Summary of A16z Podcast Episode: Are We There Yet? The Road to Autonomous Vehicles | Podcast
Podcast: A16z
6 min. read
— Description —
Discover how Waymo is revolutionizing the future of transportation with fully autonomous cars that are safer than human-driven vehicles Through machine learning and advanced sensors, Waymos technology can anticipate and react to its surroundings, making informed decisions to ensure safety With over 20 million miles of testing and billions of miles of simulation, Waymos track record speaks for itself
Experience the magic of autonomous driving and imagine a world where parking, air pollution, and wasted real estate are a thing of the past.
Are We There Yet? The Road to Autonomous Vehicles | Podcast
Table of contents
Key Takeaways
Intro
The 5 Levels Of Autonomy
The Technology Behind Waymo
LiDAR vs Video Debate
How Waymo Differentiates
Technological Unlocks On The Horizon
The Roles Of AI In Autonomous Vehicles
How Waymo Views Safety
Collaborating With Regulators
Waymo’s Expansion Strategy
Societal Unlocks Enabled By Autonomous Vehicles
Key Takeaways
- Waymo believes that a fully autonomous car will be safer than the human-driven alternative
-
Autonomous driving can be distilled down to:
- Is the car aware of what’s happening around it?
- Can it anticipate what the things around it are going to do?
- Given its surroundings, can the vehicle reason on what it should do next?
- Autonomous vehicles use machine learning to make predictions on what pedestrians will do next depending on the gate and hand motions of the given pedestrian
- Instead of debating LiDAR vs video, Saswat believes the more important question is whether or not the radars, lasers, cameras, and sensors position the vehicle better than the individual
- Waymo has driven +20M miles in testing and billions of miles of simulation
- Waymo took every fatal crash that ever occurred in Phoenix, re-simulated them, and showed that Waymo’s technology could have avoided the crash
- Our society is designed around driving; consider how much of our cities are designed around parking, the amount of air pollution idling cars cause, and the amount of wasted real estate designated for parking
- Any sufficiently advanced technology is indistinguishable from magic
Intro
- Saswat Panigrahi (@saswat101) is the Chief Product Officer at Waymo (@Waymo), an autonomous driving technology creating a new way forward in mobility
- In this conversation, Saswat Panigrahi and Steph Smith go for a ride in Waymo’s autonomous vehicle and discuss the state of self-driving, Waymo’s advantages over the competition, the differences between LiDAR and video, the future of autonomous driving, societal unlock enabled by autonomy, and more
- Check out these Podcast Notes on Marc Andreessen talking about AI sentience
- Host: Steph Smith (@stephsmithio)
The 5 Levels Of Autonomy
- Level 1 autonomy is like driver assistance
- Level 2 autonomy is lane sensing and automatic braking
- Level 3 autonomy requires human intervention within a matter of seconds
- Level 4 autonomy requires no human intervention
- Level 5 autonomy is “anywhere, anytime”
The Technology Behind Waymo
- Understanding the size of the opportunity on the other side of an innovation obstacle is the key to success
- Saswat and Waymo believe that a fully autonomous car will be safer than the human-driven alternative
- The hardest challenges were building the driver itself and measuring its performance
- Waymo built the full stack of hardware and software for its autonomous driving cars
- Waymo built the radars, cameras, lasers, hardware, and simulation infrastructure necessary to achieve autonomous driving
-
Autonomous driving can be distilled down to:
- Is the car aware of what’s happening around it?
- Can it anticipate what the things around it are going to do?
- Given its surroundings, can the vehicle reason on what it should do next?
- Lasers help identify objects in the 360-degree surrounding area, and cameras are needed to distinguish between red lights and green lights, among other things
- Radars help the car see around corners even when the lasers, cameras, and human eyes cannot
- Autonomous vehicles require an “insane” amount of machine learning
- For example, machine learning makes predictions on what pedestrians will do next depending on their gate and hand motions
LiDAR vs Video Debate
- The LiDAR vs video debate has become ideological, in a sense
- The best approach is taking a first principles approach
- LiDAR has strengths that a camera does not, and vice versa
- LiDAR might perform better at night, but cameras better identify colors during the day
- Instead of debating LiDAR vs video, Saswat believes the more important question is whether or not the radars, lasers, cameras, and sensors position the vehicle better than the individual; he believes the answer to that question is resounding “yes”
How Waymo Differentiates
- Waymo manufactures its LiDAR
- Waymo found that the best LiDAR and radio on the market were not optimized for the task of autonomous driving, so it decided to build those technologies itself
Technological Unlocks On The Horizon
- Autonomous driving applies to all forms of transportation
- Waymo has driven +20M miles in testing and billions of miles of simulation as of this writing
- Waymo’s competitors have not driven anywhere near this amount of miles
The Roles Of AI In Autonomous Vehicles
- AI helps bring down costs for autonomous driving
- AI also helps improve the effectiveness of simulations for autonomous driving
- There is AI at every level in the autonomous driving software stack
How Waymo Views Safety
- Waymo took every fatal crash that ever occurred in Phoenix, re-simulated them, and showed that Waymo’s technology could have avoided the crash
- Waymo is the first company to cross the one million miles of fully autonomous driving
- It did not record a single collision with injury in its one million miles of autonomous driving
- Waymo has recorded a total of 160 years of human driving time
Collaborating With Regulators
- Waymo exists to make driving safer
- The company has a “deep fundamental alignment” with what the regulators are trying to achieve
- Within the vehicle, the software provides explanations for why it made certain decisions
Waymo’s Expansion Strategy
- In the early days of Waymo, the company went to various cities that would challenge its driver in many different directions
- It went to 20 different cities to ensure that it was building a generalizable driver, and not one that just works in one location
- Various cities provide better testing elements than others; for example, Miami for heavy rain, Death Valley for extreme temperatures, and Tahoe for snow
Societal Unlocks Enabled By Autonomous Vehicles
- Our society is designed around driving; consider how much of our cities are designed around parking
- A parked car in a city takes up an underutilized, expensive asset (real estate)
- Think of the amount of space parking takes up in a city; parking does not contribute to the productivity of a city
- We dedicate more space in cities to sleeping cars than sleeping humans
- Consider the amount of air pollution idling cars cause in cities
- Consider the amount of real estate in cities is lost to idling or parked cars
- Any sufficiently advanced technology is indistinguishable from magic