Are Autonomous Vehicles Finally Here? Buckle up!

1. 0:00 – 2:06

   A first look at Waymo

   1. SPSaswat Panigrahi0:00

      We are in San Francisco. We're not taking you to some, you know, little test facility in a desert to show you the discounted rides-

   2. SSSteph Smith0:09

      We are on the road. We are on the road, very much

   3. SPSaswat Panigrahi0:10

      ... we are in one of the most vibrant cities on the planet. That's where we are in the journey. We're fully autonomous where we need to be.

   4. SSSteph Smith0:17

      Self-driving is here. I think if I was a human driving there, like, I don't know what would've happened. [laughs]

   5. SPSaswat Panigrahi0:22

      Uh, you were asking me, "Does it get boring for you after working here?"

   6. SSSteph Smith0:25

      Oh, I love-

   7. SPSaswat Panigrahi0:25

      I love it

   8. SSSteph Smith0:26

      ... seeing your excitement. And we are not the only ones who have noticed.

   9. SPSaswat Panigrahi0:29

      That kid's just noticing that we don't have a driver.

   10. SSSteph Smith0:31

       [laughs]

   11. SPSaswat Panigrahi0:32

       Look at that.

   12. SSSteph Smith0:32

       I love seeing so many people when they look inside-

   13. SPSaswat Panigrahi0:35

       Yeah

   14. SSSteph Smith0:35

       ... look [laughs] just like, "What's going on here?" Today, I got to ride fully autonomously with Waymo's Chief Product Officer, Saswat Panigrahi, who's actually been on this journey since 2016. We discussed so much, including the 5 levels of autonomy.

   15. SPSaswat Panigrahi0:52

       Fully autonomous. Uh, as you can see, nobody in the front seat.

   16. SSSteph Smith0:55

       Mm-hmm.

   17. SPSaswat Panigrahi0:55

       No expectation of a human to take over.

   18. SSSteph Smith0:57

       The infamous LIDAR versus video debate.

   19. SPSaswat Panigrahi1:00

       Saying you love LIDARs and hate cameras or vice versa is saying you love one wavelength [laughs] versus the other wavelength, right? It- it's not a fundamental thing.

   20. SSSteph Smith1:09

       Regulation, user experience, the role of AI in all this, but especially the question, if autonomy is truly here, now what? As a reminder, the content here is for informational purposes only. It should not be taken as legal, business, tax, or investment advice, or be used to evaluate any investment or security, and is not directed at any investors or potential investors in any a16z fund. Please note that a16z and its affiliates may also maintain investments in the companies discussed in this podcast. For more details, including a link to our investments, please see a16z.com/disclosures. [upbeat music] All right. Before we jump in, I just wanted to say that we took a 60-minute journey around San Francisco, and

2. 2:06 – 4:20

   5 levels of autonomy

   1. SSSteph Smith2:06

      we saw just about everything. We saw a construction zone, we saw someone running a red light. We literally ran into this group of bicyclists, probably, like, 20 to 30 of them. We even saw what I think was a two-year-old kid who could just tell, they could just tell that something about the front seat seemed off because there was no driver. So I hope you enjoy this episode. This really was such a special experience for me. I'll tell you more at the end of this episode why that is. But I hope you enjoy this episode, I hope you enjoy this ride as much as I did. Wow. Very clean. So what do I do? Do I just... Oh yeah, the KB up there. Nice. Oh. I, like, don't even know how [laughs] to interact with it.

   2. SPSaswat Panigrahi2:55

      [laughs] There's the KB. That's it.

   3. SSSteph Smith2:57

      All right.

   4. SPSaswat Panigrahi2:57

      I'll head in on that side.

   5. SSSteph Smith2:58

      We just jump in?

   6. SPSaswat Panigrahi3:00

      Sure.

   7. SPSpeaker3:00

      Hey there, Catherine.

   8. SSSteph Smith3:01

      Thank you.

   9. SPSaswat Panigrahi3:01

      Sure.

   10. SSSteph Smith3:02

       [door closing] This is cool.

   11. SPSaswat Panigrahi3:08

       [laughs]

   12. SSSteph Smith3:10

       Do we just click Start?

   13. SPSaswat Panigrahi3:11

       Yeah. Do you want to? Go ahead. [instrumental music]

   14. SSSteph Smith3:15

       Oh, wow.

   15. SPSpeaker3:16

       Heading to Ashbury. Please make sure your seat belt is fastened.

   16. SSSteph Smith3:20

       This is wild.

   17. SPSpeaker3:20

       Happy Friday.

   18. SSSteph Smith3:21

       [laughs] This is cool. So you can play music.

   19. SPSaswat Panigrahi3:24

       Yeah.

   20. SSSteph Smith3:24

       You can... And you can just ask it to pull over if you want.

   21. SPSaswat Panigrahi3:27

       Yeah, anytime.

   22. SSSteph Smith3:27

       At any time.

   23. SPSaswat Panigrahi3:28

       Uh, and you can call support. And, and there was an audio cue to remind you in case folks, uh, get in and they, you know, into a minute think like, "Hmm, this is... I want to speak to somebody." [laughs]

   24. SSSteph Smith3:37

       Yeah. [laughs] Like, "Hey."

   25. SPSaswat Panigrahi3:38

       But it turned out actually there wasn't, uh, you know-

   26. SSSteph Smith3:40

       Really

   27. SPSaswat Panigrahi3:40

       ... folks just got eased into it pretty quick. But somehow, uh, this, the smoothness which, with which it drives-

   28. SSSteph Smith3:46

       Yeah

   29. SPSaswat Panigrahi3:46

       ... builds an instant trust.

   30. SSSteph Smith3:48

       Totally.
3. 4:20 – 6:55

   Technology challenges

   1. SPSaswat Panigrahi4:20

      a person in the front there.

   2. SSSteph Smith4:21

      Mm-hmm, mm-hmm.

   3. SPSaswat Panigrahi4:22

      But they didn't have actual control. So that was my first.

   4. SSSteph Smith4:25

      Yeah.

   5. SPSaswat Panigrahi4:25

      And 2019 was the first time I was in a car with truly nobody on a public street, uh, just driving around.

   6. SSSteph Smith4:30

      Yeah.

   7. SPSaswat Panigrahi4:31

      And every single time, even if, of course, you know, the year before, the month before, the day before, and the hour before I was in anticipation and preparing for it-

   8. SSSteph Smith4:39

      Yeah

   9. SPSaswat Panigrahi4:39

      ... still, being inside it, it's, it's truly special.

   10. SSSteph Smith4:42

       I mean, it's so cool to just see it navigating.

   11. SPSaswat Panigrahi4:44

       Yeah, yeah.

   12. SSSteph Smith4:45

       The slight turns on the wheel. Something I have to ask about is it's felt like this is coming for a while. Like, I almost feel like, you know, the future is here.

   13. SPSaswat Panigrahi4:52

       Right, right.

   14. SSSteph Smith4:52

       Like, we're seeing this, we're sitting in a, a car with no driver.

   15. SPSaswat Panigrahi4:55

       Right.

   16. SSSteph Smith4:55

       But a lot of people would kinda say, you know, like, "This has been promised every two years," kind of thing.

   17. SPSaswat Panigrahi5:01

       Right.

   18. SSSteph Smith5:01

       And it's like we're finally at the two years in a way.

   19. SPSaswat Panigrahi5:03

       Yeah.

   20. SSSteph Smith5:04

       And so maybe you could kind of map out where we are in that kind of arc, the, the 5 levels of autonomy.

   21. SPSaswat Panigrahi5:10

       Yeah.

   22. SSSteph Smith5:10

       And also, you know, where we still have left to go.

   23. SPSaswat Panigrahi5:14

       Yeah, certainly. We are in level 4 right now.

   24. SSSteph Smith5:17

       Yeah.

   25. SPSaswat Panigrahi5:17

       So fully autonomous. Uh, as you can see, nobody in the front seat.

   26. SSSteph Smith5:20

       Mm-hmm.

   27. SPSaswat Panigrahi5:20

       No expectation of a human to take over. And, um, you know, i- in level 2 and 3, it's really, really crucial to, uh, communicate the expectations to the driver.

   28. SSSteph Smith5:30

       Yep.

   29. SPSaswat Panigrahi5:30

       'Cause it's very easy that during a normal situation the driver feels, "Hmm, car is kind of driving well, so I can, you know, pick up a book and start reading." No. [laughs]

   30. SSSteph Smith5:37

       Yeah.
4. 6:55 – 11:58

   The technology behind Waymo

   1. SPSaswat Panigrahi6:55

      fully autonomous car. I'm not supposed to be interfered with in this manner."

   2. SPSpeaker6:58

      Yeah.

   3. SPSaswat Panigrahi6:58

      So it'll, it'll pu- uh, pull over, basically.

   4. SPSpeaker7:01

      Okay, got it.

   5. SPSaswat Panigrahi7:01

      Uh-

   6. SPSpeaker7:01

      So we've talked about the kind of arc of innovation.

   7. SPSaswat Panigrahi7:04

      Yeah.

   8. SPSpeaker7:05

      You've also worked in this industry for a long time.

   9. SPSaswat Panigrahi7:07

      Yep.

   10. SPSpeaker7:08

       Since 2016, is that correct?

   11. SPSaswat Panigrahi7:09

       Yeah, that's right. That's right.

   12. SPSpeaker7:10

       Okay. So tell me a little bit more about, you know, the barriers along the way-

   13. SPSaswat Panigrahi7:14

       Yeah

   14. SPSpeaker7:14

       ... to level 4. Um, what would you say, was, is, was it the technology, the regulation, some combination? Are there major factors that have really delayed us from getting to this point where we're now at, at level 44?

   15. SPSaswat Panigrahi7:25

       Yeah, yeah. And, and one thing, you know, you mentioned the innovation arc. I, I always reflect that, you know, on any journey, uh, when you're trying to do something that has never been done before, there'll always be ups and downs, new challenge.

   16. SPSpeaker7:36

       Yeah.

   17. SPSaswat Panigrahi7:36

       Some challenges you foresaw, some you didn't.

   18. SPSpeaker7:39

       Yeah.

   19. SPSaswat Panigrahi7:39

       Uh, the key is, are you clear on how massive the, um, you know, prize or the benefit to society is-

   20. SPSpeaker7:46

       Yeah

   21. SPSaswat Panigrahi7:46

       ... on the other side of it?

   22. SPSpeaker7:47

       Mm-hmm.

   23. SPSaswat Panigrahi7:48

       'Cause then that makes it all worth it. So we were super clear from the get-go that a fully autonomous car that does not get drowsy, uh, that, uh, you know, we even have physical constraints, right?

   24. SPSpeaker7:58

       Yeah.

   25. SPSaswat Panigrahi7:58

       So even if, even when you are alert, uh, let's say you were looking for parking on this side of the street.

   26. SPSpeaker8:03

       Mm-hmm.

   27. SPSaswat Panigrahi8:04

       Your face would be turning towards that constantly looking for-

   28. SPSpeaker8:07

       Yes

   29. SPSaswat Panigrahi8:07

       ... that parking spot.

   30. SPSpeaker8:08

       Yeah.
5. 11:58 – 16:21

   LiDAR vs video debate

   1. SPSaswat Panigrahi11:58

      uh, and tremendous amount of machine learning.

   2. SPSpeaker12:00

      Yeah.

   3. SPSaswat Panigrahi12:00

      The ver- each stage here, by the way, requires an insane amount of machine learning. [laughs]

   4. SPSpeaker12:05

      And it's not the type of thing that you can just, like, put on the road and see, like-

   5. SPSaswat Panigrahi12:07

      No, no

   6. SPSpeaker12:08

      ... oh, let's see if it makes a mistake. [laughs]

   7. SPSaswat Panigrahi12:09

      Yeah. We're not trying to tell you, is this a cat image or a dog image? Like, it's-

   8. SPSpeaker12:12

      Yeah. [laughs]

   9. SPSaswat Panigrahi12:13

      ... uh, it's, it's pretty, uh... So for example, for that pedestrian, uh, in addition to seeing that they're there, acknowledging, which was this problem one, you got to look at even their gait, their hand movement, their leg movement to anticipate, are they about to take motion?

   10. SPSpeaker12:26

       Yeah.

   11. SPSaswat Panigrahi12:27

       And if you are always conservative, which was, you know, say four years ago, it was not like we couldn't detect the pedestrians. We could totally detect them. It was this nuance of are we being over-conservative? Assuming they may jump in, hence let's not move-

   12. SPSpeaker12:39

       Right

   13. SPSaswat Panigrahi12:39

       ... versus confidently moving forward.

   14. SPSpeaker12:42

       Ooh. So are we at-

   15. SPSaswat Panigrahi12:44

       Our first, uh, stop of the multi-stop.

   16. SPSpeaker12:46

       Amazing.

   17. SPSaswat Panigrahi12:47

       Yeah.

   18. SPSpeaker12:48

       Yeah

   19. SPSaswat Panigrahi12:49

       So yeah, that anticipation of what a car is gonna do, what a pedestrian is gonna do, what a child is gonna do-

   20. SPSpeaker12:54

       Mm-hmm

   21. SPSaswat Panigrahi12:54

       ... 'cause they can erratically jump through.

   22. SPSpeaker12:56

       Yeah.

   23. SPSaswat Panigrahi12:56

       What a motorcyclist is gonna do. Are they gonna lane split and speed and get round you?

   24. SPSpeaker13:00

       Mm-hmm.

   25. SPSaswat Panigrahi13:00

       So all these motion models and understanding of how people behave, how this gentleman is walking, and what their gait and motion tells you about where they're gonna go, that is the second. And third and final, what should the car do? This, this feeling that you had, that it gently accelerated but not harshly.

   26. SPSpeaker13:16

       Yes.

   27. SPSaswat Panigrahi13:16

       That takes into account not just, look at this gentleman. He is almost, has his feet almost in the crosswalk but he's not intending to cross.

   28. SPSpeaker13:23

       Yeah

   29. SPSaswat Panigrahi13:24

       ... because he has a stop sign. A, a very conservative system would just come to a stop.

   30. SPSpeaker13:28

       Mm.
6. 16:21 – 19:01

   How Waymo differentiates

   1. SPSaswat Panigrahi16:21

      position you better than individual?

   2. SPSpeaker16:23

      Mm-hmm.

   3. SPSaswat Panigrahi16:23

      The answer is yes. We can show you that there are situations in which camera will be insufficient. Now the question is a practical economic one. Is your ability to bring this public good to a large number of individuals hindered by the fact that these things are expensive?

   4. SPSpeaker16:37

      Yes.

   5. SPSaswat Panigrahi16:38

      And LiDARs, you know, 20 years ago if somebody told you that all these Toyotas are, or, or these cars are gonna have radars, you'd be like, "No, radars are expensive." Guess what? Every single c- you know, most cars have radars today. Uh, um, cameras w- on cellphones were a novelty. You know?

   6. SPSpeaker16:55

      Yes.

   7. SPSaswat Panigrahi16:55

      The old Nokia phones-

   8. SPSpeaker16:56

      Yeah

   9. SPSaswat Panigrahi16:56

      ... would have... You'd be like, "Oh, who needs a camera on a phone," right?

   10. SPSpeaker16:58

       [laughs] It's like four megapixels.

   11. SPSaswat Panigrahi16:59

       You, you may have... I'm sure, and now cellphones have better cameras than dedicated cameras of six years ago.

   12. SPSpeaker17:04

       Yep.

   13. SPSaswat Panigrahi17:04

       LiDAR is going through the same transformation.

   14. SPSpeaker17:06

       Okay.

   15. SPSaswat Panigrahi17:07

       Right now the iPhones have LiDARs, right? Uh, the amount we have been able to cost down these LiDARs in the last two years is incredible, uh, uh factor.

   16. SPSpeaker17:14

       Okay, so you're not concerned about that?

   17. SPSaswat Panigrahi17:16

       No, no. And we, we n- you know, we had, uh, four, five years ago we had that belief.

   18. SPSpeaker17:20

       Mm-hmm.

   19. SPSaswat Panigrahi17:20

       Now we have that proof. Uh, because, you know, hardware generations, you know, there's multiple examples outside of Waymo to see something that began... Chips are a great example.

   20. SPSpeaker17:31

       Yes.

   21. SPSaswat Panigrahi17:31

       Uh, cameras are a great example. So it would, it would be surprising if you had hardware that you were able to package, and then with focused effort you weren't able to. So we, we had that belief.

   22. SPSpeaker17:42

       Mm-hmm.

   23. SPSaswat Panigrahi17:42

       And now we have the proof.

   24. SPSpeaker17:43

       Right. And Waymo actually manufactures LiDAR-

   25. SPSaswat Panigrahi17:46

       Yes

   26. SPSpeaker17:46

       ... correct? And I feel like Waymo, in a way, has chosen to, to your point, like, manufacture the hardware-

   27. SPSaswat Panigrahi17:52

       Yep

   28. SPSpeaker17:52

       ... work on its own software, simulation technology. Give me the thought process there in terms of, you know, there's always this question in business.

   29. SPSaswat Panigrahi17:59

       Right.

   30. SPSpeaker17:59

       Do I build? Do I borrow? Do I buy?
7. 19:01 – 21:26

   Technological unlocks on the horizon

   1. SPSaswat Panigrahi19:01

      uh, each single time

   2. SPSpeaker19:02

      And as we look forward, um, the question becomes, you know, where do you build a moat? Because now you're not the only, uh, company that has achieved some level of autonomy, and you could imagine, like, let's just say a future where we've just achieved level 5 in many places with many companies. How do you think about what differentiates? Is it really a data moat based on the amount of training that a company has been able to do? Is it owning, like, the proprietary LiDAR that is just, you know 10X cheaper than-

   3. SPSaswat Panigrahi19:27

      Yeah

   4. SPSpeaker19:27

      ... the competitor? I'm trying to think ahead in terms of, yeah, just where, where does value really accrue in that-

   5. SPSaswat Panigrahi19:34

      Yeah

   6. SPSpeaker19:34

      ... future?

   7. SPSaswat Panigrahi19:34

      Right, right. Uh, so first and just, uh, you know, before talking about the moat, just thinking of the space we're operating in-

   8. SPSpeaker19:40

      Yeah

   9. SPSaswat Panigrahi19:41

      ... we're talking about a space with a trillion, you know, miles, right?

   10. SPSpeaker19:45

       Yes.

   11. SPSaswat Panigrahi19:45

       So it's vast. It... I- imagine, um, when, let's say, the first cars were being built, folks, uh, said, "Hey, is there space for only one provider or two providers?"

   12. SPSpeaker19:56

       Mm.

   13. SPSaswat Panigrahi19:56

       Right?

   14. SPSpeaker19:56

       Mm-hmm.

   15. SPSaswat Panigrahi19:56

       Clear. Wh- when you're talking about a space of a trillion miles of today-

   16. SPSpeaker20:00

       Yeah

   17. SPSaswat Panigrahi20:00

       ... and then you think about what is the potential value add when you have a driver that drives itself. 'Cause partly, if you think about the miles today, uh, uh, you know, we talk about a truck driver shortage, for example.

   18. SPSpeaker20:12

       Yeah.

   19. SPSaswat Panigrahi20:12

       In fact, even those commercial miles are kind of stunted by the lack of availability of-

   20. SPSpeaker20:17

       Sure

   21. SPSaswat Panigrahi20:17

       ... of drivers.

   22. SPSpeaker20:18

       Yeah.

   23. SPSaswat Panigrahi20:18

       So, so really we're talking about a space that is all cars, trucks, and all transportation. When something is that vast, uh, and the number of autonomy players today are much smaller than, uh, three years ago, if you look at it. Because it is, you know, a pretty challenging problem still.

   24. SPSpeaker20:34

       Mm-hmm.

   25. SPSaswat Panigrahi20:34

       So I think the universe is different. But still a valid question. We do, um, we do believe that there's a, um, you know, knowledge curve.

   26. SPSpeaker20:43

       Mm-hmm.

   27. SPSaswat Panigrahi20:44

       So by having driven, uh, 20 million plus miles, uh, in testing, by having, uh, done billions of miles of simulation, we become aware of problem spaces that others may not have discovered yet. And that goes into hardware design and software design and simulation design. Hardware design in particular has long lead times.

   28. SPSpeaker21:03

       Mm. Mm-hmm.

   29. SPSaswat Panigrahi21:03

       So that becomes a, you know, s- uh, flywheel effect. Uh, your question about how did you know you had to bui- build your own laser, because by that time we had already-

   30. SPSpeaker21:11

       [laughs]
8. 21:26 – 24:17

   The role of AI in autonomous vehicles

   1. SPSaswat Panigrahi21:26

      you and or I were driving-

   2. SPSpeaker21:26

      Well, I was gonna say-

   3. SPSaswat Panigrahi21:27

      Yeah

   4. SPSpeaker21:27

      ... like, I think if I was a human driving there, like, I don't know what would've happened.

   5. SPSaswat Panigrahi21:31

      And you would definitely not have been able to carry a conversation [laughs] while that was happening.

   6. SPSpeaker21:35

      No, no, no, no. [laughs]

   7. SPSaswat Panigrahi21:35

      Right?

   8. SPSpeaker21:36

      No.

   9. SPSaswat Panigrahi21:36

      Uh, so...

   10. SPSpeaker21:36

       I can barely, I can't even, you know, talk to someone next to me-

   11. SPSaswat Panigrahi21:39

       Right

   12. SPSpeaker21:39

       ... most of the time when I'm driving. Um, but okay, so talking about the technology, um, I wanna talk about safety next, but first, are there any important technological unlocks that you still see on the horizon that not just Waymo but the, you know, the industry of autonomy is still trying to solve? Is it really that cost curve, or is there something else that is still in the way of us really rolling this out more broadly?

   13. SPSaswat Panigrahi21:59

       I mean, look, we are definitely, the rate of innovation, even just within Waymo, which I can speak to most confidently, is massive. I'll, I'll give you a concrete example. Um, when we came from Phoenix to SF, we did have some work to do to, you know, adapt, uh, you know, the assertiveness in the driving here is different from the Phoenix driving. But when we went to Los Angeles, the driver worked shockingly good from the get-go.

   14. SPSpeaker22:23

       Really?

   15. SPSaswat Panigrahi22:23

       Yeah, shockingly. So it was, uh, uh... And now there's a portion of, uh, Scottsdale-

   16. SPSpeaker22:28

       Okay

   17. SPSaswat Panigrahi22:28

       ... which is a northeastern part of our Phoenix territory. It's a, you know, much denser area. Lots of restaurants, lots of shops. There we were able to go in, like, within two months. We just went there. We decided we're gonna open up Scottsdale within two months.

   18. SPSpeaker22:41

       Mm. Mm-hmm.

   19. SPSaswat Panigrahi22:41

       And the reason for that is we're truly, the driver's generalizing very well.

   20. SPSpeaker22:45

       Okay.

   21. SPSaswat Panigrahi22:45

       And the concept is pretty intuitive. Uh, and this is advancement in AI that's enabling it. But think of a driver that's capable of this kind of tight traffic navigation.

   22. SPSpeaker22:55

       Yes.

   23. SPSaswat Panigrahi22:55

       Lots of pedestrians and cyclists, but low speed of travel. That's where we are right now.

   24. SPSpeaker23:00

       Yes.

   25. SPSaswat Panigrahi23:00

       Now imagine in Phoenix, 45 miles per hour, three lane, four lane streets, lots of oncoming traffic and being able to navigate that. Pretty much every good weather city is like a linear combination of those two things, right?

   26. SPSpeaker23:13

       Okay.

   27. SPSaswat Panigrahi23:13

       So in Los Angeles, you go to-

   28. SPSpeaker23:14

       You kind of got the-

   29. SPSaswat Panigrahi23:15

       Yeah

   30. SPSpeaker23:15

       ... the best of both worlds.
9. 24:17 – 31:25

   How Waymo views safety

   1. SPSpeaker24:17

      or, you know, let's say it is an extremely foggy day, it's a rainy day-

   2. SPSaswat Panigrahi24:20

      Yeah

   3. SPSpeaker24:20

      ... you're in a new environment. Is it a different, slightly fine-tuned model based on different situations? Or is it all one aggregate that's just ingesting all of this information?

   4. SPSaswat Panigrahi24:31

      It's, um, definitely many, many deep, uh, models. Some very general, extremely deep learning models, and some specialized models w- uh, to make them really good at some very hard tasks.

   5. SPSpeaker24:42

      Specific things.

   6. SPSaswat Panigrahi24:42

      Like for example, understanding a pedestrian's intent is such a vast space. It's like understanding humans, right?

   7. SPSpeaker24:49

      I know, and really hard to understand. [laughs]

   8. SPSaswat Panigrahi24:50

      Like, it's, it takes us a lifetime to understand ourselves. Uh, so understanding human, uh, behavior and motion, there could be specific models.

   9. SPSpeaker24:57

      Mm-hmm.

   10. SPSaswat Panigrahi24:58

       Uh, there could be end-to-end models on just driving like a good ci- citizen, uh, polite to other riders. That can be a more end-to-end model. Um, being, um, you know, comfortable to rider's preferences, that can be a very, uh, you know, end-to-end model as well. So it's a mix of this. And there's AI at every layer of the stack, from perceiving the world, to predicting other people's behavior, to the driving, to the testing.

   11. SPSpeaker25:20

       Yeah.

   12. SPSaswat Panigrahi25:20

       So for example, the, you asked about fog. Uh, so what we try to do is- We both observed how other people drive in fog.

   13. SPSpeaker25:29

       Okay.

   14. SPSaswat Panigrahi25:29

       We also tried to reason about how well can we see in fog.

   15. SPSpeaker25:32

       Mm.

   16. SPSaswat Panigrahi25:32

       So, if this fog were to get a lot denser, the appropriate thing to do is I can't see that far, so I shouldn't be driving as fast as I would normally do.

   17. SPSpeaker25:40

       Yes.

   18. SPSaswat Panigrahi25:40

       So, that kind of learning, uh, is, is built in into multiple layers of the stack as well. But also, some general things that the AI surprises you. You asked about fine-tuning. One of the powerful things that deep models are telling us, as well as generative AI is telling us, is that you actually don't need to hand-tune every single thing.

   19. SPSpeaker25:58

       Okay.

   20. SPSaswat Panigrahi25:58

       It learns.

   21. SPSpeaker25:59

       Yeah.

   22. SPSaswat Panigrahi25:59

       Uh, so, so, so, so that kind of learning, um, we are doing. And one thing I will say, you asked about moat a little bit earlier. You know, the con- the high level concept of AI may be easy to understand, but really, the breakthrough engineering that you sometimes need in AI is just having the raw infrastructure to intake all this data. The amount of data you have to learn to handle to build a really-

   23. SPSpeaker26:23

       Mm-hmm

   24. SPSaswat Panigrahi26:23

       ... you know, well-learned al- algorithm is, is pretty hard.

   25. SPSpeaker26:26

       Oh.

   26. SPSaswat Panigrahi26:26

       And that's where Google's infrastructure that-

   27. SPSpeaker26:29

       Ah, that makes sense

   28. SPSaswat Panigrahi26:29

       ... we have worked with-

   29. SPSpeaker26:30

       Yep

   30. SPSaswat Panigrahi26:30

       ... is just, uh, in immense, uh, and, uh, you know, um, machine learning investments that Waymo and Google did 12 years ago, 13 years ago, is beginning to pay off in a manner that's, uh, pretty hard to, you know, just, uh, get there.
10. 31:25 – 37:12

    Collaborating with regulators

    1. SPSaswat Panigrahi31:25

       been operating for a while. There, one thing we did is we took every fatal crash that had occurred and re-simulated and showed that Waymo could avoid that.

    2. SPSpeaker31:33

       Okay.

    3. SPSaswat Panigrahi31:33

       That was one, you know, very specific data set. What we also did is we were the first company ever- Uh, uh, to cross 1 million fully autonomous miles, miles like this one

    4. SPSpeaker31:43

       Yes. Yeah

    5. SPSaswat Panigrahi31:44

       ... nobody in the front seat. So there's no debate about what the simul- you know, car could have done. It was real miles. In that, we dem- uh, published our full crash tests

    6. SPSpeaker31:53

       Okay

    7. SPSaswat Panigrahi31:54

       ... and there was not a single collision with an injury-

    8. SPSpeaker31:56

       Really?

    9. SPSaswat Panigrahi31:56

       ... in 1 million miles.

    10. SPSpeaker31:57

        Not a single one?

    11. SPSaswat Panigrahi31:57

        Not a single one.

    12. SPSpeaker31:58

        Okay.

    13. SPSaswat Panigrahi31:59

        And, um, only two of them would r- you know, meet the standards of a reportability, um, called CISS. And, yeah. All right. You're here.

    14. SPSpeaker32:10

        Oh, we're here.

    15. SPSaswat Panigrahi32:11

        Yeah, and your-

    16. SPSpeaker32:12

        Yeah, it's-

    17. SPSaswat Panigrahi32:12

        Look at that, how it stopped because it saw that they wanted to cross

    18. SPSpeaker32:14

        ... and like, I love seeing it, so many people when they look inside.

    19. SPSaswat Panigrahi32:17

        Yeah

    20. SPSpeaker32:17

        Look at [laughs]

    21. SPSaswat Panigrahi32:18

        Yeah

    22. SPSpeaker32:18

        ... just like, "What's going on there?"

    23. SPSaswat Panigrahi32:19

        Cyclist approaching.

    24. SPSpeaker32:20

        All right.

    25. SPSaswat Panigrahi32:21

        But did you see that?

    26. SPSpeaker32:21

        Oh, it's telling us, yeah.

    27. SPSaswat Panigrahi32:23

        Yeah.

    28. SPSpeaker32:23

        Okay, so I'm gonna click resume ride.

    29. SPSaswat Panigrahi32:26

        And the reason, by the way, uh, we're switching a little bit from safety to user experience design. The reason it told you cyclist approaching is what we know is even when the car is stopped, what happens is people open the door-

    30. SPSpeaker32:35

        Mm
11. 37:12 – 39:48

    Learnings from the first 2m miles

    1. SPSaswat Panigrahi37:12

       of a pedestrian if they were to jump out on this side?

    2. SPSpeaker37:14

       Mm-hmm.

    3. SPSaswat Panigrahi37:14

       We may approach the problem from different angles, but our core mission of safety is so deeply drilled into every Waymonaut that we believe that every person who meets us will see that.

    4. SPSpeaker37:27

       Yeah, and I mean, one aspect I love that, you know, it's ingesting all this data-

    5. SPSaswat Panigrahi37:31

       Yeah

    6. SPSpeaker37:31

       ... from so, like you said, almost 2 million miles now. And when you think about us as human drivers, like no one, like-

    7. SPSaswat Panigrahi37:37

       Nobody

    8. SPSpeaker37:37

       ... I certainly have not driven 2 million miles. I'm not even, you know, really processing exactly what happens when a scooter-

    9. SPSaswat Panigrahi37:43

       Right, right

    10. SPSpeaker37:43

        ... is coming up-

    11. SPSaswat Panigrahi37:43

        Right

    12. SPSpeaker37:44

        ... on the right here. And if I've spent my whole life driving in, driving in SF, which is not the case for me-

    13. SPSaswat Panigrahi37:49

        Yeah

    14. SPSpeaker37:49

        ... but some people that is, and then you drop them in Phoenix-

    15. SPSaswat Panigrahi37:51

        Yeah

    16. SPSpeaker37:51

        ... you actually, you know-

    17. SPSaswat Panigrahi37:52

        Yeah

    18. SPSpeaker37:52

        ... they, they are new to that road-

    19. SPSaswat Panigrahi37:54

        Yeah

    20. SPSpeaker37:54

        ... just like you training-

    21. SPSaswat Panigrahi37:56

        Right, right

    22. SPSpeaker37:56

        ... in a new city.

    23. SPSaswat Panigrahi37:58

        And by the way, the 2 million thing that you mentioned, that's just the fully autonomous miles.

    24. SPSpeaker38:01

        Mm-hmm. Not the-

    25. SPSaswat Panigrahi38:02

        We have billions-

    26. SPSpeaker38:02

        Billions in simulation

    27. SPSaswat Panigrahi38:03

        ... in simulation.

    28. SPSpeaker38:03

        Yeah.

    29. SPSaswat Panigrahi38:04

        So but yeah, here there was a lot of experimentation on how much is the appropriate amount of detail. So, uh, there are folks who bu- the vehicle and the sensors are seeing a lot more detail than what's-

    30. SPSpeaker38:16

        Yes
12. 39:48 – 44:16

    Driving user retention

    1. SPSaswat Panigrahi39:48

       know, too bright.

    2. SPSpeaker39:49

       Yeah.

    3. SPSaswat Panigrahi39:49

       'Cause if you want to just take a nap, we want you to take a nap.

    4. SPSpeaker39:52

       [laughs] Well, I mean, that's, that's a great point, and I just, I'm curious if there's other learnings from ... You know, again, you've rolled this out. The number one thing is safety. But then from there it's like-

    5. SPSaswat Panigrahi40:02

       Yeah

    6. SPSpeaker40:02

       ... how do you create a great product that people-

    7. SPSaswat Panigrahi40:04

       Oh, yeah

    8. SPSpeaker40:04

       ... wanna engage with, that people wanna come back to?

    9. SPSaswat Panigrahi40:06

       Right.

    10. SPSpeaker40:07

        So it's not just a novelty, right? Where you're like, "Oh, you know, I sat in an-

    11. SPSaswat Panigrahi40:10

        Yeah

    12. SPSpeaker40:10

        ... autonomous vehicle." Where they wanna use this for their commute every day or-

    13. SPSaswat Panigrahi40:13

        Yeah

    14. SPSpeaker40:13

        ... you know, for, for their, their daily lives.

    15. SPSaswat Panigrahi40:16

        Yeah.

    16. SPSpeaker40:16

        So how do you think about that other than maybe the screen? Are there other things-

    17. SPSaswat Panigrahi40:19

        Oh, yeah

    18. SPSpeaker40:20

        ... within the car?

    19. SPSaswat Panigrahi40:20

        Thousands. Thousands. And by the way, uh, speaking of modes, that's another powerful flywheel advantage, right? If you have served the first 10,000 humans who have been in a-

    20. SPSpeaker40:31

        Mm-hmm

    21. SPSaswat Panigrahi40:31

        ... fully autonomous car, the feedback they give you and the time you have to incorporate that into your learnings-

    22. SPSpeaker40:35

        Yep

    23. SPSaswat Panigrahi40:36

        ... is, is a great positive flywheel approaching, you know, 100, uh, uh, thousand tr- fully autonomous rides in a month. So that feedback does help. I'll give you just a couple of examples. We could speak for hours just on that part.

    24. SPSpeaker40:49

        Mm-hmm.

    25. SPSaswat Panigrahi40:49

        One example I'll tell you is imagine a residential street like this, this one that we're passing.

    26. SPSpeaker40:55

        Yes.

    27. SPSaswat Panigrahi40:55

        Remember when we were getting in the car, you were asking, "Hmm, which side of the street should I be on?"

    28. SPSpeaker40:59

        Yep.

    29. SPSaswat Panigrahi40:59

        Now imagine you're just getting out of home-

    30. SPSpeaker41:01

        Yep
13. 44:16 – 47:55

    Waymo’s expansion strategy

    1. SPSaswat Panigrahi44:16

       of emotions. That's okay.

    2. SPSpeaker44:17

       Yeah.

    3. SPSaswat Panigrahi44:17

       What, the physical demonstration of-

    4. SPSpeaker44:20

       Mm

    5. SPSaswat Panigrahi44:20

       ... you know, turning in or, like that example that we saw right now of somebody who ran a red light-

    6. SPSpeaker44:25

       Yes

    7. SPSaswat Panigrahi44:25

       ... if you recall.

    8. SPSpeaker44:25

       Yeah.

    9. SPSaswat Panigrahi44:26

       Those are the ones that are truly, you know, uh, could result in danger.

    10. SPSpeaker44:30

        Mm-hmm.

    11. SPSaswat Panigrahi44:30

        Which is what the vehicle positions it. It, it thinks about many small details. Where is it positioning itself? How much gap is it leaving? What's its velocity? So that it has the greatest optionality if somebody were to behave re- recklessly.

    12. SPSpeaker44:42

        Okay.

    13. SPSaswat Panigrahi44:42

        So when we are crossing that sign, we're not assuming that everybody's gonna obey the light. We're trying to monitor, hmm, their light is red-

    14. SPSpeaker44:50

        Okay

    15. SPSaswat Panigrahi44:51

        ... which means they should be slowing down. Why aren't they slowing down? That's an anomaly. Let's prepare for this anomaly. Let's protect. Let's be defensive against that anomaly.

    16. SPSpeaker44:59

        Yes. Like, something that comes to mind is just as a rider-

    17. SPSaswat Panigrahi45:01

        Yeah

    18. SPSpeaker45:01

        ... you wanting to feel confident in, in the car, and seeing it be a little more assertive-

    19. SPSaswat Panigrahi45:06

        Yeah

    20. SPSpeaker45:06

        ... is actually really reassuring at points, like when it makes sense.

    21. SPSaswat Panigrahi45:09

        Yeah.

    22. SPSpeaker45:09

        Because to your point, if it's constantly stopping, if it's-

    23. SPSaswat Panigrahi45:11

        Yeah, yeah, yeah

    24. SPSpeaker45:11

        ... constantly pulling over, then I don't have confidence that this thing-

    25. SPSaswat Panigrahi45:14

        Right, right

    26. SPSpeaker45:14

        ... is gonna know what to do in a kind of real-

    27. SPSaswat Panigrahi45:17

        Yeah, and, and you have a busy life. You wanna get where you're going, right? So but by the way, iconic place coming up.

    28. SPSpeaker45:23

        Oh, yeah.

    29. SPSaswat Panigrahi45:23

        Right?

    30. SPSpeaker45:23

        Are we at the Painted Ladies?
14. 47:55 – 48:42

    Changing regulation

    1. SPSpeaker47:55

       Mm.

    2. SPSaswat Panigrahi47:55

       We have tested in Death Valley for extreme temperatures. We have, uh, uh, tested in Tahoe for snow. So there was a testing phase in which we went to 20 cities just to make sure with enough of a diverse data set to be building off of. Look at that gentle, because this car was coming.

    3. SPSpeaker48:08

       Mm.

    4. SPSaswat Panigrahi48:08

       That tram was coming. Um, anyhow, it's beauty to see-

    5. SPSpeaker48:13

       I know. I love seeing it

    6. SPSaswat Panigrahi48:14

       ... every second ride out.

    7. SPSpeaker48:15

       It really is something that it's, it's like I'm watching every little thing that is happening.

    8. SPSaswat Panigrahi48:17

       And you can see this pedestrian crossing, the car going, another one approaching, but green light, so turned right after the pedestrian went off. Anyway, sorry. I, I, um-

    9. SPSpeaker48:26

       [laughs]

    10. SPSaswat Panigrahi48:27

        You were asking me, "Does it get boring for you after working?"

    11. SPSpeaker48:29

        Hopefully.

    12. SPSaswat Panigrahi48:30

        No, I love it.

    13. SPSpeaker48:30

        I love seeing your excitement because, I mean, this is my first time, so I really am, like, as they say, like, taking it all in.

    14. SPSaswat Panigrahi48:36

        Yeah.

    15. SPSpeaker48:36

        But it's cool. You've-

    16. SPSaswat Panigrahi48:37

        Yeah.

    17. SPSpeaker48:37

        How many times do you think you've been-

    18. SPSaswat Panigrahi48:39

        Oh

    19. SPSpeaker48:39

        ... in a-

    20. SPSaswat Panigrahi48:39

        Can't... Yeah, like just the first year in 2019,

15. 48:42 – 53:47

    Societal unlocks enabled by autonomy

    1. SPSaswat Panigrahi48:42

       I spent a ton of hours in it. Uh-

    2. SPSpeaker48:44

       Actually, were you scared at all when you were first testing it? Because now, I mean, I guess I've seen some of the data. I've seen these on the road.

    3. SPSaswat Panigrahi48:52

       Yeah.

    4. SPSpeaker48:53

       So there's a level of like, oh, I, I know these work.

    5. SPSaswat Panigrahi48:56

       Yeah.

    6. SPSpeaker48:56

       Um, but when you were first getting into the vehicle, was there any, like, fear, apprehension?

    7. SPSaswat Panigrahi49:04

       Um, not fear. You know, uh, the closest feeling I can describe is when you have prepared for an exam for-

    8. SPSpeaker49:10

       Mm

    9. SPSaswat Panigrahi49:10

       ... six years. You know, where whatever is the biggest exam you've, uh-

    10. SPSpeaker49:13

        Yeah

    11. SPSaswat Panigrahi49:13

        ... given, the f- little feeling that you got. You know that you have prepared as best as you possibly-

    12. SPSpeaker49:18

        Yeah

    13. SPSaswat Panigrahi49:18

        ... could for it. You left nothing on the table, but it is exam day. [laughs]

    14. SPSpeaker49:22

        You're like, "Oh, gosh, I hope I can show up," or-

    15. SPSaswat Panigrahi49:23

        Right, right, right

    16. SPSpeaker49:24

        ... "I hope this works." Well, it's funny because a lot of people, you know, they'll see the stats, and they'll be like, "That's the average human," but they think, for whatever reason, they outperform the average human on the road. But I'm the opposite.

    17. SPSaswat Panigrahi49:35

        Yeah.

    18. SPSpeaker49:35

        I'm like you, where I'm like, "I want this. I wa-" [laughs]

    19. SPSaswat Panigrahi49:37

        Yeah.

    20. SPSpeaker49:37

        I am not a great driver.

    21. SPSaswat Panigrahi49:38

        And by the way, the, the, the, you know, the fundamental attribution error, by the way, is that-

    22. SPSpeaker49:42

        Yeah

    23. SPSaswat Panigrahi49:42

        ... uh, you know- More than 50% of people think they're better than average, [laughs]

    24. SPSpeaker49:46

        Yeah.

    25. SPSaswat Panigrahi49:46

        Which is not supposed to be feasible, right?

    26. SPSpeaker49:48

        Exactly.

    27. SPSaswat Panigrahi49:48

        Uh, so, no, I, I... And, and, um, so maybe on the city selection, just to close that out, so yeah, LA, uh, 2 million plus town.

    28. SPSpeaker49:55

        Mm.

    29. SPSaswat Panigrahi49:56

        A huge diversity of use cases. Everything from commute trips all the way to, you know, uh, sports events, to-

    30. SPSpeaker50:03

        Yep
16. 53:47 – 54:26

    Will self-driving cars replace humans?

    1. SPSaswat Panigrahi53:47

       kinds of, uh, uh, you know, for society at large. It- it's a pandemic almost, right?

    2. SPSpeaker53:51

       Mm. Mm-hmm.

    3. SPSaswat Panigrahi53:51

       And, and this is a, a, you know, antidote to that. And, uh, then there are classes of individuals for whom this freedom is not available.

    4. SPSpeaker54:00

       Mm-hmm.

    5. SPSaswat Panigrahi54:00

       Above 65, you know, many folks can't drive anymore or, or drive while taking risk or would have the freedom of m- mobility. There are, uh, folks with visual impairments.

    6. SPSpeaker54:12

       Yes.

    7. SPSaswat Panigrahi54:13

       Then, uh, vulnerable populations being able to take jobs that they otherwise wouldn't have been.

    8. SPSpeaker54:17

       Mm-hmm.

    9. SPSaswat Panigrahi54:17

       That nighttime use case that we said.

    10. SPSpeaker54:19

        Yeah.

    11. SPSaswat Panigrahi54:19

        Giving economic opportunity, uh, so which they simply would not be comfortable, or, or you shouldn't ask somebody to

17. 54:26 – 1:00:11

    Closing thoughts

    1. SPSaswat Panigrahi54:26

       own a car before they can take their entry level job. Like that's a, you know-

    2. SPSpeaker54:30

       I mean-

    3. SPSaswat Panigrahi54:31

       ... challenging economic, uh, catch 22, right?

    4. SPSpeaker54:34

       [laughs]

    5. SPSaswat Panigrahi54:34

       Uh, so there's that. Then the third is, yes, how much, uh, you know, pollution idling cars in city centers are causing.

    6. SPSpeaker54:41

       Mm-hmm.

    7. SPSaswat Panigrahi54:41

       How much, uh, city real estate is being lost to idle assets. Uh, so yeah, layers and layers upon that, and yes, that's just in passenger vehicles. Then you, uh, consider in-city delivery, then you consider, uh, long haul trucking, where already tremendous amount of economic opportunity in the United States is being lost to lack of drivers. And by the way, when we do have drivers, we have mandatory breaks because fatigue is a real thing.

    8. SPSpeaker55:07

       We recently moved.

    9. SPSaswat Panigrahi55:08

       Yeah.

    10. SPSpeaker55:08

        I drove from near San Diego up to San Francisco.

    11. SPSaswat Panigrahi55:11

        Right.

    12. SPSpeaker55:11

        And we happened to do it at night, and-

    13. SPSaswat Panigrahi55:12

        Yeah

    14. SPSpeaker55:13

        ... just the number of truck drivers that, you know, we, we saw on the routes, I just, I knew this already-

    15. SPSaswat Panigrahi55:19

        Yeah

    16. SPSpeaker55:19

        ... but seeing it, I was like, "This, something feels wrong here." And I just imagine a future where, you know, some people might not like this, but where that's automated. That sounds beautiful to me.

    17. SPSaswat Panigrahi55:28

        Yeah. And, and there's a be- uh, you know, beautiful transition points as well, in the sense that truck driving could become a local job, which would be powerful in many, many different ways in the sense that-

    18. SPSpeaker55:40

        Mm-hmm. Oh.

    19. SPSaswat Panigrahi55:41

        Oh, you see-

    20. SPSpeaker55:41

        I see it sensed the pylon.

    21. SPSaswat Panigrahi55:43

        Yeah. And you see that?

    22. SPSpeaker55:44

        Oh my gosh, that's incred- 'cause see, I was wondering actually if it could tell-

    23. SPSaswat Panigrahi55:47

        Yeah

    24. SPSpeaker55:48

        ... if that's a hu- So they can tell that those are humans.

    25. SPSaswat Panigrahi55:50

        Yeah.

    26. SSSteph Smith55:50

        But then these are pylons

    27. SPSaswat Panigrahi55:51

        These are pylons and you-

    28. SSSteph Smith55:52

        I can sense every single one

    29. SPSaswat Panigrahi55:53

        And there was a ha- there was like a roughly written keep right and an assured going there.

    30. SSSteph Smith55:58

        Yeah, 'cause you could imagine how it might think to go left.

Episode duration: 1:00:21