The Past and the Futurist -- Toyota Research Institute Chief Technology Officer James Kuffner may be working
on a future filled with helper robots and autonomous cars, but he's still holidng onto his beloved 2004 Camry, which he drives to work every day.
James Kuffner is changing the world.
The autonomous vehicle and artificial intelligence (AI) guru, who left Google in 2016 to become the chief technology officer at Toyota Research Institute (TRI), sees a future filled with mobility on demand and robots handling day-to-day tasks.
That made us want to find out more. So, we did.
To read about Kuffner’s vision for the future, scroll down. To find out which Star Wars
robot he thinks is most useful, watch the video here.
And remember, Kuffner will be speaking to all team members during the North American Business Update meeting at 9 a.m. Central on Tuesday.
Driver’s Seat: What exactly do you guys do over at TRI?
We’re aiming to build world-class technology for artificial intelligence that will power the next generation of robots and cars that would be worthy of the Toyota brand. Our aim is really to build software that would complement the great hardware and mechanical systems Toyota is capable of manufacturing. Our goal is to use this technology to improve the quality of life around the world.
What will we see improve?
You’ll see safer, more accessible transportation for everyone in the next decade. I think you’ll also see robots that are able to do more and more things in a general-purpose way outside of factories – in homes and offices – that will be able to improve people’s quality of life.
How long do we have until the robots are in charge?
(Laughs) The machines human civilization has built over the last few centuries have helped automate physical work, and now we’re seeing computers augment cognitive tasks. As a result of all of these tools that we’re building, we’re going to see AI contributing to improving the tools that we have in order to make society better. I’m very optimistic that what we’re building will be used for the benefit of humanity. Of course, we have to think about how we can responsibly use this technology. But those are problems we can address and solve. The doomsday scenarios are exciting to think about, and they’re good counterpoints to make sure that we’re building responsibly.
What types of things would be available to help elderly people in 30 or 40 years that aren’t available now?
Having freedom of mobility. That will be available to you in the sense that you’ll be able to go see a doctor, friends or grandchildren safely and efficiently with lots of different mobility solutions, probably on demand. And you’ll also have robot assistance that could be part of your home, that could help do daily tasks like cleaning or cooking or reminding you to take your medicine. You’ll be able to talk to them in a conversational way. Maybe we’ll have new interfaces through virtual reality or haptics that will allow you to have a telepresence. Without having to physically transport yourself to a place, you’ll be able to virtually go to something like a grandchild’s birthday party. I do believe all of those things will happen over the next several decades.
What victories have TRI already had on AI and autonomous driving?
In the last 10 years, the comparison with what was possible in terms of perception, prediction and planning has been truly revolutionary. We’ve seen incredible advances in machine learning, mostly because the performance of the computers and data centers has increased so dramatically. And the algorithms have evolved along with those improvements.
Why did you leave Google to come to Toyota?
I spent two years in Japan and collaborated with Toyota while I was a professor at Carnegie Melon University. It’s also no secret that, while I was at Google, we only used Priuses and RX hybrids for research because the hybrid synergy drive is the ideal development platform. You can attach all kinds of exotic sensors to them, and if the voltage level drops, the gas engine powers on and keeps everything charged, so you can test all day and refuel in five minutes. Toyota has really done an incredible job of building the perfect platforms for AI and autonomy. I’m excited to not just be part of Toyota, but now I get to contribute to tremendous software that can have an impact at scale.
Meeting of the Minds -- James Kuffner (right) with his fellow TRI futurists, including (from left), Ryan Eustice,
vice president of Automated Driving, Chris Ballinger, CFO and director of Mobility and Blockchain Services, Gill Pratt, TRI CEO and Toyota Chairman of the Board Takeshi Uchiyamada.
What short-term goals do you have?
We built a lot of infrastructure in our first year. So now an engineer who joins TRI has powerful tools for cloud and data processing and development of software that rivals any IT giant on the planet. In the next year, we will really see the productivity gains because of those tools. They will allow us in to build a robust software base that means Toyota’s cars and robots will have world class capability in terms of robustness and reliability.
I was at an event on Saturday and, afterward, I watched people walking to their car. Chaos. I can picture an autonomous vehicle not knowing what to do in that situation. Or is that something you guys have already figured out?
The reality is that we have unsolved research problems on the critical path to launching these products. One of the unique things about TRI is that we partner with premier universities like Stanford, the Massachusetts Institute of Technology and the University of Michigan, collaborating with some of the leading researchers in the field to help us solve some of these hard problems like predicting where people are going in a swarm. As pedestrians cross the street, the car needs to be able to, in real time, make a good prediction with a probabilistic model of future trajectories. You can do it based on the rules of the road, but you can also do it based on data. And that’s where we’re really excited.
Imagine if you had all this data that captures the subtleties of how people move and the assumptions they make in different locales and different cities. When I was working on the motion planning and prediction system for the Google car, I originally set out to implement the California DMV rulebook. I said, “we can make a car that exactly follows all the rules.” It turns out that doesn’t work at all. According to the rules, you must have 10 car lengths between cars when you merge onto a freeway. But if you try to wait for that large of a gap, it’ll never come. So instead, you can take data and model how the median driver behaves – and the safest way to drive is actually the median driver. That insight allows us to scale to different cities. People drive in New York or Los Angeles or Tokyo very differently, and we don’t want to have to special case craft our software manually to handle all those behaviors. Instead, we can model them specifically by capturing data about how people move and the assumptions people make, and then we can build a system that is able to adapt. Essentially, when in Rome, drive like a Roman.
How difficult is it to develop autonomous vehicles with people still driving cars?
In a world where human drivers are out of the picture, we could remove traffic lights and all kinds of infrastructure that have been put up to help manage traffic safely. That would simplify everything. But the reality is that it’s going to take a long time. So, we’re going to be in a hybrid world of both human-driven cars and AI-driven cars for many years to come. Probably decades. Therefore, we have to build a system that’s backwards compatible with the rest of the world. And that’s why it makes the problem harder, but that is the viable path forward.
How will this change our cities?
I think there could be an incredible impact on the design of cities. Most of our cities have been designed around the car. Imagine a world where you could have cars that can relocate parking out of city centers, increasing the density of commercial or residential areas and just make them more pleasant to live. You could have a lot of land reclaimed to make our urban centers better. And it could be better for the environment as well. There’s so much time lost when you’re stuck in traffic. If we could remove a lot of those bottlenecks with better technologies, our cities could be cleaner, safer and better places to live.
By Dan Nied