00:03it's felt like this is coming for a
00:05while like I almost feel like you know
00:06the future is here like we're seeing
00:08this we're sitting in a a car with no
00:09driver and so maybe you could kind of
00:11map out where we are in that kind of Arc
00:15the the five levels of autonomy and also
00:18you know where we still have left to go
00:21yeah certainly we are in level four
00:23right now so fully autonomous uh as you
00:25can see nobody on the front seat no
00:27expectation of a human to take over and
00:30um you know in level two and three it's
00:32really really crucial to communicate the
00:35expectations to the driver cuz it's very
00:37easy that during a normal situation the
00:39driver feels H the car is kind of
00:40driving well so I can you know pick up a
00:42book and start reading no that's that's
00:44serious uh and there is an expectation
00:46take over so we are in that level four
00:48with a certain scope so right now if we
00:51were to begin heavily snowing in SF
00:54which it has Believe It or Not last
00:56season there was a little bit of snow uh
00:58we wouldn't operate could but level five
01:00is truly uh defined as anywhere anytime
01:03right I mean and even the idea of
01:05autonomy people kind of view in this
01:06binary way right like the whole idea of
01:08levels kind of puts that into
01:09perspective like level two is like Lane
01:11sensing automatic braking is level three
01:14where basically what we're doing now but
01:16there would be a human in the front or
01:18how how would you define no no I I would
01:20say a vast difference even between level
01:22three and level four huge uh it's almost
01:25like difference between um driving and
01:27flying I would say Mass difference
01:30because um this concept still that
01:33within seconds you need to take over
01:35versus now we can have this conversation
01:38look at in in uh that that would be
01:40challenging right but the Assurance
01:42level you need to get to for level four
01:44is just a univers different I would say
01:46okay so tell me a little bit more about
01:48you know the barriers along the way to
01:51level four um what would you say was is
01:53was it the technology the regulations
01:54some combination are there major factors
01:57that have really delayed us the largest
01:59challenge I would say was largely I
02:02would call it technology but I mean two
02:04different things one is building the
02:05driver itself that can drive uh under
02:08these conditions and having the high
02:10grade of uh you know performance but
02:12also measuring that is pretty hard this
02:14smooth sort of early stage drive under
02:17very tight uh constraints is now
02:20relatively with today's technology not
02:22terribly hard to build but to be able to
02:24do that at the scale uh you know 24/7
02:28mhm busy intersection
02:30at slow speed but also high speed
02:32intersections of Phoenix so in Phoenix
02:34for example the streets are wider so you
02:37don't deal with these narrow situations
02:39y but the driving speed is 45 people are
02:43sometimes going 60 on that that means
02:44you got to see a lot further so very
02:47different uh sets of challenges very
02:48diverse ones uh and the technology and
02:51it really required the full stack right
02:53we build the hardware we build the
02:55software cuz if you built just the
02:57software and waited for somebody else to
02:59deliver the the hardware the speed of
03:01learning the speed of iteration that was
03:03necessary to build something like this
03:04was so steep that it was not feasible so
03:07we had to build the lasers the cameras
03:08The Radars and the software on top let's
03:11talk about safety right I mean that is
03:13like the foundational piece of whether
03:14we can get these cars on the road how
03:17would you say you know between The
03:19Regulators the technologists like
03:22yourself who are building this and then
03:24the consumers the people the Riders how
03:26do each of them view this like concept
03:29of safety how are you designing the
03:31product with that in mind uh the
03:34analytical and quantitative version of
03:36safety you can say you know the nature
03:38of collisions you have uh the you know U
03:42probability of entering into a certain
03:43Collision under certain circumstances
03:45lot of complications in there there's
03:47the element of risky Behavior you may
03:49have gotten lucky that you didn't get
03:50into a collision but you undertook risky
03:53behavior and third and finally do you
03:55make the person feel safe so you know we
03:58could be breaking hard anytime we sensor
04:02risk yeah uh or we could feel very safe
04:05exactly you wouldn't feel very safe and
04:07I think um and then the fourth layer I
04:10would add on top of that is just because
04:12you figured out how to drive smoothly
04:14meeting the expectations of a rider you
04:16shouldn't falsely promise the true
04:20analytical safety either so for example
04:23you know designing an algorithm whereby
04:25you drive smooth on a street could be
04:27easy but you shouldn't overpromise that
04:29you can detect a pedestrian jumping out
04:31of a car that takes real you know heavy
04:34engineering that um consumer May begin
04:39believing that oh just because it can
04:40drive smoothly it can probably protect
04:42me from that and and I think being
04:44truthful about your capabilities is
04:47really really important what does the
04:48data say though can you kind of ground
04:50Us in there is a certain number of
04:52crashes that humans engage with every
04:55single year and then where the
04:57technology is at relative to that I'll
04:59give you just two examples from many
05:01safety methodologies we employ uh East
05:04Valley of Phoenix MH we have um been
05:07operating for a while there one thing we
05:09did is we took every fatal crash that
05:11had occurred and resimulated and showed
05:14that weo could avoid that okay that was
05:16one you know very specific data set what
05:18we also did is we were the first company
05:21ever uh uh to cross one million fully
05:24autonomous miles miles like this yes
05:26nobody in the front seet so there's no
05:28debate about what the simil you know car
05:30could have done it was real mils in that
05:33we published our full crash tats okay
05:36and there was not a single collision
05:37with injury in 1 million months sing not
05:40a single one okay and um only two of
05:43them would you know meet the standards
05:45of a reportability um called cisss so
05:48this element of safety it's not just
05:50about when it's driving but it's about
05:51caring about safety every