Video summary
In this discussion, Sebastian Thrun addresses the fundamental dilemma facing autonomous vehicle innovators: balancing public safety with the drive to innovate. He contrasts two distinct leadership approaches within the industry, noting that while Tesla under Elon Musk often displays an aggressive stance toward innovation, Waymo adopts a more cautious, safety-focused methodology. Thrun argues that both strategies are valid and necessary for progress in this sector, drawing parallels to the aerospace industry which has successfully managed similar risks over its century-long history. He emphasizes that while absolute zero risk is impossible—just as with nuclear energy or weapons—the goal is to balance these competing interests through rigorous procedures and methodological practices that have proven effective in creating a safer world than ever before. Thrun expresses strong personal support for Tesla's incremental approach, stating he uses Autopilot daily because it significantly enhances safety when drivers are tired. He highlights the technological shift from geometric reasoning to deep learning as a major catalyst for progress in self-driving cars. When Waymo and Google X initially relied on precise 3D mapping and geometric rules, Thrun observed that these methods were often brittle; they struggled with edge cases like curved lane markings or varying sunlight conditions because engineers had to write complex code for every scenario. In contrast, the rise of deep learning allows systems to learn from vast amounts of data in a human-like manner, recognizing patterns intuitively rather than relying on rigid pre-programmed rules. To demonstrate the power of this machine learning approach, Thrun describes his course at Udacity, which has educated over 20,000 students and trained engineers for nearly every self-driving car team globally. The curriculum focuses on tasks like lane finding using only camera images, a challenge that traditionally required extensive programming but can now be solved by labeling hours of driving data in just 24 hours. Students with no prior coding experience can build perfect lane finders within this timeframe, showcasing how machine learning dwarfs the capabilities seen even ten years ago. This capability allows vehicles to adapt and improve continuously without needing exhaustive manual rule-writing for every possible road condition or visual anomaly. Addressing Elon Musk's provocative claim that Autopilot is merely a "crutch," Thrun defends the sufficiency of camera-based perception systems by comparing them to human vision, noting that humans drive using only their eyes rather than other senses like smell or touch. He further praises the decentralized nature of Western innovation, likening it to an anthill where many independent entities try different hypotheses simultaneously without a central government dictating direction. In this competitive ecosystem, if one approach fails, others can succeed and claim victory, driving overall advancement more effectively than a centralized model that might waste time pursuing incorrect paths. Ultimately, Thrun believes that the diversity of approaches in companies like Tesla and Waymo ensures that society benefits from multiple successful innovations rather than relying on a single mandated solution.
Read the full video transcript
you know the interesting you mentioned
gutsy let me 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 Moe for example
Elon Musk and the way Moe 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 on
weibo 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 so what it
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 hundred plus year
history of doing this very successfully
yet travel is what a hundred times are
safe per mile then ground travel and
then cars and there's a reason for it
because people have found ways to be
very methodological about insuring
public safety while still being able to
make progress on important aspects for
example like yell and noise and 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 never
guarantee to 100% absolute safety other
than just not doing it but I think I'm
very proud of his
of the end states I mean we've we've
dealt with much more dangerous
technology like nuclear energy and and
kept that safe - 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 love of
four-level files a 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-two
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 four 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 in terms of the right approach I
think the 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 self-driving
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
over 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 ten thousand 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 infect
I think which we've if credits over
20,000 or so people on self-efficacy
kills so every every self-driving car
team in the world now use 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 oh you're
open your eyes and you would 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 shining so when the
Sun shines and this is right and this is
a straight line I missed quite a
straight line because the vote is curved
and and do we know that there's a six
feet between lane markings or not or
twelve feet whatever it is and now the
voted students are doing they would take
machine learning so instead of like
writing these crazy rules for the lane
marker is their 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 find us and
that's completely amazing to me we've
seen progress using machine learning
that completely Dwarfs anything that I
saw 10 years ago what are your thoughts
on Elon Musk's statement provocative
statement perhaps that light air is a
crutch so this geometric way of thinking
about the world may be 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 and 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 heads because we only have eyes
and we mostly 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 prove is there that eyes must
be sufficient in fact we could even
drive a car if someone put a camera out
and then give us the camera image with
known agency we would be able to drive a
car and that way it the same way so a
camera is also 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 a noun little tries to forage
for food right 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 of any
directions someone's gonna succeed and
you're gonna come back and and claim
victory and get the Nobel Prize about
everything and equivalent is and then
they will 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 in C Corp you get investor money
and go out and try it out and who knows
is gonna win
you