00:00we are in San Francisco we're not taking
00:03you to some you know little desk
00:06facility in your desert to show you that
00:11the vibrant cities on the planet that's
00:13where we are in the journey we're fully
00:15autonomous where we need to be
00:17self-driving is here I think if I was a
00:20human driving there like I don't know
00:21what would have happened yeah you were
00:23asking me it doesn't get boring for you
00:24afterwards okay I love seeing your
00:26excitement and we are not the only ones
00:28who have noticed that kid's just
00:30noticing that we don't have a driver
00:32I love seeing so many people when they
00:35look inside look at that they're just
00:36like what's going on there today I got
00:38to ride fully autonomously with waymo's
00:41Chief product officer saswat hanigrahi
00:44who's actually been on this journey
00:45since 2016. we discuss so much including
00:50the five levels of autonomy fully
00:52autonomous as you can see nobody in the
00:54front seat no expectation of a human to
00:56take over the infamous lidar vs video
00:59debate saying you love lidars and hate
01:03cameras or vice versa is saying you love
01:05one wavelength versus the other
01:06wavelength right it's not a fundamental
01:09thing regulation user experience the
01:11role of AI and all this but especially
01:14the question if autonomy is truly here
01:23as a reminder the content here is for
01:25informational purposes only should not
01:27be taken as legal business tax or
01:29investment advice or be used to evaluate
01:31any investment or security and is not
01:33directed at any investors or potential
01:35investors in any a16z fund please note
01:38that a16z and its Affiliates may also
01:40maintain investments in the companies
01:42discussed in this podcast for more
01:44details including a link to our
01:46investments please see a16c.com
01:59all right before we jump in I just
02:01wanted to say that we took a 60 Minute
02:03Journey around San Francisco and we saw
02:06just about everything we saw a
02:09construction zone we saw someone running
02:11a red light we literally ran into this
02:14group of bicyclists probably like 20 to
02:1630 of them we even saw what I think was
02:19a two-year-old kid who could just tell
02:22they could just tell that something
02:23about the front seat seemed off because
02:26there was no driver so I hope you
02:28enjoyed this episode this really was
02:30such a special experience for me I'll
02:33tell you more at the end of this episode
02:34why that is but I hope you enjoyed this
02:37episode I hope you enjoyed this ride as
02:44very clean so what do I do do I just
02:49oh yeah the KB up there nice oh
02:53I like don't even know how to interact
02:57all right we just jump in
02:59hey there Catherine thank you
03:09we just click Start yeah do you wanna go
03:22this is cool so you can play music you
03:25can and you can just ask it to pull over
03:27if you want anytime at any time and you
03:29can call support and there was an audio
03:30cue to remind you in case folks get in
03:33and they you know into a minute think
03:35like this is somebody but it turned out
03:39actually that wasn't uh you know folks
03:41just got eased into it pretty quick you
03:42know somehow uh this the smoothness
03:45which which it drives yeah it's an
03:47instant trust totally so I mean so I get
03:51carsick pretty easily but to your point
03:52like it doesn't feel jerky at all right
03:55um it feels really smooth but I mean I'm
03:57so curious because you've been working
03:58in this space for so long does knowing
04:01kind of like how the sausage is made
04:03does that make it any less magical like
04:06when you got into this car for the first
04:07time and there was no driver sitting
04:09there yeah were you also like oh my gosh
04:11like what's going on here oh absolutely
04:13I'm totally like a kid on this one so
04:15you know it was um uh 2017 was probably
04:19the first time there was a person in the
04:21front there but they didn't have actual
04:23controls so that was my first yeah 2019
04:25was the first time I was in a car with
04:27truly nobody on a public Street uh just
04:29driving around yeah and every single
04:32time even if of course you know the year
04:34before the month before the day before
04:36and the hour before I was emancipation
04:38and preparing for it yeah still being
04:40inside it it's it's truly special I mean
04:43it's so cool to just see it navigating
04:44yeah the slight turns on the wheel
04:46something I have to ask about is it's
04:49felt like this is coming for a while
04:50like I almost feel like you know the
04:52future is here like we're seeing this
04:53we're sitting in a car with no driver
04:55but a lot of people would kind of say
04:59this has been promised every two years
05:00kind of thing and it's like we're
05:02finally at the two years in a way and so
05:04maybe you could kind of map out where we
05:07are in that kind of Arc the five levels
05:09of autonomy yeah and also you know where
05:13we still have left to go yeah certainly
05:15we are in level four right now so fully
05:18autonomous uh as you can see nobody in
05:20the front seat no expectation of a human
05:24um you know in level two and three it's
05:26really really crucial to communicate the
05:28expectations to the driver yeah because
05:30it's very easy that during a normal
05:32situation the driver feels the car is
05:34kind of driving well so I can you know
05:36pick up a book and start reading no yeah
05:38that's serious uh and there is an
05:40expectation to take over so we are in
05:41that level four with a certain scope
05:44so right now if you were to begin
05:47which it has I believe it or not last
05:50season there was a little bit of snow uh
05:52we wouldn't talk we couldn't but level
05:54five is truly uh defined as anywhere
05:56anytime right I mean and even the idea
05:58of autonomy people kind of view in this
06:00binary way right like the whole idea of
06:02levels kind of puts that into
06:03perspective like level two is like Lane
06:05sensing automatic braking right is level
06:07three where basically what we're doing
06:09now but there would be a human in the
06:11front or how how would you no no I would
06:13say a vast difference even between level
06:16three and level four huge uh it's almost
06:18like difference between
06:21um driving and flying I would say you
06:22know it's a massive difference because
06:25um this concept still that within
06:27seconds you need to take over versus now
06:30we can have this conversation right in
06:33uh that would be challenging right but
06:35the Assurance level you need to get to
06:37for level four is just a universe
06:39different I would say and if
06:41too like if I if I let's just
06:45down a hill reason and feels like the
06:46car is not stopping like could I take
06:48this over if I wanted to uh so if you
06:51did the car would say hmm I'm being
06:53interfered with I'm a fully autonomous
06:56car I'm not supposed to be interfered
06:58with in this manner yeah so it'll pull
07:00over basically got it so we've talked
07:03about the kind of Arc of innovation
07:04you've also worked in this industry for
07:06a long time since 2016 is that correct
07:09that's right that's right okay so tell
07:10me a little bit more about you know the
07:12barriers along the way to level four and
07:15what would you say what is was it the
07:17technology the regulation some
07:18combination are there major factors that
07:20have really delayed us from getting to
07:22this point where we're now at level four
07:24yeah yeah and the one thing you know you
07:27mentioned the Innovation Arc I always
07:28reflect that you know on any Journey
07:31when you're trying to do something that
07:33has never been done before there will
07:35always be ups and downs new challenge
07:36some challenges you for saw some you
07:39didn't yeah the key is are you clear on
07:42how massive the you know price or the
07:45benefit to societies on the other side
07:47of it because then that makes it all
07:49worth it so we were super clear from the
07:51get-go that a fully autonomous car that
07:54does not get browsy uh that uh you know
07:57we even have physical constraints right
07:58so even if even when you are alert uh
08:01let's say you were looking for parking
08:02on this side of the street your face
08:05would be turning towards that constantly
08:06looking for that parking spot so you
08:09wouldn't be able to see you have
08:10yourself yeah yeah exactly you can't
08:11keep turning back and forth uh and so we
08:14we're fundamentally convinced that a
08:16fully Artemis driver is going to be
08:18safer so once we started that yes the
08:22largest challenge I would say was
08:24largely I would call it technology but I
08:27mean two different things one is
08:28building the driver itself that can
08:30drive under these conditions and having
08:32the high grade of uh you know
08:34performance we're also measuring that is
08:36pretty hard this smooth sort of early
08:38stage drive under very tight uh
08:41constraints is now relatively with
08:44today's technology not terribly hard to
08:46build but to be able to do that at the
08:49scale uh you know 24 7
08:51busy intersections at slow speed but
08:54also high speed intersections of Phoenix
08:56so in Phoenix for example the streets
08:59are wider so you don't deal with these
09:01narrow situations yep but the driving
09:04speed is 45 people are sometimes going
09:0660 on that that means you got to see a
09:08lot further it's a very different sets
09:11of challenges very diverse ones yeah and
09:13the technology and it really required
09:15the full stack right we built the
09:17hardware we build the software because
09:19if you built just the software and
09:21waited for somebody else to deliver the
09:23hardware the speed of learning the speed
09:25of iteration that was necessary to build
09:26something like this was so steep yeah it
09:29was not feasible so we had to build the
09:31lasers the cameras The Radars and the
09:33software on top and the massive
09:35simulation infrastructure I was gonna
09:36say there's so many moving parts and
09:38actually maybe let's talk about that
09:40technology like so this is my first time
09:42in a fully autonomous vehicle but I've
09:44seen them around my area they're driving
09:46around I see no driver but I also see
09:48the swirly thing on top I see a bunch of
09:51different kind of appendages to the car
09:53so maybe you could just break that down
09:55like what is happening how is all this
09:57technology coming together and what are
09:59the bits and pieces that you've added
10:01onto the car that allow it to be
10:03autonomous absolutely
10:05um so fundamentally you can think of it
10:08um is the car aware of what's happening
10:10around it and then can it anticipate
10:14what the things around it are gonna do
10:15yeah and then reasoning on what it
10:18should these are sort of the three
10:19components in perceiving what's around
10:22us think of the example we're just
10:24discussing you're trying to look for a
10:26parking so you're focused on that task
10:28this car with those appendages as you
10:31mentioned can see three football fields
10:34away 360 degree and it's getting a
10:37snapshot multiple times a second okay
10:38right and it's relying on a combination
10:42of the state-of-the-art laser camera The
10:46Radars all strategically positioned okay
10:48so to give you an idea lasers give you a
10:50very precise understanding of everything
10:52around you it's the smallest detail so
10:54if there was a child you know an inch
10:57out of this pole it will be able to mark
10:59oh this is a demarcated uh you know
11:01child away from that pool it'll get to
11:05but you know uh the uh cameras are
11:08needed to distinguish between the red
11:09light and the green light yeah and The
11:12Radars can almost see around corners
11:14even when the laser and the camera or
11:16our human eyes can't because they can
11:18sense uh objects coming in okay and so
11:21we took an approach that we want to
11:23combine the best strengths of each of
11:26these modalities to create the best
11:28picture of what you can see around the
11:30world that you're just just incredibly
11:33better than a human possibly could both
11:35due to the attention span the range the
11:37Fidelity and the combination of these
11:39sensors coming together so that's what
11:41we see but then there's a harder
11:43challenge of anticipating what the
11:44person will do take a look at that
11:46pedestrian they're standing pretty close
11:48to the crosswalk right but you don't
11:49know if they're going to move exactly
11:50are they going to jump in or are they
11:52going to stay are they going to jaywalk
11:54or are they going to obey the light yeah
11:55this requires a deep deep understanding
11:58and tremendous amount of machine
12:00learning the for each stage here by the
12:02way it requires an insane amount of
12:04machine learning and it's not the type
12:05of thing that you can just like put on
12:07the road and say like no oh let's see if
12:08it means yeah we're not trying to tell
12:13it's pretty uh so for example for that
12:15pedestrian in addition to seeing that
12:18they're there acknowledging which was
12:19this problem one you got to look at even
12:22their gate their hand movement their leg
12:24movement to anticipate are they about to
12:26take motion and if you are always
12:28conservative which was you know say four
12:30years ago it was not like we couldn't
12:31detect the British we could totally
12:33detect them it was this Nuance of Are We
12:36being over conservative assuming they
12:37may jump in and so let's not move right
12:39versus confidently moving forward
12:42so are we at our first uh stop of the
12:46multi-stop second opens the door
12:49so yeah that anticipation of what a car
12:51is going to do what a pedestrian is
12:53going to do what a child is going to do
12:54because they can erratically jump
12:55through yeah what a motorcyclist is
12:57going to do are they going to Lane split
12:58and speed and get around you so all
13:01these motion models and understanding of
13:03how people behave how this gentleman is
13:05walking and what their gait and motion
13:07tells you about where they're gonna go
13:08that is the second and third and final
13:10what should the car do this this feeling
13:13that you had that it gently accelerated
13:15but not harshly yes that takes into
13:17account not just look at this gentleman
13:19he is almost has his feet almost on the
13:22Crosser but he's not intending to cross
13:23yeah because he has a stop sign a very
13:26conservative system would just come to a
13:28stop but here we sort of asserted
13:30ourselves a little bit because we went
13:32for it that's that's such a great point
13:35so I learned to drive and
13:37I've live and actually waiting for self
13:38Drive driving and it took a little too
13:41um as part of that oh yeah
13:44absolutely let's go you can see here for
13:47example where they're jaywalking yeah
13:49and you see in this you within the last
13:52minute you saw an example in which we uh
13:54looked and we noticed that this
13:56pedestrian is not going to cross so we
13:57didn't come to an abrupt halt we went
13:59through smoothly and later we just
14:00noticed that somebody's jaywalking and
14:02we'd dealer to them that delicate Nuance
14:04is uh and like this lady you actually
14:06don't know she kind of looks like she's
14:08went back and you can see here we are
14:09tracking them so you know if there was
14:11an anxious passenger you know is this
14:13car seeing it we give them the feedback
14:14that we are we're seeing this pedestrian
14:16crossing right here on the screen we
14:18tell them why we're slowing because what
14:20happens is also people zone out they
14:22take their space they speak to their kid
14:24if they're picking them up after soccer
14:26practice or take a phone call and then
14:28they notice when the car is stopped and
14:29they're like why are we stopped and here
14:31we try to give them the feedback why
14:33we're stopped it's first upside now
14:35or we're healing for this you know truck
14:38that uh passed by so I want to get to
14:40the Comfort like how do you pass the
14:41autonomy evolve this product because it
14:45um but first I feel like you know the
14:46one the Rivalry in self-driving is lidar
14:49versus you know Tesla Vision or the
14:51video processing how has Wayman thought
14:53about that decision of what you know
14:55some people pose it as expensive
14:57Hardware more simple software because
14:59you have so much Fidelity from lidar
15:01versus a bunch of cameras and just like
15:04a lot heavier processing so you know
15:07what went into that decision and also
15:08how are you thinking about that moving
15:10forward like is there a future where
15:11maybe actually you don't need all of the
15:14same sensing systems yeah a great
15:16question so you know personally I have
15:18found the you know lighter and video
15:20debate almost takes like an ideological
15:22sense you know for hard problems The
15:25Innovation Arc that we're talking about
15:27the best approach is taking a first
15:29principles approach right without uh you
15:32know neither love nor hate for a
15:34specific technology that it's a
15:36technology you yeah you shouldn't get to
15:38that level of um it's it's a tool and
15:41lidar clearly has strengths that a
15:43camera doesn't example at night time
15:45even the best cameras will have some
15:48challenges and Camera clearly has
15:50strengths that the lidar doesn't you
15:51know the red green example that we
15:53mentioned and similarly radar has
15:55strengths that lidars and cameras don't
15:58so saying you love lidars and hate
16:01cameras or vice versa is saying you love
16:03one wavelength versus the other
16:05wavelength right it's it's
16:07um almost uh you know
16:09um it's it's it's not a fundamental
16:12thing okay but uh there is a practical
16:15question so clear on the first
16:17principles does it does the combination
16:21position you better than individual the
16:23answer is yes we can show you that there
16:25are situations in which camera will be
16:28now the question is a practical economic
16:30one is your ability to bring this public
16:33good to a large number of individuals
16:35hindered by the fact that these things
16:37are expensive yes and lidars you know 20
16:41years ago if somebody told you that all
16:42these Toyotas are or these cars are
16:44going to have radar Shield like no
16:45Radars are expensive guess what every
16:47single you know most cars have Radars
16:49today uh um cameras on cell phones were
16:54a novelty you know the old Nokia phones
16:56would have you'd be like oh who needs a
16:57camera on the phone right you may have
16:59I'm sure now cell phones have better
17:01cameras than dedicated cameras of six
17:03years ago yep ladder is going through
17:05the same transformation right now the
17:07iPhones have lidars right uh the amount
17:10we have been able to cost down these
17:11lidars in the last two years is
17:13incredible okay so you're not concerned
17:15about that and we we you know we had uh
17:18four or five years ago we had that
17:20belief now we have that proof uh because
17:22you know Hardware Generations you know
17:25there's multiple examples outside of
17:27waymo to see something that again chips
17:29are a great example our cameras are a
17:32great example so it would it would be
17:35surprising if you had a hardware that
17:37you were able to package and then with
17:39focused effort you weren't able to so we
17:41had that belief and now we have the
17:43proof right and waymo actually
17:45manufactures lidarb correct and I feel
17:48like waymo in a way has chosen to to
17:50your point like manufacture the hardware
17:52work on its own software simulation
17:54technology give me the thought process
17:56there in terms of you know there's
17:58always this question in business do I
18:00build do I borrow do I buy yeah and um
18:03given that this is a capital intensive
18:06business how do you think about that
18:07where should you Outsource and where
18:09where do you really need that
18:10fundamental technology yourself great
18:12question the first thought we had let's
18:14build this all ourselves that was not
18:15the first thought but we said okay uh
18:18let's see what's the best out there
18:20absolute best even you know
18:24easing a little bit the cost
18:25requirements say we were willing to pay
18:27yeah let's we'll pay for it yeah what's
18:30the best out there and what we found is
18:32that the absolute best lidar out there
18:37um your radar out there was not
18:42the task of autonomous driving okay and
18:47um we had to build it each Hardware
18:49generation we do evaluate that we try to
18:51see okay have Radars
18:54um you know evolve to the point that we
18:55could use something off the shelf so
18:57there that build versus Y is a pretty
19:00practical Choice uh each single time and
19:02as we look forward the question becomes
19:04you know where do you build a moat
19:06because now you're not the only uh
19:08company that has achieved some level of
19:10autonomy and you could imagine like
19:12let's just say a future where we've just
19:14achieved level five in many places with
19:16many companies how do you think about
19:18what differentiates is it really a data
19:20mode based on the amount of training
19:21that a company has been able to do is it
19:24owning like the proprietary lidar that
19:26is just you know 10x cheaper than the
19:28competitor I'm trying to think ahead in
19:30terms of yeah just where where does
19:32value really accrue in that future right
19:35right so first and just you know before
19:37talking about the more just thinking of
19:39the space we're operating in yeah we're
19:41talking about a space with a trillion
19:43you know miles right so it's vast it
19:47um when let's say the first cars were
19:50being built folks uh said hey is there
19:53space for only one provider or two
19:55providers right here when you're talking
19:57about a space of a trillion miles of
20:00today and then you think about what is
20:02the potential value add when you have a
20:04driver that drives itself because partly
20:07if you think about the miles today uh
20:09you know we talk about a truck driver
20:11shortage for example yeah in fact even
20:13those commercial miles are kind of
20:15stunted by the lack of availability of
20:17bus drivers yeah so so really we're
20:19talking about a space that is all cars
20:22trucks and all Transportation when
20:24something is that vast and the number of
20:27autonomy players today are much smaller
20:29than three years ago if you look at it
20:31because it is you know a pretty
20:33challenging problem still so I think the
20:36the universe is different but still a
20:38valid question we do
20:40um we do believe that there's a um you
20:42know knowledge curve so by having driven
20:46uh 20 million plus miles in testing by
20:49having uh done billions of miles of
20:51simulation we become aware of problem
20:54spaces that others may not have
20:56discovered yet and that goes into
20:58Hardware design and software design and
21:00simulation design Hardware design in
21:01particular has long lead times so that
21:03becomes uh you know a flywheel effect
21:06your question about how did you know you
21:08had to build your own laser because by
21:10that time we had already driven 10
21:12million miles and we were like we're
21:14gonna need that things
21:19yeah that person basically ran the red
21:21light it was red for them
21:23and and this is the kind of we can talk
21:25if you and or I were I was gonna say
21:27like I think if I was a human driving
21:29there like I don't know what would have
21:31happened and you would definitely not
21:33have been able to carry a conversation
21:34right uh so I can barely I can't even
21:38you know talk to someone next to me most
21:40of the time when I'm driving
21:41um but okay so talking about the
21:43technology I want to talk about safety
21:45next but first are there any important
21:46technological unlocks that you still see
21:48on the horizon that not just waymo but
21:51the you know the industry of autonomy is
21:53still trying to solve is it really that
21:55cost curve or is there something else
21:56that is still in the way of us really
21:58rolling this out more broadly and look
22:00we are definitely the rate of innovation
22:02even just within way more which I can
22:03speak to most confidently is massive
22:05I'll give you a concrete example
22:08um when we came from Phoenix to SF we
22:12did have some work to do to you know
22:13adapt you know the assertiveness and the
22:15driving here is different from the
22:17Phoenix driving but when we went to Los
22:20the driver worked shockingly good from
22:22the get-go really yeah shockingly so it
22:25was uh uh and now there's a portion of
22:27uh Scottsdale which is uh northeastern
22:30part of our Phoenix territory it's a you
22:32know much denser area lots of
22:34restaurants lots of shops there we were
22:36able to go in like within two months we
22:38just went there we decided we're gonna
22:39open up Scottsdale within two months and
22:41the reason for that is we're truly the
22:43drivers generalizing very well okay and
22:46the concept is pretty intuitive and this
22:48is advancement in AI That's enabling it
22:50but think of a driver that's capable of
22:53this kind of tight traffic navigation
22:55yes lots of pedestrians and cyclists but
22:58low speed of travel that's where we are
22:59right now yes now imagine in Phoenix 45
23:02miles per hour three lane four-lane
23:05streets lots of oncoming traffic and
23:07being able to navigate that
23:09pretty much every good weather city is
23:11like a linear combination of those two
23:13things right so in Los Angeles The Best
23:15of Both Worlds exactly so when you go to
23:17West Hollywood you're much more like a
23:19SF style driving lots of pedestrians
23:21cyclists and so on you go to the LA's uh
23:25you know more faster boulevards it's a
23:27lot like Phoenix so once you have solved
23:29these two you just the AI is just much
23:32more generalizable the second area I
23:34would say you already mentioned cost
23:35down is making the simulator a lot
23:38better I'll give you an example we have
23:40you know billions of miles of simulation
23:42in good weather yeah if you want to test
23:44how would we do in rain imagine being
23:47able to simulate terrain so that you can
23:49take all the learnings in good weather
23:50all the tough situations you encountered
23:52and now test yourself in Rain what if
23:55rain was a complicating Factor on top of
23:57that what if I add a cyclist into that
23:59tough situation more what if all these
24:01combinatorial questions being able to
24:04realistically simulate that that's also
24:06a huge area yeah I mean it's a little
24:08foggy today but I also wonder you know
24:10you talked about Ai and obviously this
24:11is running off of an algorithm that's
24:13been trained on all these miles
24:15is it one algorithm or you know let's
24:18say it is an extremely foggy day it's a
24:19rainy day you're in a new environment is
24:22it a different slightly fine-tuned model
24:24based on different situations or is it
24:27all one aggregate that's just ingesting
24:29all of this information it's
24:31um definitely many many deep uh models
24:34some very general extremely deep
24:36learning models and some specialized
24:38models were to make them really good at
24:40some very hard like for example
24:43understanding uh pedestrian sentient is
24:46such a vast space it's like
24:48understanding humans
24:50takes us a lifetime to understand
24:52ourselves so understanding human
24:54behavior and motion there could be
24:56specific models there could be
24:58end-to-end models on just driving like a
25:01good citizen uh polite to other riders
25:03that can be a more end-to-end model
25:05being you know comfortable to writers
25:08preferences that can be a very you know
25:10android model as well so it's a mix of
25:12this and there's AI at every layer of
25:15the stack from perceiving the world to
25:17predicting other people's behavior to
25:18the driving to the testing so for
25:21example though you asked about fog so
25:23what we try to do is we both observed
25:27how other people drive and fall we're
25:29also try to reason about how well can we
25:31see in fog so if this forward were to
25:35the appropriate thing to do is I can't
25:37see that far so I shouldn't be driving
25:38as fast as I would normally do yes so
25:41that kind of learning is is built in
25:43into multiple layers of the stack as
25:45well but also some general things that
25:47the AI surprises you you asked about
25:49fine tuning one of the powerful things
25:51that deep models are telling us as well
25:53as generative AI is telling us is that
25:55you actually don't need to hand tune
25:57every single thing it works yeah so so
26:00so that kind of learning
26:02um we are doing and one thing I will say
26:04you asked about remote a little bit
26:07you know the con the high level concept
26:10of AI may be easy to understand but
26:12really the Breakthrough engineering that
26:14you sometimes need in AI is just having
26:16the raw infrastructure to intake all
26:18this data the amount of data you have to
26:20learn to handle to build a really you
26:23know well-earned algorithm is pretty
26:26hard and that's where Google's
26:28infrastructure that we have worked with
26:30is just uh in immense uh and the uh you
26:34know machine learning Investments That
26:36waymo and Google did 12 Years over 13
26:39years ago is beginning to pay off in a
26:40manner that's uh pretty hard to you know
26:43just uh get there so that's some more
26:45that's actually a great point because I
26:46mean we're again we're in a space where
26:48you need to react in milliseconds right
26:50and so you need to not just be able to
26:52train this algorithm but to interpret
26:55live yeah and process that information
26:57live let's talk about safety right I
26:59mean that is like the foundational piece
27:01of whether we can get these cars on the
27:03road right right and something I'd love
27:04to hear from you on is how different
27:08parties interpret safety being on the
27:11road you could say just inherently isn't
27:13safe you can get in a crash you can die
27:15people unfortunately that happens every
27:17day so how would you say you know
27:19between The Regulators the technologists
27:23like yourself who are building this and
27:25then the consumers the people the Riders
27:27how do each of them view this like
27:30concept of safety how are you designing
27:32the product with that in mind yeah I
27:35mean I think you you uh definitely
27:37touched on multiple aspects this is the
27:39analytical and quantitative version of
27:42safety you can say you know the nature
27:44of collisions you have the you know
27:47probability of entering into a certain
27:49Collision under certain circumstances a
27:51lot of complications in there and good
27:53news is many you know regulatory bodies
27:56do have their you know teams that study
27:58crashes and you know there are great
28:00databases of crashes and so on there's
28:02the element of risky Behavior you may
28:04have gotten lucky that you didn't get
28:06into a collision but but you undertook
28:08risky Behavior so for example in that
28:11situation we were in before we could see
28:12that those two individuals were gonna
28:15jaywalk yes now you could have argued
28:18that well the car has the right of way
28:19we were driving right no but that's
28:21risky Behavior right it's it's
28:23preventative Behavior so you can't see
28:26that solely with the presence or absence
28:28of collisions were you a good driver and
28:30third and finally do you make the person
28:32feel safe so you know we could be
28:36breaking hard anytime we sense a risk
28:41exactly you wouldn't feel very safe and
28:44I think um and then the fourth layer I
28:47would add on top of that is just because
28:49you figured out how to drive smoothly
28:51meeting the expectations of a rider you
28:54shouldn't falsely promise the true
28:57analytical safety either so for example
29:00you know designing an algorithm whereby
29:02you drive smooth on a street could be
29:04easy but you shouldn't over promise that
29:06you can detect a pedestrian you jumping
29:08out of a car that takes real you know
29:11heavy engineering that um consumer May
29:14begin believing that oh just because it
29:17can drive smoothly it can probably
29:19protect me from that and and I think
29:21being truthful about your capabilities
29:23is really really important and what
29:25waymo has tried to see their reaction
29:26yeah yeah and you can see how tight this
29:29space is struck on the right a car just
29:31went by a pedestrian just crossed and
29:34there's somebody just going in and
29:36noticing that this fire truck didn't
29:37have its sirens on so it didn't try to
29:40so slight change into that situation
29:43which we can test in simulation we would
29:45say hmm right now the politest thing to
29:47do is let that pedestrian pass and you
29:49know wait for this turn but when the
29:51sirens are on it's a different
29:52environment you know so anyhow and
29:55that's another data point right that you
29:56need to take in it's sound it's not just
29:58visual oh yeah that's the sensor we
30:00didn't discuss there are microphones
30:01that can not only hear that there's a
30:03siren but also uh you know point at
30:06where the siren is coming from yeah
30:09um that's fascinating what does the data
30:11say though can you kind of ground Us in
30:13there is a certain number of crashes
30:15that humans engage with every single
30:17year and then where the technology is at
30:20relative to that yeah so actually
30:24um so both the Metrology by which we
30:26evaluate our safety which is a
30:27combination of many many this will be
30:29interesting there's a bunch of uh
30:33look at that it went around a double
30:38yeah look at all of them we were telling
30:40you who are we waiting for see they just
30:44look at that we can detect all of them
30:47we're giving you feedback that we can
30:48yeah you can literally sense every
30:50single one yeah oh and them coming
30:51around the corner yeah yeah oh and we're
30:55because we can see that their handles
30:57are turning leftward so we can
30:59understand that they are likely going to
31:00go that way so we don't have to come to
31:02a stop and so balance between you know
31:05making progress because that's what a
31:08rider expects and not yeah exactly and
31:10Ultra conservative behavior that may or
31:12may not be warranted to your earlier
31:16um so what we did is I'll give you just
31:18two examples from many safety
31:19methodologies we employ uh East Valley
31:23we have been operating for a while there
31:26one thing we did is we took every fatal
31:29crash that had occurred and re-simulated
31:31and showed that WeMo could avoid that
31:33okay that was one you know very specific
31:35data set what we also did is we were the
31:37first company ever uh uh to cross one
31:41million fully Artemis miles miles like
31:43this yes nobody in the front seat so
31:45there's no debate about what the similar
31:47you know car could have done it was real
31:49months in that we published our full
31:52crash stats okay and it was not a single
31:55collision with an injury in one million
31:57months not a single one okay and um only
32:00two of them would you know meet the
32:02standards of reportability
32:07yeah all right you're here oh we're here
32:11yeah and you look at that how it stopped
32:13because it saw that they wanted to cross
32:14and like I love seeing so many people
32:16when they look inside like they're just
32:18like what's going on there all right
32:21telling us yeah okay so I'm gonna click
32:25and the Reason by the way uh we're
32:27switching a little bit from safety to
32:29user experience design the reason I told
32:31you cyclist approaching is what we knew
32:33is even when the car is stopped what
32:34happens is people open the door if the
32:36cycle I mean so this element of safety
32:40it's not just about when it's driving
32:42but it's about caring about safety every
32:45well I I love the kind of
32:48um analogy used about you know even a
32:51driver we only have two eyes they're
32:52right at the you know on the front of
32:53our face if we're looking a certain
32:55direction if we're focused on you know
32:57music or a podcast in the car our senses
32:59are you know as much as we like to
33:01believe as humans we're all special and
33:03we're you know the best drivers we are
33:05variable in ourselves in terms of what
33:07we're focused on even if your two kids
33:09in the back start screaming at each
33:11other or something by the way I know
33:13you're relatively new to us yeah I hope
33:16you're enjoying the view as well you can
33:18buy some iconic locations the city
33:20absolutely I love to think about these
33:22like second third order effects like you
33:24know people taking City tours in these
33:26cars I mean we've basically turned this
33:27into a recording studio which you would
33:29never think of you know in the past I
33:31can take confidential work calls yeah
33:33that was not a thing I couldn't do
33:34before right it's so good such a good
33:36point I saw um I was looking through the
33:38reviews on the app and I saw this one
33:40that I just thought was so you know it's
33:42just you get a sense of the user that
33:45wants to be in a car like this and they
33:47were like they called it basically Uber
33:48and Lyft for introverts and you know
33:51there's just these little things that
33:52you don't think about because again
33:53we're fixated on safety and that is so
33:55foundational but then you really do once
33:58you have covered safety once it becomes
34:00safe in The Eyes Of The Regulators the
34:03it opens up all these doors oh
34:05absolutely you know we work with so many
34:07Founders in nascent spaces AI web3 space
34:11I would say those um three Industries
34:14also have a fair amount of pushback
34:15which autonomous vehicles do as well
34:17right and in many cases rightfully so
34:19especially autonomous vehicles like
34:20we're talking about people's lives on
34:22the line and so have there been any
34:25learnings since you've been working in
34:26this space for a while where you're
34:28trying to meet the safety need but
34:30you're also trying to push regulation
34:32and welcome you know this technology
34:34into different cities yeah yeah uh you
34:37know it may sound like an idealistic
34:39answer but I do believe that
34:42you know there are applications that
34:45have a fundamentally different use case
34:46or a promise and then they are trying to
34:48make sure they're not harming the public
34:50or you know you know or unintended
34:53the reason we exist is to make driving
34:56safer so we have a deep
34:59fundamental alignment with what
35:01sometimes The Regulators are trying to
35:03achieve and now we may have different
35:05inputs we may have different data we may
35:07have been approaching it from a
35:08different angle but such a so I believe
35:11that every at least every conversation I
35:12have had with anybody who is in state
35:15federal or local government or even
35:17outside of regulators just uh you know
35:20firemen local law enforcement we very
35:24quickly in that conversation
35:26I begin to appreciate our team begins to
35:29appreciate and they begin to appreciate
35:30that we're trying to do the same thing
35:32here okay and that is a powerful
35:36Baseline to begin constructive
35:38conversations out of but if for example
35:40the goal was something else and by the
35:43way what about safety right and that
35:46would be a very challenging conversation
35:47we try to say this is what we're trying
35:49to do folks this is what we have
35:50measured this is the left yeah this and
35:53just transparently sharing the data
35:55um safety or is it collisions all right
35:57well in collisions we have
35:59published our accident reconstruction we
36:01have done 20 million miles of testing
36:03billions of miles of simulation and
36:06we're telling you every single uh
36:07contact we have had in one million
36:09Optimus Miles by the way uh this week
36:11we're about to cross 2 million first
36:13company again yeah first company again
36:1524 7 including daytime not filtering out
36:18any of the challenging situations dense
36:20downtowns 24 7 2 million miles more than
36:24160 years of human driving worth of data
36:26we will just share it with the world and
36:29then they can see for themselves that
36:32uh you know a safer driver and um and
36:35and if ever there was an event about
36:37which they asked us we would
36:38transparently share with them yeah so I
36:41think that gives a fundamentally good
36:43basis and I genuinely believe that you
36:46know uh even the word pushback all right
36:49it's almost like internal debates that
36:51way um when we debate you know how
36:53should we design this thing yeah we come
36:55at it from different angles somebody may
36:57see the user's expectation of smoothness
36:59of Drive somebody may see more what's
37:02that Scooter's intent you see how the
37:04scooter swung in from the right
37:06kind of came in between a bus and
37:09themselves whatever their expectations
37:11what's the expectation of a pedestrian
37:13if they were to jump out on this side we
37:15may approach the problem from different
37:16angles but our core mission of safety is
37:20so deeply drilled into every waymonaut
37:23that we believe that every person who
37:25meets us will see that yeah and I mean
37:27one aspect I love that you know it's
37:29ingesting all this data from so like you
37:32said almost 2 million miles now and when
37:34you think about us as human drivers like
37:36no one like I certainly have not driven
37:382 million miles I'm not even you know
37:40really processing exactly what happens
37:42when a scooter is coming up right on the
37:44right here and if I've spent my whole
37:46life dreaming and driving in SF which is
37:48not the case for me but some people that
37:49is and then you drop them in Phoenix you
37:51actually you know they they are new to
37:54that road just like you training right
37:56in a new city and by the way the two
37:58million thing that you mentioned that's
38:00just the fully autonomous miles not the
38:02billions simulation yeah so but yeah
38:05here there was a lot of experimentation
38:07on how much is the appropriate amount of
38:10detail so uh there are folks who will be
38:13the vehicle and the sensors are seeing a
38:15lot more detail than what's being shown
38:17here right we're we're seeing many many
38:19points per square inch of detail here so
38:22we tried to uh experiment with how much
38:25detail we put in here and there are
38:27folks who when in the early days we had
38:30a version here that would show a lot
38:31more detail and they would engage like
38:34this right in fact they would look more
38:36inside than outside the window and keep
38:38cross checking did it see that code did
38:40it see that thing yeah and what we came
38:42to a balance with is we want to put
38:45people at ease and invite them to use
38:47this space and so they you know this
38:49much kind of and it pops up so when the
38:51vehicle stop you will notice it'll try
38:53to explain why it's stopping like is it
38:55a stop sign or is it yielding to a
38:56pedestrian because we realize through
38:58lots of experimentation that that's when
39:00people want to take a look at the screen
39:03okay so for example if you're a go by
39:06yourself we want you to get lost in the
39:09beauty of San Francisco around you we if
39:12you are checking emails it's all right
39:13go go do that and you will take your
39:15head off your phone or from viewing you
39:18know the Painted Ladies when you see why
39:20are we stopped and we will tell you well
39:22there are two pedestrians on the right
39:23there is a car Crossing on the left
39:25there's a you know car parking in the
39:28front we'll explain that to you but then
39:30you can see here who are we slowing down
39:32for you see that little gentle highlight
39:35lots of design experimentation goes into
39:37that because we don't want to be in your
39:39face we want to be gentle soft and
39:41you'll notice that in the night this
39:43will go into dark mode because the
39:45ambient lighting is reduced so we don't
39:47want to be you know too bright because
39:49if you want to just take a nap
39:52well I mean that's that's a great point
39:55and I just I'm curious if there's other
39:57learnings from you know again you've
39:59rolled this out the number one thing is
40:01safety but then from there it's like how
40:02do you create a great product that
40:04people want to engage with that people
40:05want to come back to so it's not just a
40:07novelty right where you're like oh you
40:09know I sat in an autonomous vehicle
40:10where they want to use this for their
40:12commute every day or you know for their
40:15their daily lives so how do you think
40:17about that other than maybe the screen
40:18are there other things oh yeah with
40:20thousands thousands and by the way uh
40:22speaking of mods that's another powerful
40:24flywheeler Advantage right if you have
40:28the first 10 000 humans who have been in
40:31a fully autonomous car the feedback they
40:33give you and the time you have to
40:34incorporate that into your learnings is
40:36a great positive flywheel approaching
40:39you know hundreds uh a thousand fully
40:41Artemis rides in a month so that
40:44feedback does help I'll give you just a
40:46couple of examples we could speak for
40:47hours just on that part one example I'll
40:50tell you is imagine a residential street
40:52like this this one that we're passing
40:55remember when we were getting in the car
40:56you were asking which side of the street
40:58should I be on yeah now imagine you're
41:00just getting out of home yeah going to
41:02work some writers you would imagine
41:04expect the vehicle to pull up right on
41:06their side of the street
41:09actually it turns out to be largely
41:12so what so because the way you would
41:14achieve that is let's say you're coming
41:15from this direction this direction you
41:17would make a U-turn yes and come back to
41:19them but most spread you know on narrow
41:21residential streets folks are like that
41:23was not necessary you know you know I
41:26could it takes me a second to walk
41:27across the street I could have entered
41:29on the other street but you didn't need
41:34so but that same reasoning does not hold
41:38on a street like this one yes yeah
41:40because you're like why are you making
41:41me cross you know two lanes of traffic I
41:44was going to the coffee shop on the
41:46other side just drop me in front of the
41:48coffee shop this Nuance so you can see
41:51how it's not just a single rule that you
41:53code in you don't say hey always park on
41:55the side that The Pedestrian is on no it
41:58depends on the context if you're on a
42:00busy street like that and somebody's
42:01going to a business they would like to
42:03be dropped very likely in front of the
42:04business they're going to whereas if
42:06they are getting out of home going to
42:07work and it's a narrow residential
42:09street just go to whichever side is
42:11closest yeah that's just a thing just
42:13that rule training it required speaking
42:16to many riders where they expect
42:18learning that and then being able to
42:20articulate that from a machine learning
42:22standpoint and an overall rules-based
42:24stand that's just one example many more
42:26are there any other learnings about what
42:28makes people feel comfortable or want to
42:30come back like one thing that's coming
42:31up is like I can see that driving wheel
42:34right and I can imagine especially once
42:36L5 is hit then you know we don't really
42:39need the steering wheel right you don't
42:41need the same car design because in the
42:45past you know for 100 years cars have
42:47been designed around the driver right
42:48and now we don't have a driver so you
42:51know it introduces all these questions
42:52but I'm curious I know we haven't
42:54removed the steering wheel per se but
42:56are there other dynamics that just from
42:59us growing up in cars designed a certain
43:01way that we expect certain things and
43:03then are there other things where you're
43:04actually like oh no we can start to get
43:06rid of some of this yeah yeah
43:08um so yeah I mean specifically about the
43:10steering wheels block currently by uh
43:13regulations Beyond a certain number of
43:14vehicles and so on and uh you know our
43:17next Generation vehicle that we did
43:19design with CVT and Julie is a pretty
43:22powerful platform part with the writer
43:24in mind so we spend months and years
43:26with uh designers on on all teams trying
43:29to visualize that I personally do
43:31believe that uh you know thinking of all
43:34the screen and software aspects and he's
43:37like waving the car yeah
43:39and and that was beautiful wasn't it
43:42like it's it's polite but it's also
43:44responsive and that that uh you know you
43:47know today we spoke about everything
43:49from safety to artificial intelligence
43:52to design to user understanding in that
43:55three seconds there all of that got
43:57exercise right because we were confident
43:59that we're not you know there's enough
44:00Gap there's no imminent contact we saw
44:04a collaborative fellow occupant of the
44:06road not an anniversary aerial one
44:07because sometimes yeah
44:14yelling is uh expression of emotions
44:16that's okay yeah the physical
44:18demonstration of you know turning in or
44:21like that example that we saw right now
44:23somebody who ran a red light if you
44:25recall those are the ones that are truly
44:28you know a good result in danger which
44:31is what the vehicle positions it it
44:33thinks about many small details where is
44:35it positioning itself how much Gap is it
44:37leaving what's its velocity so that it
44:38has the greatest optionality if somebody
44:41were to behave recklessly okay so when
44:43we're crossing that sign we're not
44:45assuming that everybody is going to obey
44:46the light we're trying to monitor their
44:49light is red okay which means they
44:51should be slowing down why aren't they
44:54slowing down that's an anomaly let's
44:55prepare for this anomaly let's protect
44:57let's be defensive against that normally
44:59like something that comes to mind is
45:00just as a rider you wanting to feel
45:02confident in in the car and seeing it be
45:05a little more assertive is actually
45:06really reassuring at points when it
45:08makes sense yeah because it's your point
45:10if it's constantly stopping if it's
45:11constantly pulling over then I don't
45:13have confidence that this thing is going
45:15to know what to do in a kind of yeah and
45:17you have a busy life you want to get
45:19where are you going right so by the way
45:21iconic Place coming up oh yeah why are
45:23we at the Painted Ladies yeah
45:25I think I see them up there yeah yeah
45:28that's by the way that's another thing
45:30you were asking early on about where are
45:32we in the journey we are in San
45:35Francisco we are in the most so this is
45:39I'm not you know we're not taking you to
45:41some you know little test facility in a
45:44desert to show you that this is
45:49most vibrant cities on the planet and if
45:52you wanted to come by to Los Angeles we
45:54would take you to the most important
45:56part of Los Angeles and in Phoenix we'll
45:59take you to downtown to Scottsdale and
46:00when you land at the airport we will
46:02pick you up at the airport that's where
46:04we are in the journey we're fully
46:05autonomous where we need to be something
46:07that it tells you you're finding a spot
46:08to pull over look at that right and it
46:12sees all those little kids there do you
46:15and it's more cautious because it
46:16understands that kids can jump out more
46:18erratically they're unpredictable yes oh
46:22and vehicle approaching so that's what
46:23you were saying earlier about so you
46:25know not to open your door exactly
46:27exactly now you can continue please make
46:29sure all right so those are the Painted
46:33but yeah I love the point that we're
46:34just we're fully on the road we're not
46:37doing you know we're not in the middle
46:38of nowhere practicing
46:40um and you should see the eyes of that
46:42kid just noticing that we don't have a
46:43driver like he's young enough where he
46:46he can't even articulate but he's like
46:47this isn't my pattern recognition is
46:50awesome and by the time he grows up this
46:52will be the more yeah well I mean you
46:54mentioned we're so you're an SF you're
46:56in Phoenix yeah um Los Angeles L.A how
46:59have you decided which markets to
47:01address first is it just a matter of
47:03what cities will welcome this technology
47:05or what goes into that calculus in the
47:09very very early days we tried to make
47:10sure that we're uh picking a city that
47:12you know challenged the system in very
47:14different directions because we were in
47:15our development stage we wanted to get
47:17so we tested in 20 cities just to make
47:19sure that from the very early days we're
47:21building a generalizable driver not one
47:23that just works in one location but
47:25generalizes double parked truck by the
47:28way and so is it it's even sensing maybe
47:30the lights blinking you know when for
47:32example somebody's unparking we notice
47:34their nose is starting to Jet out it's
47:37just so bizarre to see a wheel moving on
47:40its own like that like it truly does
47:41look kind of fake in a way
47:44um but yeah so you're trying to find
47:46cities that kind of test the system push
47:49it forward and and then what we found is
47:51that uh we have tested for example in
47:53Miami as well for the heavy rain we have
47:55tested in Death Valley for extreme
47:57temperatures we have uh tested in Tahoe
48:00for snow so there was a testing phase in
48:01which we went to 20 cities just to make
48:03sure with enough of a diverse data set
48:06to be building off of look at that
48:07gentle because this car was coming that
48:11um anyhow this beauty to see
48:14it really is something that it's like
48:16I'm watching and you can see this
48:18pedestrian crossing the car going
48:20another one approaching but green light
48:22so right after The Pedestrian went off
48:24anyway sorry uh you were asking does it
48:28get boring for you after working I love
48:30seeing your excitement because I mean
48:32this is my first time so I really am
48:34like as they say like taking it all in
48:36but it's cool you've how many times do
48:38you think you've been off like just the
48:40first year in 2019 I spent a ton of
48:42hours on it uh actually were you scared
48:45at all when you were first testing it
48:46because now I mean I guess I've seen
48:50some of the data I've seen these on the
48:52road yeah so there's a level of like oh
48:54I know these work yeah
48:56um but when you were first getting into
48:59the vehicle was there any like fear
49:02apprehension not fear you know the
49:05closest feeling I can describe is when
49:07you have prepared for an exam for six
49:10years you know whatever is the biggest
49:12exam you've given this little feeling
49:14that you got you know that you have
49:16prepared as best as you possibly could
49:18for it you left nothing on the table but
49:21it is exam day you're like oh gosh I
49:23hope I can show up right right well it's
49:25funny because a lot of people you know
49:27they'll see the stats they'll be like
49:28that's the average human but they think
49:30for whatever reason they outperform the
49:33average human on the road but I'm the
49:34opposite I'm like you where I'm like I
49:36want this I am not by the way the the
49:39you know the fundamental attribution
49:41error by the way is that uh you know
49:43more than 50 of people think they're
49:45better than average they're just not
49:47supposed to be feasible right so no I I
49:51um so maybe on the city selection just
49:52to close that out so yeah la uh 2
49:55billion plus 10 a huge diversity of use
49:58cases everything from commute trips all
50:00the way to you know uh sports events to
50:03you know sightseeing tours that you were
50:05mentioning San Francisco similarly both
50:07SF and LA are top two among the top five
50:11right-hailing markets in the U.S and
50:14among the top in the world right Phoenix
50:16is the fastest growing city in the
50:18United States it's uh Phoenix uh uh
50:22airport is among the top 10 busiest
50:24airports in the world so clearly
50:26commercial uh so when we commit to a
50:28city for a launch right that's different
50:30than the uh going to a city for data
50:32collection and testing something else
50:33that that's coming to mind is I wonder
50:35you know we are in the early stages
50:37we're only in a few cities but as
50:39consumers do see these on the roads like
50:41this is something that you know some
50:43people really will want to see in their
50:45cities and so have you seen any shift in
50:48terms of regulators you know maybe being
50:51a little against it to actually being
50:52like this is a competitive Advantage if
50:55my City offers this I I you know we
50:57definitely see that in the you know the
50:59place where we've been the longest is
51:00Phoenix East Valley and everybody from
51:03passengers do uh you know neighbors who
51:07even haven't taken a ride they notice
51:09that it's a much more polite driver to
51:12uh um law enforcement all the way to you
51:16know City and Merrell level all the way
51:18to state level absolutely I love
51:20thinking about just what does this
51:22histories that can unlock so much
51:25because people do spend so much time in
51:28cars I think that was an amazing
51:30beautiful thing we saw from remote work
51:31is just the second third order effects
51:33like what do we have when we get that
51:35commute back but then there still are
51:37people spending a lot of time on the
51:39roads and so I'd love to hear from you
51:42like what are some of those impacts The
51:44Wider impacts maybe it's on the way the
51:46insurance industry Works maybe it's on
51:48trucking maybe it's on City design now
51:50that we have data like that's another
51:51aspect we now have data about how people
51:54really move around a city and interact
51:56and how things are designed parking
51:58right so you know maybe pick and choose
52:01it's pretty vast yeah right so uh I I do
52:05believe it's it's uh pretty uh profound
52:09and only some of those aspects we can
52:11see and some we will be shocked by what
52:13what it will do because again like you
52:16said normally is the car designed around
52:19life is designed around driving right
52:22like look look at how much parking space
52:24you know that that house right there
52:27costs insanely higher amount in dollar
52:30per square mile as I'm sure as a new ssf
52:32Resident you were aware of I think of
52:34the department so there are so many
52:36things in our cities that are designed
52:39around uh Assumption of not only a human
52:42driver but also of a highly
52:44underutilized expensive asset just
52:47sitting there all these vehicles just
52:49add up the cost of these vehicles and
52:52think of how much space in the city they
52:54may be consuming while not adding to the
52:57productivity of the city because at this
52:59instant I'm sure each of these vehicles
53:01added to the mobility and freedom of
53:03individuals that's great
53:04but at this instant they are not
53:07you know utilized they're not adding
53:09value not neither to their Rider they're
53:11giving a promise that when the person
53:13who's gone in for a eight hour work day
53:14when they come out it'll still be there
53:16I heard this quote the other day it was
53:17like we dedicate in cities more space to
53:19sleep in cars and sleeping humans which
53:21is kind of crazy and first and foremost
53:24there are people who are traveling today
53:27that we believe over time
53:29the just the roads will get safer and
53:31that in itself you know I I know I'm
53:34saying safety so many times but that's
53:36truly is Central right 1.35 million
53:38people are killed on the streets but
53:40that's that's how is
53:43how is that acceptable right so that's
53:45first and foremost for all kinds of uh
53:48you know for Society at large it's a
53:50pandemic almost right and and this is a
53:53uh you know antidote to that and uh then
53:56there are classes of individuals for
53:58whom this freedom is not available above
54:0165 you know many folks can't drive
54:04anymore or or drive while taking risk or
54:07would have the freedom of Mobility there
54:10are uh folks with visual impairments yes
54:12then uh vulnerable populations being
54:15able to take jobs that they otherwise
54:17wouldn't have been that nighttime use
54:18case that we said giving Economic
54:20Opportunity uh so which they simply
54:22would not be comfortable or you
54:24shouldn't ask somebody to own a car
54:27before they can take their entry level
54:28job like that's uh you know challenging
54:31economic uh catch-23 right uh so there's
54:35that then the third is yes how much uh
54:38you know pollution idling cars in City
54:40centers are causing how much uh City
54:43real the state is being lost to idle
54:45assets so yeah layers and layers upon
54:48that and yes that's just in passenger
54:51vehicles then you consider in-city
54:53delivery then you consider a long-haul
54:56Trucking where already tremendous amount
54:58of Economic Opportunity in the United
55:00States is being lost to lack of drivers
55:02and by the way when we do have drivers
55:04we have mandatory brakes because fatigue
55:06is a real thing we recently moved I
55:08drove from near San Diego up to San
55:10Francisco and we happened to do it at
55:12night and just the number of truck
55:14drivers that you know we we saw on the
55:17route I just I knew this already but
55:19seeing it I was like this something
55:21feels wrong here and I just imagine a
55:23future where you know some people might
55:25not like this but we're that's automated
55:27that sounds beautiful to me yeah and and
55:29there's a you know beautiful transition
55:31points as well in the sense that truck
55:34driving could become a local job which
55:36would be powerful in many many different
55:38ways in this instead
55:40oh oh you see it since the pylon and you
55:43see that oh my gosh that's incredible
55:45because see I was wondering actually if
55:47it could tell if that's a human so they
55:49can tell that those are humans yeah but
55:51then I use every single one and there
55:53was a hand there was like a roughly
55:55written keep right and an issued going
55:57there yeah because you could imagine how
55:59it might think yeah you may think
56:01there's enough space there so I think
56:03construction zones by the way are uh
56:07um change to otherwise structured word
56:09right is there no construction zone is
56:11the same right yeah it's like each one
56:13is unique and the degree of uh
56:15intelligence required to figure that out
56:17is pretty it pretty you know substantial
56:20as well because it's suddenly you're
56:21breaking the prior structure of the
56:23world yeah so I mean you mentioned how
56:28Vehicles can actually make our world
56:32do imagine a future we're actually right
56:36now it's it's like most humans drive or
56:39actually becomes illegal or the minority
56:41of people who are able to drive because
56:44we get the technology so far along where
56:47it's just again it's a no-brainer for us
56:50to have the tech drive us around instead
56:52of Vice Versa the mission we have is
56:54being able to provide this option safe
56:58and easy for people and things to move
56:59around I think it's good to have the
57:03all right so maybe to close things off
57:04I'd love to hear you know you've been
57:05working in this space for a while what
57:07gets you excited uh you know seeing the
57:10writers uh the first time the fifth time
57:13and what they say about us that gets me
57:15excited amazing well thank you so much
57:17this was an excellent first ride I'm so
57:20excited to do this more in San Francisco
57:22I love the music too that they're
57:24they're welcome by the way we had lots
57:26of music as well we were having a
57:27conversation so I didn't show you but
57:29yeah a lot of effort has gone into this
57:30one as well yeah and I can't wait I
57:32imagine like this being personalized as
57:34well in the future you know the climate
57:37from your app we did both of these
57:39screens as well because there would be
57:41cases in which somebody would be in
57:42front synchronizing all of that but
57:44amazing for another time well this was
57:46great and I guess we just hopped out
57:47yeah yeah we're here
57:58this is great I guess I shouldn't it
58:01knows I'm here it knows I'm imagining
58:03myself on the screen yes a little
58:09so how did you find the ride I really
58:11enjoyed it I mean I feel like
58:14it's funny because I only really noticed
58:16the lack of driver for maybe two minutes
58:19and then as you saw I was so like
58:21involved in the conversation and you
58:23know you don't even notice the right
58:24driver there and that's that's what we
58:27what's that quote where it's like any
58:28sufficiently advanced technology is
58:31indistinguishable for magic and it
58:33really does like wow we're here yeah
58:35we're here we just did it I'm well I'm
58:37glad to have been there when you had
58:38your first experience yeah well thank
58:39you so much this was so cool yeah
58:44all right if you made it to the end here
58:46I just wanted to say thank you this
58:49episode in particular was actually
58:51really special to me I got my first
58:54learner's permit when I was 16 in Canada
58:56and I actually waited until I was 29 to
59:00get my driver's license because quite
59:02frankly it scared me and I was waiting
59:04for self-driving and it's finally at
59:07least in some places like San Francisco
59:09it has arrived and I am so excited I am
59:12so happy that I could share this with
59:14you and if you are just as excited as I
59:18am to see how this whole thing unfolds
59:20let us know in the comments what you are
59:22most excited about how you think
59:23autonomous vehicles might most reshape
59:26Society because there really are so many
59:28implications whether it's public
59:30infrastructure energy Finance shopping
59:32I'm so interested to see how this
59:34technology finally comes into play and
59:36on that note if you'd like to dig even
59:38deeper I'm going to leave you with an
59:40eight-part series that one of our
59:42partners at a16z created Five years ago
59:45but let me tell you it's incredible to
59:47see how much of this still holds up and
59:51actually it's still forward-looking all
59:53right on that note thank you again so
59:56much for joining me we'll see you next
59:58time and we'll see you on the road
01:00:03thanks for listening to the a16z podcast
01:00:05if you like this episode don't forget to
01:00:07subscribe here on YouTube to get our
01:00:08exclusive video content we'll see you