The Secrets of Self-Driving Cars from Autonomous System Maestro David Silver

Inside this episode....

David Silver is sharing his journey in Self-Driving Cars, from his first internship at Ford Motors, to creating the Self-Driving Car Nanodegree with Sebasitan Thrun, to building self-driving cars at Cruise and Kodiak.

A few samples:

• How David started in self-driving cars, and how he got his first intership at Ford (04:55)
• The email David got that sent him to Udacity to build the Self-Driving Car Engineer Nanodegree (07:14)
• How Sebastian Thrun made the instant decision to purchase a Self-Driving Cars (14:14)
• The difference between coding and teaching self-driving cars (18:27)
• How David joined Voyage, and how he experienced Voyage's acquisition by Cruise (24:00)
• The biggest challenge in working for tech giants like Cruise (27:50)
• and many, many more... including questions from people in the chat about the worth of a Ph.D, building a portfolio, how to get a job at Kodiak, David's best recommendations on skills to learn, what skills are looked for by recruiters, and more more more...

Interview Sample: How Udacity Built its Self-Driving Car

Well, were you in the self-driving car space around 2017? You know, when self-driving cars were absolute madness hype? When everyone and their mother, including me, would leave it all behind to build autonomous cars?

Well, I do remember those crazy years, and to me, there were marked by projects like the Google Car, startups like NuTonomy, and even educational platforms like Udacity. If you remember, Udacity, as education platform, had built its own self-driving car at the time, and was driving it across Silicon Valley to make some noise.

You may know about this, but you may not know how they built it. It turns out, a couple of years ago, I have interviewed David Silver, the head of Udacity's self-driving car program at the time, and read what he revealed to me:

— JEREMY: How hard was it to actually build it [the self-driving car]? And what were the main challenges when building it?

— DAVID: "Yeah, that turned out to be a really challenging thing to build. And it was 100% foreseeable that building a real self-driving car that students could put their code on was going to be a lot of work.

But it was a really special part of the course and kind of became the mascot of Udacity for a while. We gave the car a name, the car's name was Carla. And Carla was wrapped in Udacity colors and he had a Udacity logo.

And actually we hired somebody, Vienna, who I think I saw her name in the chat here, was our kind of our self-driving car engineer and would work with students to run their code on the car at the test track that we had. We had a very small test track, parking lot basically, close to the office.

And we used a lot of open source software. It was kind of cobbling together or integrating stuff that other people had done and then connecting it together in a way that would work for students. So I remember, you know, just a kind of fun story:

There was a company called Autonomoustuff, still is, I guess, that particularly at that time was: the way that software engineers who wanted to work on self-driving cars got a car to work on as they would go to Autonomoustuff. Actually, at that time it wasn't even how they got a car; it was how they got sensors.

Like if you wanted to buy one LIDAR unit, one laser array, or one radar unit, the big automotive suppliers were not necessarily interested in selling you one unit, you know, on eBay or whatever. But Autonomoustuff would kind of buy them in bulk and then sell them out onesie-twosies to whoever wanted them.

We had the sales rep from Autonomoustuff drive over to our office to talk to us, and he drove in a car that was kind of their demo car. And he said, 'Look, if you want to buy some sensors, we can coach you on how to put a self-driving car together.'

And Sebastian Thrun said, 'Well, how much would it cost for us to just keep this demo car that you drove over?'

And the sales rep said, 'We've never actually sold anybody a car before, but if you give me your credit card, I'm sure we can make a deal.' And Sebastian whipped out his wallet and was like, 'Here you go.' And that wound up being the car that we purchased.

Before that, we'd been looking at buying a car from a Ford dealership and building it from the ground up. But we got that car, and then we used some open-source software from a community called Autoware based out of Japan that does a lot of open-source software for education, autonomous vehicle software for education.

And then we cobbled it together with a bunch of code we wrote so that students could send us modules to plug into the car and actually get the car driving around the test track. So it was a big project; it was a lot of fun, and it was kind of at the cutting edge at the time."

And there you have it:

• Car from Autonomoustuff (who had to walk back home I guess 🤷🏻‍♂️)
• Software from Autoware

I don't know about you, but I totally picture Sebastian Thrun pulling out his credit card and buying a self-driving car in 5 minutes.

The story is incredible, because you realize that sometimes, the biggest decisions are not made after months of hard work and study, they can just take a second. It wasn't a careful study of all the possible options they had, it wasn't a detailed listing of everything they could possibly do.

Just 'forward intent', trust in the company they bought from, and a lot of motivation.

There is probably a big lesson inside, where we should all learn how to make decisions faster.

Next Steps