Kind: captions
Language: en
following is a conversation with
Sebastian Thrun he's one of the greatest
roboticists computer scientists and
educators of our time he led the
development of the autonomous vehicles
at Stanford that one 2005 DARPA Grand
Challenge and placed second in the 2007
DARPA urban challenge he then led the
Google self-driving car program which
launched the self-driving car revolution
he taught at the popular Stanford course
on artificial intelligence in 2011 which
was one of the first massive open online
courses or MOOCs as they're commonly
called that experience led him to
co-found Udacity an online education
platform if you haven't taken courses on
it yet I highly recommended their
self-driving car program for example is
excellent
he's also the CEO of Kitty Hawk a
company working on building flying cars
are more technically Evie tall's which
stands for electric vertical takeoff and
landing aircraft he has launched several
revolutions and inspired millions of
people but also as many know he's just a
really nice guy it was an honor and a
pleasure to talk with him
this is the artificial intelligence
podcast if you enjoy subscribe I need to
give it five stars and Apple podcasts
follow it on Spotify supported on
patreon or simply connect with me on
Twitter and Lex Friedman spelled Fri D
ma M if you leave a review on Apple
podcast or YouTube or Twitter consider
mentioning ideas people topics you find
interesting
it helps guide the future of this
podcast but in general I just love
comments with kindness and
thoughtfulness in them this podcast is a
side project for me as many people know
but I still put a lot of effort into it
so the positive words of support from an
amazing community from you really helped
I recently started doing ads at the end
of the introduction I'll do one or two
minutes after introducing the episode
and never any ads in the middle they can
break the flow of the conversation I
hope that works for you and doesn't hurt
the listening experience I provide time
stamps for the start of the conversation
that you can skip to but it helps if you
listen to the ad and support this
podcast by trying out the product
service being advertised this show is
presented by cash app the number one
finance app in the App Store I
personally use cash up to send money to
friends but you can also use it to buy
sell and deposit Bitcoin in just seconds
cash app also has a new investing
feature you can buy fractions of a stock
say $1 worth no matter what the stock
price is brokerage services are provided
by cash app investing a subsidiary of
square and member has site PC I'm
excited to be working with cash app to
support one of my favorite organizations
called first best known for their first
robotics and Lego competitions they
educate and inspire hundreds of
thousands of students in over 110
countries and have a perfect rating and
charity navigator which means the
donated monies used the maximum
effectiveness when you get cash out from
the App Store or Google Play and use
coal export gas you'll get ten dollars
in cash up we'll also donate ten dollars
the first which again is an organization
that I've personally seen inspired girls
and boys the dream of engineering a
better world and now here's my
conversation or Sebastian Thrun you
mentioned that the matrix may be your
favorite movie so let's start with the
crazy philosophical question do you
think we're living in a simulation and
in general do you find the thought
experiment interesting define simulation
I would say maybe we are not but it's
completely irrelevant to the way we
should act putting aside for a moment
the fact that it might not have any
impact on how we should act as human
beings for people studying theoretical
physics these kinds of questions might
be kind of interesting looking at the
universe's information processing system
the universe is an information
processing system is a huge physical
biological chemical computer there's no
question but I live here and now I care
about people okay about us what do you
think is trying to compute and I think
there's an intention I think it just the
world evolves the way it devolves and
it's it's beautiful is unpredictable and
I'm really grateful to be alive
spoken like a true human which last time
I checked that was oh that in fact this
whole conversation is just a touring
test to see if if indeed if indeed you
are you've also said that one of the
first programs of the first few programs
you've written was a wait for a TI 57
calculator yep
maybe that's early eighties I don't
wanna date calculators anything early
eight is correct yeah so if you were to
place yourself back into that time into
the mindset you are in because you have
predicted the evolution of computing AI
the internet technology in in the
decades that followed I was super
fascinated by Silicon Valley which I
seen on television once and thought my
god this is so cool they build like D
Rams there and CPUs how cool is that and
as a college students a few year later a
few days later I decided to be study
intelligence and study human beings and
found that even back then in the 80s and
90s that artificial intelligence is what
fascinated me the most I was missing is
that back in the day the computers are
really small they're like the brains you
could build well not anywhere bigger as
a cockroach and cock-horse aren't very
smart so we weren't at the scale yet
where we are today
did you dream at that time to achieve
the kind of scale we have today did that
seem possible I always wanted to make
robots smart I felt it was super cool to
build an artificial human and the best
way to build not official you want to be
a robot because that's kind of the
closest if you could do unfortunately we
aren't there yet there were words today
are still very brittle about as
fascinating to study intelligence from a
constructive perspective it built
something to understand you build what
do you think it takes to build an
intelligent system and an intelligent
robot I think the biggest innovation
that we've seen as machine learning and
it's the idea that their computers can
BC teach themselves
let's give an example I'd say everybody
pretty much knows what a wok and we
learn how to walk in the first year two
of our lives but no scientist has ever
been able to write on the rules of human
gait we don't understand that we can't
put we have in our brain somehow we can
practice it we understand it that we can
articulate that we can't pass it on by
language and that to me is kind of a
deficiency of today's computer
programming even you could program a
computer they're so insanely dumb that
you have to give them rules for every
contingencies very unlike the way people
learn but learn from data and experience
computers are being instructed and
because it's so hard to get this
instruction set right we pay software
engineers two hundred thousand dollars a
year now the most recent innovation
which has been to make for like 3040
years is an idea that computers can find
their own rules so they can learn from
falling down and getting up the same way
children can learn from falling down and
getting up and that revolution has led
to a capability that's completely
unmatched today's computers can watch
experts do their jobs whether you're a
doctor or lawyer pick up the
regularities learn those rules and then
become as good as the best experts so
the dream of in the 80s of expert
systems for example had at its core the
idea that humans could boil down their
expertise on a sheet of paper so sort of
reduce sort of be able to explain to
machines how to do something explicitly
so do you think what's the use of human
expertise into this whole picture do you
think most of the intelligence will come
from machines learning from experience
without human expertise input so the
question for me is much more how to
express expertise um you can express
expertise providing a book you can
express expertise by showing someone
what you're doing you can express
expertise by applying it by by many
different ways and I think the expert
systems was our best attempt in AI to
capture expertise in rules there someone
sat down and say here the rules of human
gait here's when you put your big toe
forward and your heel backwards and
Yahoo stop stumbling and as we now know
the set of rules a set of language that
he can command is incredibly limited
the human brain doesn't deal with
language it is with that subconscious
numerical perceptual things that we
don't even ever survey off
now when a AI system watches an expert
do their job and practice their job it
can pick up things that people can't
even put into writing into books or
rules and that's where the real power is
we now have AI systems that for example
look over the shoulders of highly paid
human doctors like dermatologist or
radiologists and they can somehow pick
up those skills that Noah can express in
words so you were a key person in
launching three revolutions online
education and Thomas vehicles and flying
cars or vetoes so high level and I
apologize for all the philosophical
questions that's no policy necessary how
do you choose what problems to try and
solve drives you to make those solutions
a reality I have two two desires in life
I want to literally make the lives of
others better or as few of them say
maybe joke indeed what make the world a
better place if you believe in us it's
as funny as it sounds and second I want
to learn I want to get in the circus I
don't want to be in a dropping with it
because if I meant job that I'm good at
the chance for me to learn something
interesting is actually minimized so I
want to be in a job I'm bad at that's
really important to me so I'm in a bill
for example but people often call flying
cars is that electrical vertical takeoff
and landing vehicles I'm just no expert
in any of this and it's so much fun tool
to learn on the job what actually means
to build something like this now it's
saying that the stuff that I done lately
after I finished my professorship at
Stanford the video focused on like what
has the maximum impact on society like
transportation is something has
transformed the 21st 20th century more
than any other invention of my opinion
even more than communication and cities
are different workers different women's
rights are different because of
transportation and yet we still have a
very suboptimal transportation solution
where we kill 1.2 or so million P
every year in traffic it's like the
leading cause of death for young people
in many countries we have here extremely
inefficient resource wise just go to
your average neighborhood city and look
at the number of parked cars that's a
travesty in my opinion or where we spend
endless hours in traffic jams and very
very simple innovations like a
self-driving car or what people call a
flying car could completely change this
and it's there I mean the technology is
it's basically there yet close your eyes
not to see it
so lingering on autonomous vehicles
fascinating space some incredible work
you done throughout your career there so
let's start we'll start with DARPA
I think the DARPA challenge there's a
desert and then urban to the streets I
think that inspired an entire generation
of roboticists and obviously sprung this
whole excitement about this particular
kind of four wheeled robots were called
autonomous cars self-driving cars so you
led the development of Stanley the
autonomous car that one that erased the
desert the DARPA challenge in 2005 and
junior a car that I finished second in
the DARPA Grand Challenge also did
incredibly well in 2007 I think what are
some painful inspiring or enlightening
experiences from that time that stand
out to you oh my god painful were all
these incredibly complicated stupid bugs
that had to be found we had a face where
the stanley hour or carded i eventually
won the DARPA Grand Challenge but every
30 miles just commit suicide and we
didn't know why and it ended up to be
that in the sinking of two computer
clocks occasionally a clock went
backwards and that negative time elapsed
screwed up the entire internal logic but
it took ages to find this they were like
bugs like that I'd say enlightening is
the Stanford team immediately focused on
machine learning and on software where's
everybody else seem to focus on building
better hardware our knowledge had been
you a human being with an existing
rental car can
we drive the course I have to might have
to build a better rental car I just
built it should replace the human being
and the human being to me was a
conjunction of three steps we had
extensors eyes and ears
mostly eyes we had brains in the middle
and then we had actuators our hands in
our feet now the extras I used to build
the sensors like she also use it a bit
what was missing was the brain so he had
to build a human brain and nothing
nothing clear them to me that that the
human brain is a learning machine so why
not just train our robot so it you would
build a massive machine learning into
our machine and with that were able to
not just learn from human drivers we had
the entire speed control of the vehicle
was copied from human driving but also
have the robot learn from experience
where it made a mistake and go to
recover from it and learn from it you
mentioned the pain point of software and
clocks
synchronization seems to seems to be a
problem that continues with robotics
it's a tricky one with drones and so on
Oh what what does it take to build a
thing a system with so many constraints
you have a deadline no time you're
unsure about anything really it's the
first time that people really do even
explore yeah it's not even sure that
anybody can finish when we were talking
about the race of the desert the year
before nobody finished what does it take
to scramble and finish a product that
actually a system that actually works we
were very lucky we did a small team that
core of the team of four people it was
four because five couldn't comfortably
sit inside carpet for food and I as a
team leader my job was to get pizza for
everybody and wash the car and stuff
like this and repair the radiator and it
broke and debug the system and we were a
kind of open mind that we had like no
egos involved in this you just wonder to
see how far we can get or we did really
really well was time management we were
done with everything a month before the
race and we froze the entire of a month
before the race and it turned out
looking at other teams every other team
complained if they just one more week
they would have won and we decided
that's gonna fall into a mistake you're
gonna be early and we had an entire
month to shake that system
and we actually found two or three minor
bugs in the last month that we had to
fix and we were completely prepared in
the race occurred okay so first of all
that's such an incredibly rare
achievement in terms of being able to be
done on time or ahead of time what do
you how do you do that in your future
work what advice do you have in general
because it seems to be so rare
especially in highly innovative projects
like this people work till the last
second but the nice thing about the
topic one challenge is that the problem
was incredibly well-defined we were able
for a while to drive the old topic van
challenge course which had been used the
year before and then at some reason we
were kicked out of the region so we had
to go to different desert the Sonoran
Desert and be able to drive desert
trails just at the same time so there
was never any debate about like what is
actually the problem we didn't sit down
and say hey should we build a car or a
plane if we had to build a car that made
it very very easy then I studied my own
life and life of a dozen guys that the
typical mistake that people make is that
there's this kind of crazy bug left that
they haven't found yet and and it's just
there regretted and it back would have
been trivial to fix it was haven't fixed
it yet they didn't want to fall into
that trap so I build a testing team we
had a testing tena build a testing
booklet of 160 pages of tests we had to
go through just to make sure we shake
all the system appropriately Wow and the
testing team was with us all the time
and dictated to us today we do railroad
crossings tomorrow over do we practice
the start of the event and in all of
these we thought oh my god has long
solved trivial and I mean tested it out
oh my god it doesn't were a well for us
and why not oh my god it mistakes the
and the rails for metal barriers we have
to fix this yes so it was easy a
continuous focus on improving the
weakest part of the system and as long
as you you focus on improving the
weakest part of the system you
eventually build a really great system
let me just pause Allah is to me as an
engineer is super-exciting that you were
thinking like that especially at that
stage as brilliant that testing was such
a core part of it it may be to linger on
the point of leadership
I think it's one of the first times you
were really a leader and you've led many
very successful teams since then what
does it take to be a good leader I would
say I'm most of all just take credit for
the work of others right that's that's
very convenient than that because I
can't do all the things myself I'm an
engineer at heart so I I care about
engineering so so I I don't know what
the chicken in the egg is but as a kid I
love computers because you could tell
them to do something and they actually
did it it was very cool and you could
like in the middle of a night wake up at
1:00 in the morning and switch on your
computer and what you told you to
yesterday I would still do that was
really cool
unfortunately that it didn't quite work
with people so you go to people and tell
them what to do and they don't do it
mm-hm
and they hate you for it or you do it
today and then you go a day later and
you stop doing it so you have to so then
a question really became how can you put
yourself in the brain of the of people
as opposed to computers and it has the
computers as super dumb then so dumb if
if people were as dumb as computers i
wouldnt want to walk with them mmm
but people are smart and people are
emotional and people have pride and
people have a spur a shion's so how can
i connect to that and that's the thing
where most of leadership just fails
because many many engineers turn manager
believe they can treat their team just
the same way I can treat your computer
and it just doesn't work this way it's
just really bad so how did how can i how
are can i connect to people and in turns
out as a college professor
the wonderful thing you do all the time
is to empower other people like your job
is to make your students look great
that's all you do you're the best coach
and it turns out if you do a fantastic
job is making a students look great they
actually love you and their parents love
you and they give you all the credit for
stuff you don't deserve since that all
my students who are smarter than me all
the great stuff invented at Stanford
versus their stuff not my stuff and they
give me credit and say oh Sebastian but
just making them feel good about
themselves so the question really is can
you take a team of people and what does
it take to make them to connect to what
they actually want in life and turn this
into product affection it turns out
every human being that I know has
incredibly good intention
I've really never really met a person
with bad intentions I believe every
person wants to contribute I think every
person I've met wants to help others
it's amazing how much of a urge we have
not to just help ourselves but to help
others so how can we empower people and
give them the right framework that they
can accomplish this if in moments when
it works it's magical because you'd see
the confluence of people being able to
make the world a better place
and driving enormous confidence and
pride out of this and that's when when
my environment works the best these are
moments where I can disappear for a
month and come back and things still
work it's very hard to accomplish but in
when it works is amazing so I agree very
much it's not often heard that most
people in the world have good intentions
at the core their intentions are good
and they're good people that's a
beautiful message it's not often heard
we make this mistake and this is a
friend of mine eggs water token us that
we we judge ourselves by our intentions
in others by the actions and I think the
the biggest skill I mean here in Silicon
Valley were full of Engineers I have
very little empathy and and I kind of
befuddled why it doesn't work for them
the biggest skill I think that that
people should acquire is to put
themselves into the position of the
other and listen and listen to what the
other has to say and they'd be shocked
how similar they are to themselves and
they might even be shocked how their own
actions don't reflect their intentions I
often have conversations with engineers
yes they look hey I love you doing a
great job and by the way what you just
did has the following effect are you
aware of that and then people would say
oh my god not I wasn't because my
intention was that they'd say yeah I
trust your intention you're a good human
being but just to help you in the future
if you keep expressing it that way then
people just hate you and I've had many
instances we say oh my god thank you for
telling me this because it wasn't my
intention to look like an idiot wasn't
my intention to help other people I just
didn't know how to do it
simply by the way there's a no-fail
carnegie 1936 how to make friends and
how to influence others has the entire
pipe or just read it and you're done and
usually apply it every day and I wish I
could I was good enough to apply it
every day but it says simple things
right like be positive remember previous
name smile and eventually have empathy
like really think that the person that
you hate and you think is an idiot is if
you just like yourself it's a person
who's struggling who means well and who
might need help and guess what you need
help I've recently spoken with Stephen
Schwarzman I'm not sure if you know who
that is but do so and he said I'm a list
no but he he said sort of to expand on
what you're saying that one of the
biggest things you can do is hear people
when they tell you what their problem is
and then help them with that problem
he says it's surprising how few people
actually listen to what troubles others
and because it's right there in front of
you and you can benefit the world the
most and in fact yourself and everybody
around you by just hearing the problems
and solving them I mean that's my my
little history of engineering that is
while I was engineering with computers I
didn't care all what the computers
problems for just I just volumize
everyone to do it and it doesn't work
with me you've become the mean say to do
AI do the opposite but let's return to
the comfortable world of engineering
thinking you can you tell me in broad
strokes in how you see it because you're
the course starting at the core of
driving it the technical evolution of
autonomous vehicles from the first DARPA
Grand Challenge to the incredible
success we see or the program you
started with Google self-driving car and
way more in the entire industry that
sprung up all the different kinds of
approaches debates and so on well the
idea of self-driving car goes back to
the 80s there was a team in Germany on
the team at Carnegie Mellon that it's
very pioneering work but back in the day
I'd say the computers were so efficient
that even the best professors and
engineers in the world basically stood
no chance it then folded into a phase
where the US government spent at least
half a million dollars that I could
count on research projects but the way
the procurement works a successful stack
of paper describing lots of stuff that
no one's ever gonna read was a
successful product of a research project
so so we trained our researchers to
produce lots of paper that all changed
for the DARPA Grand Challenge and I
really gotta credit the ingenious people
at DARPA and the US government in
Congress that took a complete new
funding model where they said that's not
fun effort let's fund outcomes and it
sounds way trivial but it there was no
tax code that allowed did the use of
congressional tax money for a price it
was all effort based so if you put in a
hundred dollars in you could charge 100
hours you put in a thousand dollars and
you could build a thousand hours by
shading the focus in city making the
price we don't pay you for development
we pray for the accomplishment
they drew in they automatically drew out
all these contractors who are used to
the drug of getting money power and they
drew in a whole bunch of new people and
these people are mostly crazy people
there were people who had a car and a
computer and they wanted to make a
million bucks the million bucks was the
official price money was then doubled
and they felt if I put my computer in my
car and program it I can be rich and it
was so awesome like like half the team's
there was a team that was surfer dudes
and they had like two surfers on the
vehicle and brought like these fashion
girls super cute girls like twin sisters
and and you could tell these guys were
not your common I felt very offended who
like gets all these big multi-million
and billion other countries from the US
government and and there was a great we
set the universities moved in I was very
fortunate at Stanford that I just
received tenure so I couldn't be fired
whenever I do otherwise I would have
done it and I had enough money to
finance this thing and I was able to
attract a lot of money from
from third parties and even car
companies moved in they kind of moved in
very quietly because they were super
scared to be embarrassed that they a car
would flip over but Ford was there and
Volkswagen was there and a few others
and GM was there so it kind of reset the
entire landscape of people and if you
look at who's a big name in suffering
cars today these were mostly people who
participated in those challenges
ok that's incredible can you just
comment quickly on your sense of lessons
learned from that kind of funding model
and the research that's going on
academia in terms of producing papers is
there something to be learned and and
scaled up bigger these having these
kinds of grand challenges that could
improve outcomes so I'm a big believer
in and focusing on kind of an end-to-end
system I'm a really big believer in an
insistence building I've always built
systems in my academic career even
though I love math and an abstract stuff
but it's all derived from the idea of
let's solve your problem and it's very
hard for me to be an academic
and say let me solve a component of a
problem like if someone this feels like
not monitoring logic or AI planning
systems where people believe that a
certain style of problem-solving
is the ultimate end objective and and I
would always turn it around and say hey
what problem put my grandmother care
about that doesn't understand computer
technology and doesn't want to
understand how could I make her love
what I do because only then do I have an
impact on the world
I can easily impress my colleagues
that's that's that that is much easier
but impressing my grandmother is very
very hard so I've always thought if I
can build a self-driving car and and my
grandmother can use it even after she
loses her driving privileges or Sheldon
can use it or we save maybe a million
lives a year they would be very
impressive and then there's so many
problems like these like there's a
problem of curing cancer or I'll if
twice as long once the problem is
defined of course I can solve it in
society like it takes sometimes tens of
thousands of people to to find a
solution there's no way you can fund an
army of ten thousand at Stanford so
you're going to be the prototype it's
bit of meaningful prototype and the
DARPA Grand Challenge was
beautiful because it told me what this
prototype had to do I didn't need to
think about what it had to do it is said
to read the rules and it was really
really beautiful and it's most beautiful
you think what academia could aspire to
is to build a prototype that's the
system's level that solves it gives you
an inkling that this problem could be
solved with this project that's all I
want to emphasize what academia really
is and I think people misunderstand it
first and foremost academia is a way to
educate young people first and foremost
the professor is an educator no matter
away what a small suburban college or
whether you are a Harvard or Stanford
professor that's not the way most people
think of themselves in academia because
we have this kind of competition going
on for citations and and publication
that's a measurable thing but that is
secondary to the primary purpose of
educating people to think now in terms
of research most of the great science
the great research comes out of
universities you can trace almost
everything back including Google to
universities so there's nothing we do
fundamentally broken here it's a it's a
good system and I think America has the
finest University system on the planet
we can talk about reach and how to reach
people outside the system it's a
different topic but the system would
serve as a good system if I had one wish
I would say it'd be really great if
there was more debate about what the
great big problems are on the side and
focus on those and most of them are
interdisciplinary unfortunately it's
very easy to fall into a inner
disciplinary viewpoint where your
problem is dictators but what your
closest colleagues believe the problem
is it's very hard to break out and say
why there's an entire new field of
problems so give an example um prior to
me working on self-driving cars I was a
roboticist in a machine learning expert
and I wrote books on robotics something
called probabilistic robotics the survey
methods driven kind of viewpoint of the
world I build robots that acted in
museums as tour guides that bug let
children around it's something that it's
time was moderately challenging when I
started working on cars several
colleagues told me Sebastian you're
destroying your career because in our
field of robotics cars are looked like
as a gimmick and they're not expressive
enough they can only push the throttle
and and in the brakes there's no
dexterity there's no complexity it's
just too simple and no one came to me
and said Wow if you solve that problem
you can save a million lives right
among all robotic problems that I've
seen in my life I would say the self
having car transportation Havana has the
most hope for society so how come the
robotics community was all over the
place and of us become because we
focused on methods and solutions and not
on problems like if you go around today
and ask your grandmother what bugs you
what really makes you upset
I challenge any academic and to do this
and then realize how far your research
is probably away from that today at the
very least that's a good thing for
academics they deliberate on the other
thing that's really nice in Silicon
Valley is Silicon Valley is full of
smart people outside academia right so
there's the Larry page's and magaz
archive books in the world who are
anywhere as smart or smarter than the
best academics I met in my life and what
they do is they they are at a different
level they build the systems they build
they build the customer-facing system
they built things that people can use
without technical education and they are
inspired by research they're inspired by
scientist they hire the best PhDs from
the best universities for a reason so I
think this kind of vertical integration
that between the real product the real
impact and the real thought the real
ideas there's actually working
surprisingly balanced Silicon Valley it
did not work as well in other places in
this nation so when I worked at Carnegie
Mellon we had the world's finest
computer science university but there
wasn't those people in Pittsburgh that
would be able to take these very fine
computer science ideas and turn them
into massive the impact for products
that symbiosis seemed to exist pretty
much only in Silicon Valley and maybe a
bit in Boston in Austin yeah with
Stanford that's it was it's really
really interesting
so if we look a little bit further on
from the the DARPA Grand Challenge and
the launch of the Google self-driving
car what do you see is the state the
challenges of autonomous vehicles as
they are now is actually achieving that
huge scale and having a huge impact on
society I'm extremely proud of what what
has been accomplished and again I'm
taking a lot of credit for the work for
us and I'm actually very optimistic and
and people have been kind of worrying is
it too fast as to slow I salute there
yet and so on it is actually quite an
interesting hard problem and in that a
self-driving car to build one that
manages 90% of the problems encountered
in everyday driving is easy we can
literally do this over a weekend to do
99% might take a month then there's 1%
left so 1% would mean that you still
have a fatal accident every week very
unacceptable so now you work on this 1%
and the 99% of there were any 1% is
actually still a relatively easy but now
you're down to like a hundredth of one
percent and it's still completely
unacceptable in terms of safety so the
variety of things you encounter are just
enormous and that gives me enormous
respect for human being available to
deal with the couch on the highway right
or the DNI headlight or the blown tire
that we'd never never been trained for
and all of a sudden I have to handle in
an emergency situation and often do very
very successfully it's amazing
from that perspective how safe driving
actually is given how many millions of
miles we drive every year in this
country we are now at a point where I
believed it in already is there and I've
seen it I've seen it in way more I've
seen it in activist engine crews and in
a number of companies in unvoyage where
vehicles not driving around and
basically flawlessly I able to drive
people around in limited scenarios in
fact you can go to Vegas today and order
a Seminole lift and if you got the right
setting off your app you'll be picked up
by a driverless car now there's still
safety drivers in there but that's a
fantastic way to kind of learn what the
limits of Technology today and there's
still some glitches
but the gifts have become very very rare
I think the next step is gonna be to
down cost it to harden it did that
entrapment it sends us are not quite an
automatic weights than that yet and then
you read about the business models to
really kind of go somewhere and make the
business case and the business case is
hard work it's not just oh my god we
have this capability people that's gonna
buy it you have to make it affordable
you have to give people that find the
social acceptance of people none of the
teams yet has been able to or gutsy
enough to drive around without a person
inside the car and that's that the next
magical hurdle will be able to send
these vehicles around completely empty
in traffic and I think I'm gonna wait
everyday wait for the news that vamo has
just done this so you know the
interesting you mentioned gutsy I mean
let me ask some maybe unanswerable
question may be edgy questions but in
terms of how much risk is required some
guts in terms of leadership style it
would be good to contrast approaches and
I don't think anyone knows what's right
but if we compare Tesla and way mo for
example Elon Musk and the way mo team
the there's slight differences in
approach so on the Elon side there's
more I don't know what the right word to
use but aggression in terms of
innovation and I'm way mo side there's
more sort of cautious safety focused
approach to the problem what do you
think it takes what leadership at which
moment is right which approach is right
look I'm I don't sit in either of those
teams so I'm unable to even verify like
somebody says correct right in the end
of the day every innovator in in that
space will face a fundamental dilemma
and I would say you could put aerospace
Titans into the same bucket yes which is
you have to balance public safety with
your drive to innovate and this country
in particular in states has a
plus your history of doing this very
successfully yet travel is what a
hundred times are safe per mile than
ground travel and then cars and there's
a reason for it because people have
found ways to be very methodological
about ensuring public safety while still
being able to make progress on important
aspects for example like yell and noise
and fuel consumption so I think that
those practices are pruned and they
actually work we live in a world safer
than ever before and yes they will
always be the provision that something
was wrong there's always the possibility
that someone makes a mistake or there's
an unexpected failure we can't never
guarantee to 100 percent absolute safety
other than just not doing it
but I think I'm very proud of the
history of of United States I mean we've
we've dealt with much more dangerous
technology like nuclear energy and kept
that safe too we have nuclear weapons
and we keep those safe so so we have
methods and procedures that really
balance these two things very very
successfully you've mentioned a lot of
great autonomous vehicle companies that
are taking sort of the level 4 level
file they jump in full autonomy or the
safety driver and take that kind of
approach and also through simulation and
so on there's also the approach that
Tesla autopilot is doing which is kind
of incrementally taking a level 2
vehicle and using machine learning and
learning from the driving of human
beings and trying to creep up trying to
incremental improve the system until
it's able to achieve level 4 autonomy so
perfect autonomy in certain kind of
geographical regions what are your
thoughts on these contrasting approaches
when suppose of all I I'm a very proud
Tesla and I literally used the autopilot
every day and it literally has kept me
safe is a beautiful technology
specifically for highway driving when
I'm slightly tired because then it turns
me into a much safer driver and that I'm
a hundred percent confident it's the
case
and tells us the right approach I think
that the biggest change I've seen since
I went away one team is is this thing
called deep learning deep learning was
was not a hot topic when I when I
started way more or Google suffering
cars it was there in fact we saw the
Google brain at the same time in Google
X so I invested in deep learning but
people didn't talk about it wasn't a hot
topic and nowadays there's a shift of
emphasis from a more geometric
perspective where you use geometric
sensors they give you a full 3d view
when you do a geometric reasoning about
all of this box over here might be a car
towards a more human like oh let's just
learn about it this looks like the thing
I've seen 10,000 times before so maybe
it's the same thing
machine learning perspective and that
has really put I think all these
approaches on steroids at Udacity we
teach a course in self-driving cars we
can in fact I think we'd be if credit is
over 20,000 or so people on self-driving
car skills so every every self-driving
car team in the world now uses our
engineers and in this course the very
first homework assignment is to do Lane
finding on images and lane finding
images for layman what this means is you
you put a camera into your car or you
open your eyes and even know where the
lane is right so so you can stay inside
the lane with your car humans can do
this super easily you just look and you
know where the line is just intuitively
for machines for long term of a super
heart because people would write these
kind of crazy rules if there's like
vineland marcus and he's for fight
really means this is not quite wide
enough so let's all it's not right or
maybe the Sun is trying so when the Sun
shines and this is right and this is a
straight line I missed quite a straight
line because the ball is curved and and
do we know that there's really six feet
between lane markings or not or 12 feet
whatever it is and now the very students
are doing they would take machine
learning so instead of like writing
these crazy rules for the lane marker is
they say let's take an hour driving and
label it and tell the vehicle this is
actually the lane by hand and then these
are examples and have the Machine find
its own rules but for lane markings are
and within 24 hours now every student
there's never done any programming for
in this space can write a perfect Lane
finder as good as the best commercial
line
and that's completely amazing to me
we've seen progress using machine
learning that completely Dwarfs anything
that I saw ten years ago yeah and just
as a side note the self-driving car
nanodegree the fact that you launch that
many years ago now maybe four years ago
three years ago three years ago is
incredible that it that's a great
example of system level thinking sort of
just taking an entire course I teach
each other solve entire problem I
definitely recommend people it's been
super popular and it's become actually
incredibly high quality we build it with
Mercedes and and and various other
companies in that space and we find that
engineers from Tesla and vamo are taking
it today
the insight was that two things one is
existing universities will be very slow
to move because the departmental ice and
there's no department for self-driving
cars so between Mickey and EE and
computer science getting these folks
together into one room is really really
hard and every professor listening he
ever know that probably agree to that
and secondly even if if all the great
universities just did this which none so
far has develop a curriculum in this
field it is just a few thousand students
they can partake because all the great
universities are super selective so how
about people in India how about people
in China or in the Middle East or
Indonesia or Africa right should those
be excluded from the skill of building
self-driving cars are there any dumber
than we are any less privileged and the
answer is we should just give everybody
the skill to build a self-driving car
because if we do this then we have like
a thousand self-driving car startups and
if 10 percent succeed that's like a
hundred that means hundred countries now
we have self-driving cars and be safer
it's kind of interesting to imagine
impossible to qualify but the number the
you know over a period of several
decades the impact that has like a
single course like a ripple effect of
society if you just recently thought the
Android and who was creator of cosmos
show it's interesting to think about how
many scientists that show launched yes
and so it's really
in terms of impact I can't imagine a
better course than the self-driving car
course that's you know the there's other
more specific disciplines like deep
learning and so on that Udacity is also
teaching but self-driving cars it's
really really interesting course yeah
and it came at the right moment it came
at a time when there were a bunch of
aqua huggers aqua hire as a acquisition
of a company not for its technology or
its products or business but for its
people
so aqua hire means maybe the company of
70 people they have no product yet but
they're super smart people and he pays
certain amount of money so I took back
the highest like GM Cruise and uber and
and others and did the math and said hey
how many people are there and how much
money was paid and as a lower bound is
tomato value of an self-driving car
engineer in these acquisitions to be at
least 10 million dollars right so think
about this you you get just have a skill
and you team up and build a company and
you're worth now is 10 million dollars I
mean it's kind of cool I mean but what
other thing could you do in life to be
worth 10 million dollars within a year
yeah amazing but to come back for a
moment on to deep learning and its
application in autonomous vehicles you
know what are your thoughts on Elon
Musk's statement provocative statement
perhaps that lighter is a crutch so this
geometric way of thinking about the
world maybe holding us back if what we
should instead be doing in this robotics
but in this particular space of
autonomous vehicles is using camera as a
primary sensor and using computer vision
or machine learning is the primary way
to look up to Commons I think first of
all we all know that people can drive
cars without light us in their hands
because we only have eyes and we most
you just use eyes for driving maybe we
use some other perception about our
bodies accelerations occasionally our
years certainly not our noses so that
the existence proof is there that eyes
must be sufficient in fact we could even
drive a car someone put a camera out and
then give us the camera image with known
agency you would be able to drive a car
and that way it the same way so cameras
sufficient secondly I really love the
idea that in in the Western world we
have many many different people trying
different hypotheses it's almost like an
anthill like if another Idol tries to
forage for food but you can sit there as
two ands and agree what the perfect path
is and then every single ant marches for
the most like the location of food is or
you can even just spread out and I
promise you the spread out solution will
be better because if the discussing
philosophical intellectual ends get it
wrong and they're all moving the wrong
direction they're gonna waste a day and
then you're gonna discuss again for
another week
whereas if all these ants go in a random
direction someone's gonna succeed and
you're gonna come back and claim victory
and get the Nobel Prize about everything
antipholus and then they'll march in the
same direction and that's great about
society that's great about the Western
society if you're not plant-based you're
not central base we don't have a Soviet
Union style central government that
tells us where to forge we just Forge we
start and seek or you get investor money
and go out and try it out and who knows
is gonna win I like it in your when you
look at the long term vision of
autonomous vehicles
do you see machine learning as
fundamentally being able to solve most
of the problems so learning from
experience I'd say we should be very
clear about what machine learning is and
is not and I think there's a lot of
confusion for this today is a technology
that can go through large databases of
repetitive patterns and find those
patterns so in example we did a study at
stand for two years ago where we applied
machine learning to detecting skin
cancer and images and we harvested or
built a data set of 129,000 skin photo
shots that were all had been biopsied
for what the actual situation was and
those included melanomas and carcinomas
also included rashes and other skin
conditions lesions and then we had a
network find those patterns and it was
by and large able to then detect skin
cancer with an iPhone as accurately as
the best board-certified Stanford level
dermatologist
we proved that now not this thing was
great in this one thing I'm finding skin
cancer but it couldn't drive a car so so
the difference to human intelligence as
we do all these many many things and we
can often learn from a very small data
set of experiences but as machines still
need very large data sets and things
should be very repetitive no that's
still super impactful because almost
everything we do is repetitive so that's
gonna we transform human labor but it's
not this almighty general intelligence
we're really far away from a system that
will exhibit general intelligence to
that end I actually commiserate the
naming a little bit because artificial
intelligence if you believe Hollywood is
immediately mixed into the idea of human
suppression and and machine superiority
I don't think that we don't see this in
my lifetime I don't think human
suppression is a good idea I don't see
it coming I don't see the technology
being there what I see instead is a very
pointed focused pattern recognition
technology that's able to extract
patterns relation large data sets and in
doing so it can be super impactful and
super impactful let's take the impact of
artificial terrorism on human work we
all know that it takes something like
10,000 hours to become an expert if
you're gonna be a doctor or lawyer or
even a really good driver it takes a
certain amount of time to become experts
machines now are able and have been
shown to observe people become experts
in observe experts and then extract
those rules from experts in some
interesting way they could go from law
to sales to driving cars to diagnosing
cancer and then giving that capability
to people who are completely new in
their job we now can and that's that's
been done has been done commercially in
many many instances that means we can
use machine learning to make people an
expert on the very first day of their
work like think about the impact if if
your doctor is still in the first 10,000
hours you have a doctor who's not quite
an expert yet
who would not want the doctor who is the
world's best expert and now we can
leverage machines to really eradicate
the error and decision making error and
lack of expertise for human doctors they
could save your life
if we can link on that for a little bit
in which way do you hope machines in the
medical in the medical field could help
assist doctors you mentioned this sort
of accelerating the learning curve or
people if they start a job or in the
first 10,000 hours can be assisted by a
machine how do you how do you envision
that assistance looking so we built this
this app for an iPhone that can detect
and classify and diagnose skin cancer
and we proved two years ago there is as
pretty much as good or better than the
best human docto