain situations.
Ultimately, Jack’s goal is to drive better than a human, and like any human he’s developing a unique driving style.
Autonomous vehicles have 5 levels, and there currently aren’t any Level 3 cars on the market. To give a bit of background here are the levels:
Level 0 – People do everything
Level 1 – Most functions are controlled by the driver but a few things like steering or acceleration and be done by the car. Think, cruise control the car isn’t staying in its lane, it’s just maintaining a speed.
Level 2 – The car helps you drive, but you need to keep an eye on it. The car can maintain a speed, slows down to avoid other cars and stays in its lane but you need to be ready to intervene in case something happens. Tesla’s “auto pilot” is the most famous example.
Level 3 – We can expect to see these on the road in the next 5 years. The cars won’t just stay in their own lane and avoid other cars: they’ll be able to make decisions. Instead of braking to avoid hitting that slow-moving car, they’ll look around to change lanes. Audi is putting some features of level 3 autonomy into the Traffic Jam Pilot of their upcoming A8. Level 3 cars still need a human around, in case the weather messes with their sensors or the painted lines disappear, so we’re still waiting for the day when napping at the wheel is a good thing.
Level 4 – They don’t need no stinking humans! The car can handle any situation by itself, even in an emergency. It can also guide itself but you’ll still need to be there to start the autonomous functions and the car might ask the driver to intervene, but it’s not dependent on it.
Level 5 – Full Autonomy! The car doesn’t need a driver to be present at all. You can call it to you or send it somewhere, you don’t need to be inside the car at all. Steering wheel optional!
Audi’s engineers work with psychologists and lawyers to help Jack evolve, become a better driver and figure out what to do when faced with any number of life and death situations.
The one topic that I didn’t really get to dive into on this trip was the ethical implications of Autonomous drive. I’m not talking about Jack stealing jobs, but the situations where you need to figure out which life to save. Imagine you’re driving in between two cyclists one with a helmet one without and someone brakes in front of you. Do you hit the cyclist with the helmet because he has a better chance of survival? Or do you hit the car in front of you because dying yourself might be the outcome that you personally could live with?
Of course, there is a great TedTalk discussing the ethical dilemma of self-driving cars. However, leave me a comment if this is something you’d like to see me dive into the next time I’m given the chance to meet decision makers in the self-driving community.
Autonomous Car Meets an Artificially Intelligent Robot
Robots have a bad rap, Hollywood is constantly making them out to be killers that as soon as they become aware the first thing they’ll want to do is destroy humanity. Hanson Robotics is building life like robots specifically to build trusted relationships with people. Sophia’s AI software is being infused with kindness and compassion so she can learn to be empathetic. She’ll have to learn to understand or feel what another person is experiencing.
Sophia is disturbingly “Westworld;” she is clearly a viable step in the direction of robots being contributing members of our society. The breakthroughs in deep learning are what have enabled robots like Sophia to learn empathy and cars like Audi’s Jack to be able to see.
Artificial Intelligence is human intelligence exhibited by machines and it’s been around since the 1950s, but until recently we didn’t have enough data to train the machines or the computing power to process that data. Nvidia has a great article explaining the difference between Machine Learning, Deep Learning, and AI.
Sophia and Jack are in similar positions where they are working to earn the trust of humans. Both are overcoming a skepticism. Jack can’t possibly drive as well as a human and once a robot becomes self-aware the first thing they’ll do is murder their creator.
Sophia’s “brain” runs on Hanson’s MindCloud, a deep neural network and cloud-based AI software and deep learning data analytics program. The Depp Learning, AI and cognitive architecture that makes up Sofia’s neural network allows the robot to maintain eye contact, recognize faces, process and understand speech and hold relatively natural conversations.
What role does Deep Learning play in Self-Driving Cars?
Deep down I can’t help but get excited about the future and understanding how this change in processing data has allowed for Piloted Drive. In the video below, I give you a brief history of deep learning and share part of a longer discussion with Audi’s Head of pre-development automated driving functions, Dr. Klaus Verweyen about the limitations of Deep Learning and Piloted Drive.
When will Jack be on our roads?
This year Audi will be launching the A8 with a Level 3 autonomous drive features appearing in their Traffic Jam Pilot.
Currently, we have Level 2 autonomy appearing in our cars which means it can only “help” you drive. The driver is always in control even if their hands and feet are completely disengaged and that car is able to maintain its speed, slow down to avoid other cars or stay in its lane. Level 2 means that the car can give up control at any time and it’s up to the driver to take over. Tesla’s “autopilot” system is the best-known example, even if some drivers treat it like a more advanced system.
The Level 3 Autonomy that will appear in the Traffic Jam Pilot means that the driver will be able to turn their attention to a task other than driving in specific situations. Basically, rather than needing to be available to take control at all times the driver is expected to be available for occasional control, but with sufficiently comfortable transition time. If you watched the video of my first autonomous drive (at the start of this article) you’ll see that Jack asked me to take back control of the car when we enter into a construction zone.
Traffic Jam Pilot will work if you’re traveling between 0 and 65 km/h (40.4 mph). The car will accelerate, brake and steer itself. When the traffic jam dissolves or the road ends, you’ll be prompted to take back control.
It’s going to be exciting to see how the exact limitations of the Traffic Jam Pilot are going to be communicated when the Audi A8 is released later this year. Personally, the biggest problem that I had during my drive was not knowing what to do with my hands. You can’t physically use your phone because you can’t have anything between you and the air bag for safety reasons, so the exact role of the car in increasing my productivity is going to be interesting.
It’s inspiring to think of the artificial intelligence that will be entering into our daily lives. Well, almost –as one of the top luxury cars on the market you’re going to need a pretty big stack of cash to afford it. The future is here, it’s just not evenly distributed yet.