AI and self-driving vehicle risks are worth the rewards, says former GM R&D chief

LAS VEGAS — During the time Larry Burns led R&D and planning at General Motors, from 1998 until 2009, his team developed many advanced technologies that eventually came to market. Perhaps the biggest: the skateboard chassis that underpins just about every electric vehicle produced today.

After his GM career, Burns didn't slow down. He advised companies such as Google and Waymo, worked as an engineering professor at the University of Michigan and authored several books. He continues to work as a consultant on mobility and other transportation issues. Burns, 72, spoke with Staff Reporter Richard Truett on the sidelines of CES here this month. Here are edited excerpts.

Q: There have been a lot of negative headlines regarding self-driving technology. Some companies are scaling back, some have failed. What's your take on autonomous driving? Will the industry ever get to Level 5, in which there is no role for a human driver?

A: I don't think Level 5 is ever going to happen. That should never have been the goal. Look at the aerospace industry. There are times when an aircraft shouldn't fly and times when a car shouldn't drive. Level 4 [in which there are still controls for a human to take over] is what the goal is.

Look at the progress that has been made in the last year by Waymo. The company has expanded from Phoenix into San Francisco and L.A. Waymo is doing more and more rides without a driver in the car. They've made great progress, and they've made it by focusing on transparency and safety as the primary factor in advancing their progress.

I rode from Houston to Palmer in Texas in an autonomous Class 8 truck. It's incredible what the truck could do. You could say autonomous vehicles have taken a lot of hits, but if you look at the history of innovation, that happens with every technology. There are a lot of companies that jump in. Not everybody survives. The story should be covered through the lens of the companies who are leading the progress, not the ones who dropped out of the race. This is a marathon. If someone said to me no one's going to make it, I would be concerned. It's a question of when, not whether.

We hear the term AI every day. Could you give a simple definition of artificial intelligence?

The terms AI, artificial intelligence, and ML, machine learning, get interchanged a lot. If you go back five or six years, most of what we were seeing was machine learning. It was the simple fact that we have loads and loads of data that can be collected, analyzed, stored and used to make things better. And why not do that? Now what we are seeing with artificial intelligence is the ability to make more inferences, the ability to take all of that data and all of that stored knowledge and connect the dots.

How could AI be used in product development at an automaker?

Maybe AI is the wrong label. I think about it the way I think about engineering. When I started at General Motors Institute (now Kettering University), I didn't have a pocket calculator; I had a slide ruler. Then we got pocket calculators, and then we got computers, and then we were able to do things like computational fluid dynamics and finite element analysis.

And when I led R&D at General Motors, I began to wonder did people really need to understand calculus anymore because all of the knowledge of calculus was being embedded in code, and engineers were able to use that code to explore way more edge cases to make designs lighter and engines more efficient. I thought it was a good thing to apply that.

AI has also been in the headlines because of the possibility that it can be misused.

I don't think AI is something that we should be afraid of. It should be something we should embrace in terms of mankind moving forward. But, like many technologies, we need to manage it to make sure it delivers real benefits without the real risks.

How can AI be dangerous?

Because of misinformation. When I was a young scientist, I had peer reviews. If I wanted to publish a research paper, I had to send it off to the Journal of Engineering Research. From there, it would go to a number of my peers who would read it and make sure it was correct. Once they all agreed it was good scientific work, they would publish it. But now I think the world has moved so fast that there is a lot of technical information getting out there without peer reviews. That can be risky, and I worry about that, to be honest.

What's the best use of AI for automobiles?

The whole future of the industry is about design innovation. Because innovation is at the heart of competitive advantage in the industry. When you look at design innovation, you want to focus it on the total experience of the people you are trying to innovate for. AI can give engineers a lot more insight than what companies can get from written surveys. I think you get insight from observing. AI can help everyone become better designers and better innovators. Engineers will be able to look at a whole bunch more design opportunities. It's also understanding people at a level of depth and what motivates us.

For much of your career, you've been paid to, in a sense, predict the future and help companies develop technologies for products many years from the market. What else do you see coming?

The industry is going through a major transformation that most of us talk about in terms of electrification and autonomous driving. But there's another transformation coming, and that is even bigger. And that is: Do we even have to travel at all to live our lives? Think about remote work. Think about telehealth. Think about online learning. Think about e-commerce. Think about how you are able to gauge an activity today without having to travel. In the past, you had to travel and were often constrained to a schedule. I think the future of how we access what we want to do in our lives and how we use transportation to get that access is the single biggest strategic question all companies will have to answer.