Wayve combines road data, OEMs, and simulation

Transcript

Transcribed by machine (Whisper) from the episode’s audio, so a word may be off here and there.

1. 8:21

   Yeah, let's talk about demos. Let's talk about data. So I'm going to jump to one to just highlight that last point. This is in the center blog, the validation blog. So we'll come back to some of these videos. But just speaking to the diversity, here is a very simple scene that you can see anywhere in the world. And it's stopping at a stoplight. Now, the way that those stoplights look in the US and Germany is slightly different. The nuanced behavior for how far you might stop behind a stop line is not only geography dependent, but it's also local rules dependent, style dependent.

2. 9:00

   And culturally, you know, they say it's a California stop compared to... Yeah. And this really gets back to sort of the overall thrust of what we're putting out there and we're trying to say, which is it is literally impossible to write rules for how this should work. We didn't learn to drive by digesting. We digested, well, it depends on where you grew up, but like we digested a rules book that said okay here's you know the driver test and we took it but ever since then when we were 16 or 14 or whatever we've learned we've extrapolated from those rules we've taken trips to countries where you drive on the other side of the road and we've become better drivers and that's exactly how we view the driving problem um for uh wave i want to um highlight one other great example in here where um i think it might be in the first post here we go okay so um

3. 9:50

   So this example is about many things, but I just want to highlight one thing that is kind of a little bit different, but also a little bit the same throughout the world, which is construction scenes. So I think when you look at these scenes, you probably have a sense of what to do from just looking at the image. Okay, it's like there's a light. I'm not that great of a driver, so honestly, I look at these scenes, I'm like, thank God I'm not driving in London or Japan right now.

4. 10:15

   well this is this is why we say that the the driver should be better than better than humans because i i was actually looking at this image on the left i'm like do i do i actually know what to do here so um the point is that construction scenes are all a little bit different you have to understand language you have to understand that language both from a system one thinking level um reactionary but also a system two thinking level which is like long-term strategy so like it's not just enough to parse the sign but you also need to understand like what are the consequences of parsing that sign how do i make sure i'm not driving over here into this like left region um and the and then if you look at the right you know there's there's signs in um in uh in different characters uh there's a person there that's like maybe it doesn't even look like

5. 11:01

   he's necessarily trying to tell you what to do so this is an example of the diversity of data that we have. And the point is that if we learn how to drive in situations like this in Japan on the right and this in the UK on the left, we're going to make a generally better driver that works worldwide. How do you gather all of this data? We have a fantastic operations team that thinks about that night and day, 24-7, 365 globally. So that's like pillar one is our operations team.

6. 11:31

   Now, Wave is decidedly not a hardware and operations business. It's a very hard business to be in. People who have been working on self-driving, especially RoboTaxi Forever, have massive operations footprints. Kudos to them for how they've built that out. But that's not how Wave is planning on solving the problem. We're planning on solving the problem by having a minimal internal operations footprint purely for the purpose of geographic and vehicle diversity data collection.

7. 11:58

   and also... High-fidelity human reference data. So we have expert drivers. We see how they drive. However, we want to primarily generate data with simulation over time. Maybe not primarily, but over time. That's one aspect of getting data. Two is partnering with our OEMs, who have a much richer set of data than we do. And three is purchasing data. So there's a rich set of data out there that we can go license or purchase from a variety of vendors. Dashcam data, for example.

8. 12:28

   We have a huge corpus of crazy stuff that's happened on highways, right? So if we can get that and we can incorporate something that happened on a highway 10 years ago into our corpus, then we can build a better driver.