00:00hi and welcome to the a 16z podcast this
00:02conversation looks at where we are now
00:05in self-driving cars and in the big
00:07trend of autonomous and self-driving
00:09cars of all levels from data to mapping
00:11to interfaces to security and more
00:13moderated by a 16z partner Frank Chen
00:16the conversation part of our November
00:192017 summit event includes in the order
00:22in whose voice you will hear first
00:24Kassar Yunus CEO of Applied intuition
00:27James Wu CEO of deep math and Taggart
00:30Matheson director of product at lyft so
00:33on the scale from crazy 13 years to
00:36myopic never when does self-driving
00:39start it's not gonna be one of these
00:41things where we wake up one day and like
00:43autonomy is hair you know every two
00:46years every one year we're gonna just
00:47make that incremental move and I think
00:49when a ct6 launches with super cruise or
00:52the Tesla autopilot and then maybe ten
00:54years from now we'll wake up and you
00:55have that option to go to the dealership
00:56and buy a car but that's everything we
00:58think autonomy does the major OEMs in
01:02Europe u.s. and Japan other places if
01:05you look their comm table it's about
01:072020 2021 they're gonna start shipping
01:11level four vehicles I think it's
01:13probably happened much sooner than 13
01:17years for everybody to have the
01:19capability to use a level 4 technology
01:22go over the next couple of years you're
01:24gonna start to see very small pockets
01:25where these vehicles are in fact
01:28operating at a little level 4 capability
01:31but I think the challenge here is
01:32there's a cost aspect if you look at the
01:34sensor suite on these vehicles it's not
01:36economically feasible to once you get
01:38into the world of a large scale
01:40deployment of vehicles it's not just
01:42about the car being able to navigate the
01:44environment it's also just kind of the
01:46fleet operations as well from pre
01:48positioning to fueling it's gonna take a
01:50number of years before a large majority
01:53of these vehicles are autonomous but
01:55you're going to see over the next couple
01:56of years is small pockets of innovation
01:59where over time more and more vehicles
02:02are going to be operating in greater
02:03situations in terms of level four and
02:06then if you think about sort of the long
02:08pole in the tent do you think it's going
02:10to be technology it won't work is it
02:13you can't edit this maybe the sensors
02:15are too expensive is it policy the
02:17government's won't allow it or something
02:19else what will cause it to come more
02:21slowly if anything yeah I keep coming
02:24back to the economics in that scale
02:26first of all no one's at true level 5 or
02:30even level 4 at this point and to do
02:33that at scale in all different kinds of
02:36conditions from rain to snow even in a
02:39geographically confined area at a point
02:42where you've figured out the economics I
02:44think is gonna be you know quite a
02:46difficult challenge we'd say
02:48infrastructure is a big thing and I
02:49don't mean infrastructure in the
02:50traditional sense we're done with
02:52infrastructures like software
02:53infrastructure and in the actual product
02:55getting out to market sensor Suites it's
02:57like a real problem I think that that
02:59hasn't been solved yet and I think I
03:01would say secondly its cultural norms
03:03like you know I think the old adage in
03:05automotive is like the stop sign came
03:06out ten years after the you know after
03:09the Model T and so there's a lot of
03:11regulatory hurdles that need to be kind
03:13of defined and overcome well me I think
03:16the major bottleneck at this stage is
03:20probably talents in the next few years
03:23because suddenly you need a lot of
03:26robotic experts a lot more computer
03:29science engineers scientist and this
03:32will be a bottleneck for current state
03:35this is the heart of like you know
03:37making it happen I think that's what
03:39you're really hearing is to make these
03:41long-term visions happen there's just a
03:43lot of these individual hurdles which
03:46they're non-trivial so most of the
03:48discussion around self-driving cars on
03:50the benefit side is around convenience
03:52you don't have to drive yourself in
03:53safety right we're gonna stop killing
03:55people on the roads with these dangerous
03:56cars what other ways are our cities or
03:59our lives going to get better
04:01aside from convenience and safety that
04:03nobody's talking about crime rate will
04:05drop in my opinion because basically
04:08every car will became a moving sensor
04:11collecting tons of real-time information
04:13it's like a surveillance system right
04:15and also the housing price public on
04:17that change because people don't have to
04:20like hack together and also all these
04:21parking space will become available
04:24space for housing there's a huge amount
04:27there's a hundred Tier one suppliers
04:29right now that make over a billion in
04:31revenue year I'm from Detroit I grew up
04:35I really wonder like what's gonna happen
04:37to those companies yep
04:38I think the real bloodbath in this
04:39ecosystem is not gonna be at the
04:42manufacturers level cuz consumers have
04:43some affiliation to like a Tesla vehicle
04:46or like a Mercedes Benz etc it happens
04:48at the tier ones but where there's that
04:50destruction there's an amazing
04:51opportunity you have these massive
04:53companies which are gonna be almost
04:55useless in in this new economy and you
04:58have this new huge market so I think
04:59that untapped opportunity is I think
05:02there's still huge huge potentials for
05:03soft core specific vehicle kind of
05:06applications and I'll kind of play on
05:08top of that and we see this today with
05:10with ride-sharing it's getting time back
05:12and so I go down to Palo Alto all the
05:16time and I can't tell you how quickly I
05:17want to have an autonomous vehicle
05:18driving I spend 30 to 40 minutes on like
05:22the best day getting down to Palo Alto
05:23but the hard reality is how can I
05:25reimagine that time and we were talking
05:28about Netflix earlier in stranger things
05:30the beauty would be Netflix already
05:32knows that I'm on season 2 episode 4 the
05:35car should already know that it's you
05:36know 40 maybe an hour long and it should
05:39just pop in that video for me and
05:41whether it's like something you know
05:42focused on entertainment or just having
05:45that connectivity from a work
05:46perspective or connecting with friends
05:48the point here is that you can get that
05:50time back and I think that's going to be
05:51pretty magical I love this idea that
05:53sort of personalization your fleet
05:55remembers what you were doing what you
05:56were using becomes the buying criteria
05:59right so it's not the door thunk it's
06:01the do you remember that I was watching
06:03stranger things all right
06:04good I have seen now probably half a
06:06dozen startups that are working on
06:08basically remote driving whether it's
06:10remote driving a car or remote driving a
06:12truck on the assumption that the
06:13algorithms won't be able to figure out
06:15what to do in all cases so is this gonna
06:18be a thing how long will it last I think
06:21it's gonna be a thing if you think about
06:22the challenge of self-driving vehicles
06:24it's kind of the unknown unknown if the
06:26situation or the scenario that you never
06:28tested or was aware of and so you know a
06:31car gets into a situation where it's not
06:33able to classify or understand the scene
06:35around it you can imagine a Tele
06:37operator quickly going in there and
06:39helping like to some extent labeling
06:41and letting the car kind of understand
06:43the context of the environment
06:44- maybe even drawing a line to say look
06:47vehicle here's a valid path for
06:49navigation however I would argue on the
06:51tella operations you need to be very
06:53very careful here where I don't think
06:55they're gonna be driving the car I would
06:57almost argue that it helps the car
06:58understand the scenario or the scene and
07:00then leave it up to the vehicle to then
07:02do that navigation and motion control
07:04the the concern I would have is one in
07:06terms of latency so the operator doesn't
07:09actually see something that happens in
07:10the scene and then the other thing just
07:11comes it's yet another vector for
07:14potential people taking over the
07:16vehicles security-wise sure to me I
07:19think self-driving is really not try to
07:21get rid of the driver this actually get
07:24people more efficient like instead of
07:26driving one car one person is driving or
07:30managing thousands of cars by one person
07:32I think that's some driving still a
07:35technology is a tool to your neighbor
07:37human being it's like the DevOps
07:40operator see in data centers right right
07:42so now like one CEO guy can manage
07:44millions of machines that dick
07:46time-lapse is gonna be a critical part
07:48of that software infrastructure I think
07:49less in the traditional passenger space
07:51I think those edge cases relatively are
07:54less than if you're doing like automated
07:56construction equipment caterpillar has
07:58done things like this in coal mines
07:59we're having a Tele operator it kind of
08:03always on call managing those fleets I
08:04think becomes really that's where you're
08:06in kind of an unbounded environment but
08:09it's one of these like I think just
08:11chunks that will exist and probably for
08:13ultimately I think those companies will
08:14become fleet management companies but
08:16yeah I think it's gonna happen with it
08:18out and to add to that actually even we
08:21have this kind of operations for planes
08:23for trains for subway systems even for
08:25shuttles already there
08:27so we were honored to have governor doug
08:29ducey from arizona join us at the
08:31conference he's on the record saying
08:32look I want Arizona be the friendliest
08:34state for these types of experiments so
08:38if you could do one or two things as the
08:40governor what would you do so that we
08:43get to this world faster well I mean I
08:45think Arizona I think Nevada I think
08:47Michigan in Florida I think there are
08:48states who are like very actively asking
08:52that question and I think they're doing
08:53a lot of the things that you too
08:54expect we build advanced simulators for
08:56autonomous vehicles and you know we very
08:59much engage with regulatory kind of
09:01bodies both the federal and state level
09:02to really answer this long-term question
09:04of it the DMV goes away and you have
09:07these over there updates to self-driving
09:08cars how do we make sure that we don't
09:10have some regression in the soft crow
09:12ends up creating a scenario which puts
09:15the entire industry a step back that's
09:17kind of where the I getting a real
09:19bleeding edge of the conversation is
09:21it's not the ten thousand vehicles it's
09:23really the what is it when we have a
09:25million vehicles and you have updates in
09:28software happening every single day
09:29once you have thousands of players doing
09:31really edge case you know essentially
09:33autonomous robots not necessarily only
09:35self-driving that's gonna be a big
09:37question the first time we're
09:39introducing really intelligent robots
09:41into the world and it's the first time
09:43we as a society have to regulate these
09:45robots and in a way that is you know not
09:49constraining innovation but at the same
09:51time not putting the public at harm so
09:54let me push on that a little which is do
09:56you think the government should have a
09:57role on over-the-air software updates
09:59right so today we do lots of
10:01over-the-air updates there are iPhones
10:02and androids and two windows government
10:04has no role in that would you argue that
10:06the government should in the
10:08self-driving car world because the
10:09stakes are higher do you need to have
10:10some basic standards so whether that's
10:12cybersecurity whether that's there's a
10:14multitude of things that we can talk
10:15about there and the professional quote
10:18unquote companies that have brands that
10:19have so much riding on it that can't
10:21afford to have a real issue that's not
10:23the problem it's the long tail select
10:25FinTech regulation it's not regulating
10:27necessarily the Goldmans of morgan
10:29stanley's of the world it's this you
10:30know individual person who's trying to
10:31sell penny stocks to people who
10:34shouldn't be buying them they're not
10:35accredited investors that's where
10:36regulations really come in it's to
10:38protect the uninformed and so that kind
10:41of version of this world is whether
10:43that's construction equipment or that's
10:44shuttles or whether that's passenger
10:46vehicles definitely people can die
10:48Google Maps is a pretty sophisticated
10:49product but we would do who like updates
10:51where we'd have P zero bugs introduced
10:53into like the public domain and we would
10:55immediately fix those you can't do that
10:57in autonomous vehicles I couldn't agree
10:59more on like making sure that there are
11:02standards I think the challenge here is
11:04that you don't want a patchwork of
11:05different standards you know one fit one
11:07correct to stay on track
11:08this isn't something that you should
11:09just leave it up to a group of
11:12organizations I think this is a
11:13collaboration between these companies
11:16today as well as both at the state and
11:18federal level getting to some level of
11:20certification and standards to ensure
11:23that these are safe vehicles I think the
11:25government can do two major things to
11:28help the self-driving industry move
11:30forward one direction is to promote
11:32cooperation and provide digital
11:36infrastructure either add investment in
11:39digital infrastructure building or
11:41promote and help companies jumping in to
11:44help be with the digital infrastructure
11:46by digital infrastructure I mean like HD
11:48maps and the other direction is to draw
11:51the boundary and limit where we can
11:54actually play with self-driving
11:56technologies for example we can probably
11:59have some certain straight paved with a
12:02self-driving lane instead of carpool
12:04lanes that's kind of giving people some
12:07kind of boundary where the technology is
12:09I want to turn the line of questioning a
12:12little and talk about sort of our
12:14personal experiences so what would you
12:16say to the pitchfork people right like I
12:18love driving driving as independence
12:20it's part of my self-identity I identify
12:22with my car what would you say to
12:24somebody like that who surfaces those
12:26types of objections to this feature well
12:28I mean I'm as much of a car guy is
12:30anybody else I grew up in the Detroit
12:31area and went to GM I for undergrad
12:33worked at General Motors for five years
12:34I love the auto industry I think mixing
12:37those emotional sentiments with what's
12:39right for society it shouldn't and is
12:41not an individual decision it ultraman's
12:43of being a decision that we all make
12:44together by the choices that we make in
12:46the way that we allocate capital both as
12:48private investors and there's individual
12:49citizens whether you like it or not it's
12:52gonna happen the second thing I would
12:53say is the reality is it's not that
12:56we're gonna take away people's cars I'd
12:59like to go to the track it's a hobby of
13:00mine and I don't think that in the
13:02duration of my life is going away and so
13:04the people were really into cars I don't
13:06think they have anything to worry about
13:07and for everybody else who doesn't want
13:09to do a commute it's a wonderful change
13:10so I don't think it's like this binary
13:12you know we have to have a revolution to
13:14inject this product into the consumers
13:16what's probably gonna happen is car
13:18ownership is not gonna go to zero what
13:20will happen is this really becomes more
13:22and less of a utility and so if you want
13:25your hot rod or if you want your you
13:27know muscle car from the 60s I don't
13:29think that anyone's going to stop you
13:31from doing that I think the interesting
13:32question becomes over a long enough
13:34period of time do states do cities allow
13:37potentially you know roadways or
13:40arteries for transportation that have
13:42fully autonomous to have human drivers
13:44in there as well just from a safety
13:45perspective you can still have your
13:47muscle car in 20 years can you drive
13:49that muscle car on the same streets I
13:51don't know I think that's an interesting
13:52question like you don't see horses going
13:54down 101 so often right just to follow
13:56but I will point out that for people who
13:59love to ride horses there are still
14:00places for you to ride horses today and
14:03also I think for self-driving cars it
14:07will have a rely on data especially
14:09mapping data so I think for certain
14:12space like rural regions or tough road
14:16and people seldom drive we don't have
14:19enough data there is probably still need
14:21human drivers there for a long time so
14:23this issue of security and we've been
14:26talking around the edges I'm going to
14:27sort of hit head-on which is how do we
14:30deal with it are we gonna have to have
14:32open source so everything is open to
14:34inspection are we going to need to get
14:36government regulations in here the same
14:38sort that would govern over-the-air
14:39updates like what's the answer to how do
14:42we make sure that these things can't be
14:44hijacked the biggest you know
14:45correlation here or analogy here would
14:48be the problems that we currently have
14:49with the web and and and mobile web I
14:51mean we have massive data breaches so
14:53common that that a we can't even list
14:56them anymore I think we will have them I
14:58think the key point will be there will
15:00have to be some version of overrides
15:02that are still local to the car that's
15:04why you won't see this like iRobot type
15:06of environment emerging immediately
15:08until that infrastructure gets built up
15:10we go through some of these growing
15:12pains step backs and then I think
15:14ultimately we'll have that maybe in 20
15:16years that future that we think of with
15:18no steering wheel no brakes I don't
15:20think anyone here is gonna claim that
15:21they're gonna be able to solve security
15:23I think it's going to be a very big
15:25challenge and anyone looking at this
15:26space is absolutely digging into that
15:28and that's one of the reasons why I was
15:30talking about remote operators and how
15:31you limit the surface area by which
15:33someone can attack the vehicle the
15:35minute you have a No
15:36system is the minute someone can get in
15:37but even if you have a closed system
15:38there's the physical area of the car
15:40whether you're gonna you know open up
15:42the backseat maybe there's some access
15:44and now you can get into the you know
15:46the brakes or the steering it's gonna be
15:48a problem that we're gonna have to solve
15:49and there will be a lot of opportunities
15:52in that space for startups we are
15:55focusing on data security so when you
15:57look at the self-driving wiko it's
15:59actually collecting a lot of data with a
16:01super high precision Center meter level
16:03procedures and also privacy information
16:06so data ownership and the data security
16:08is a major concern we need to make sure
16:10the data should stay where it should
16:13stay and not get into the hands of
16:15people like try to abuse it alright my
16:18last question each one of you is
16:19basically working in a company that's
16:21trying to bring this vision to life and
16:23so you've made the ultimate bet right
16:25your time and energy and talent are
16:27headed in this direction what are you
16:28most excited about saving lives
16:30I really ninety so my percent of vehicle
16:34accidents can be avoided and so I think
16:37this is a huge opportunity for us to
16:38change that I mean we cannot overstate
16:41the amount of people that get you know
16:44injured and maimed and died every year
16:46that's insane the number is something
16:48like a 747 falling out of the sky every
16:50day and everyone dying that's the
16:53equipment how many people died and like
16:55if that happened for three or four days
16:56and any other like in the air
16:58transportation industry we would
16:59definitely land all planes that say what
17:01the hell is happening that we have 747s
17:03falling out of the sky and automotive we
17:05just kind of like saying well you know
17:06what let me just a text something to my
17:08friend as I get in the car so I think
17:10it's yeah that's the biggest benefit by
17:12far nothing compares to it for us is
17:14saving time so basically we want to
17:17enable our players serious customers
17:20serious players in this self-driving
17:22space to get to state we're way more the
17:25other forerunners are by providing this
17:28HTML solution to them good and with that
17:32that's a wrap thank you very much for