← Lex Fridman

Chris Urmson: Self-Driving Cars at Aurora, Google, CMU, and DARPA

Episode 2019-07-222019-07-22EnglishMembers only

Guests: Chris Urmson

Listen to this episode00:00 / 44:59
Reading settings
Reading preferences
Text size
Download in-depth read
Read offline

English · In-depth read · Current version

Downloads are not available yet.

Text language

In-depth read · Key ideas, reasoning, and examples, organized by topic.

§01Introduction

Lex Fridman introduces Chris Urmson (the transcript mishears the name as Chris Sampson), former CTO of Google's self-driving car team and a key engineer behind Carnegie Mellon's entries in the DARPA Grand Challenges and winner of the DARPA Urban Challenge. He is now CEO of Aurora Innovation, which he founded with former Tesla Autopilot director Sterling Anderson and Uber's former autonomy and perception lead Drew Bagnell. Fridman calls him a longtime voice of reason in a field shrouded in mystery and hype.

§02Lessons from the DARPA challenges and from Red Whittaker

Asked what he learned from the DARPA Grand and Urban Challenges at CMU with Red Whittaker, Urmson says the key realization was that it could be done. As a grad student he faced a dichotomy: the task seemed nearly impossible and so would be cool, but CMU was the only robotics institute around, so falling on their faces would be embarrassing. Having the will to attempt it, and succeeding after a couple of tries, was exciting. He believed it could be done, since one does not attempt the completely impossible, and credits a benefit of naivete: not knowing how hard something is lets you try things wiser people would not.

On the biggest pain points, he says the joy of the field is that every part is hard and you must be good at each. In retrospect the Grand Challenge should be easy today: the world was static with no other actors, out in the desert with good GPS, so it could be mapped roughly. Getting the vehicle controllable was heavy engineering, and the uncertainty of the route was part of the challenge, since DARPA gave only waypoints and a corridor width. The most devious option would have been a kilometer-wide corridor across scrub and rocks, but it became driving along trails, relevant to the application.

From Red Whittaker he learned two things about leadership: to go try really hard things, where the opportunity is, and to see people for who they can be rather than who they are. Whittaker would empower undergraduates and graduate students to lead and do great things that others might dismiss, a trust-but-verify confidence in what people can become that Urmson calls powerful.

Keep reading as a member.

You've read the free preview. Become a member to read the rest, with side-by-side reading and downloads where available.