
[Music]
we are experiencing a renaissance of
robotics and artificial intelligence in
recent years we’ve seen computers drive
our cars and trucks translate and
interpret natural languages and defeat
the world’s champions of chess go and
jeopardy ibm’s watson is now or soon
will be the world’s leading oncology
expert and the new AI is fueling the
robot revolution in my own work i’ve
been able to contribute to this
revolution in areas including
manufacturing we’re feeling is required
for a mechanical assembly or grinding
and in autonomous vehicles which will
soon be commercial products offering us
greater safety and convenience in
machines that sense and interpret their
environments with applications such as
voice controlled smart wheelchairs and
in robotic li assisted minimally
invasive surgery with the promise of
faster more precise operations with
lower trauma and less risk of infection
as well as service robots such as tour
guides and sentries and in humanoid
robots that can operate in domestic
environments and can navigate over
difficult terrains and can operate
controls like bowels levers switches and
buttons or operate power tools designed
for humans or behaved as co-workers
collaborating with humans these results
from my lab and many others around the
world are now emerging into new robots
and emerging into our lives of course
there are applications in dull
monotonous and repetitive tasks but
there are also new applications in tasks
that require superhuman strength speed
precision reliability Total Recall or
unflinching focus of attention why is
this happening why now and how will it
change our lives as a career robot
assist I’ve lived through the robot
depression I used to advise my students
don’t call yourself a robot assist you
won’t get a job but in recent years I’ve
seen that turned upside down my students
are getting great jobs specifically for
the robotics expertise from my
perspective this has come about from two
important changes one
is the evolution of the new AI
specifically deep learning and the
second is a fundamental change in the
way that we program robots let me first
give you a little background on deep
learning what is it where did it come
from the foundations of deep learning
come from brain science we all know that
brains are made up of neurons
interconnected with synapses and from
there we generate all thought action
memories perception emotions feelings
but exactly how this is done has been a
long-standing challenge neuroscientists
has studied neurons down to the
molecular level and we understand in
detail how they communicate with release
of neurotransmitters and responsiveness
of ion channels scientists have even
demystified the act of learning by
tagging ion channels and watching their
migration through dendrites in an
installation into synapses this is
watching learning under a microscope the
physiological manifestation of learning
a biological change you by listening to
this talk are biologically changed and
we take our knowledge of these and we
wrap them up into mathematical models
and we program these into computer
simulations and so neurons are the
building blocks of brains and simulated
neurons are the building blocks of the
new AI but it’s not enough to understand
how neurons work we have to understand
how they’re interconnected how are their
synapses formed and for this we get
knowledge from biology for example the
horseshoe crab the horseshoe crab is an
ancient creature with primitive eyes and
scientists have been able to disentangle
the axons from their primitive light
sensors and found that they’re connected
in patterns creating the equivalent of
computing tissue that performs low-level
image processing these synapses are not
an accident or random these are dictated
by the DNA of the creature because these
synaptic values perform useful
computations and we learned more from
the Nobel prize-winning work of Hubel
and Wiesel studying the visual cortex of
the cat they found more of these innate
connections among neurons notably there
are collections of neurons in the retina
that can detect light dark changes like
the horseshoe crab and these bits
project onto deeper neurons that
assemble these into line segments at
different displacements and angles and
these project onto deep
so to assemble them into geometric
patterns and so we see not only are
there patterns of connections at
synapses but there’s also an
organization by layers of abstraction so
we’ve learned more about how at least
ensembles of neurons are connected
together but are we ready to build
brains maybe we get insight and
inspiration from biologists studying the
behavior of creatures an example is the
ghost crab the ghost crab has a behavior
in that it lives in a burrow in the sand
and emerges periodically to enter the
sea if you chase it it runs quickly to
the sanctuary of its burrow it appears
to have a strong reflex for
self-preservation and good strategies
for survival but if while the crab is
out you cover its burrow at the metal
plate and when it really chase it it
runs towards its burrow and it stops on
top of the plate and at this point you
can just walk over and pick it up and so
although the ghost crab has a biological
brain it reveals itself to be
indistinguishable from a robot it has a
program which is normally effective but
it has no room for creativity or
innovation and so we’re embolden to try
to build our primitive brains and we put
together what we’ve learned our
knowledge of neurons in simulations
patterns of interconnections in synapses
and layers of abstraction and these form
what are called
deep neural networks now the trick to
getting useful computations have it out
of a deep neural network is to discover
good synaptic weights like the horseshoe
crab the process of this discovery is
called deep learning now we knew about
many of these discoveries much earlier
but they were never successful until
recently and part of the reason is this
discovery of good synaptic values
requires intensive computations and
we’ve been helped recently by the
emergence of new computing hardware
notably the graphical processing unit or
GPU these were originally developed for
computer gaming but they were found to
be effective for the intensive
computations for discovery of synapses a
second critically important development
was the emergence of big data we all
know that companies such as Facebook
Amazon and Google are amassing huge
repositories of labeled information and
the ease of gathering data over the
Internet
leads to the creation of new
repositories springing up around the
world and this is essential for training
the new networks because they still
learn very slowly we require both
intensive computations and literally
millions of examples but the results
have been stunning now computers can
interpret images semantically they can
say that as a pedestrian that is a
bicyclist that is a car that is my Road
and this is necessary for effective
autonomous vehicles or for robots
behaving intelligently in factories or
in homes applications of deep learning
are springing up everywhere facial
recognition recognition of emotions
interpretation of natural languages even
patent searches and an avalanche of new
applications on the horizon and these
dramatically better brains are spurring
the rise of the robots the second
important development for the rise of
robots was a fundamental change in how
we program them as built into the robot
operating system or Ross in my career
I’ve seen decades of robot projects that
started from scratch and invested all of
their resources only to recreate results
of the past we were not building on
foundations we were not making progress
by reinventing the wheel we’re not
protecting our advances and reusing them
at integrating them with others and Ross
has changed all of this by bringing
modern software engineering techniques
to robots ross was innovated at
Stanford’s artificial intelligence lab
and was nurtured for nine years by
Google and it’s now emerged as the de
facto standard robot programming
environment around the world I was not a
founder of Ross but I am a convert I’m a
user a contributor a teacher an
evangelist most recently I’ve also
written a textbook in pursuit of my
desire to train and recruit the next
generation of robot assists the number
of robots the variety of robots using
Ross is exploding old robots are being
converted to Ross new robots are born
with Ross the existence of Ross has led
to the creation of new robot startup
companies and Ross has spread throughout
the world in academia in industry in
research labs and now we’re seeing
experts in Tokyo in Edinburgh in
Cleveland who
contributing their expertise in modules
like Lego blocks that can be assembled
in new novel systems in California in
Australia and in China and for the first
time we’re able to create robots that
are more sophisticated than any one
person or anyone research group could do
in a lifetime
dramatically better robots are coming
out and we can expect dramatic changes
in our life one of the things we expect
will make a difference is jobs a recent
McKinsey study estimated that half of
all human labor can be replaced by a
robot or machine enabled with artificial
intelligence they also identified over
70 professions in which 90% of human
content can be eliminated ranging from
stock pickers to truck drivers a recent
study showed that loss of manufacturing
jobs in the United States was
attributable to automation not
outsourcing robots are taking our
manufacturing jobs and it’s not just in
manufacturing it’s across the spectrum
for example in logistics like Amazon
automated distribution centers or in
design with 3d printers in home
automation in security and defense
agriculture and food health care energy
Elon Musk warns that there will be fewer
and fewer jobs that a robot cannot do
better than he human and Stephen Hawking
worries that there will only be jobs for
the most exceptional humans but
throughout modern history has been
predicted the technological advances
would lead to massive job layoffs and
historically this has not been true and
the 1800s the Luddites protested
automation in the textile industry by
smashing weaving machines but ultimately
these machines increased employment in
the industry and increased wages overall
the Industrial Revolution elevated the
lives of the common worker in the 20th
century United States we began with 40%
of our workforce involved in farming and
with introduction of technology this
fell to 2% by the end of the century but
there was not a massive jobs collapse
instead there was jobs growth there was
wealth increases we largely moved into
manufacturing including in sectors that
had been undreamed-of in the past and
the introduction of computer into
business was predicted to lead to
massively
and certainly businesses became more
efficient but there were not massive
layoffs there was increase in employment
and increased salaries today you can’t
run a business without a computer and
the existence of the computer is made
startups easier and as we lost our
manufacturing jobs we gained new jobs in
the service sector to a large extent we
gave up dull dirty and dangerous
manufacturing jobs for knowledge workers
and so we have reason to believe there
will not be a robot apocalypse but a
robot utopia a world where robots do all
of the dirty dull and dangerous work for
us and that humans working together with
robots increase the value of human
labour increased productivity and
increased wealth and we will still need
doctors therapists poets writers
researchers teachers musicians composers
comedians actors athletes dancers and
artists and countless others who elevate
and enrich our lives that we can expect
with our newfound wealth we’ll be able
to enjoy these arts and entertainment
and travel and leisure robots will
become our laborers and also our
assistants and our advisors our servants
and our Guardians
and our companions robots will become
commonplace and indispensable in our
lives and this will give birth to a new
basic industry and generate millions of
jobs involved in the creation of ever
more capable machines tomorrow’s robot
assists will face a fundamental grand
challenge to create robots in our image
but that ultimately surpass human
strength speed precision intelligence
and most importantly compassion and
ethics addressing this grand challenge
will require a collaboration among
contributors with greater
multidisciplinary breadth than any time
in history of course we’ll need more
engineers and computer scientists but
we’ll also need philosophers linguists
psychologists psychiatrists neuro
scientists biologists skill experts and
countless more knowledge workers the
result of this grand challenge will be a
transformational elevation of human life
and so yes the robot revolution is here
but it’s not to be feared but to be
embraced as our hope for the future
thank you
[Applause]
we are experiencing a renaissance of
robotics and artificial intelligence in
recent years we’ve seen computers drive
our cars and trucks translate and
interpret natural languages and defeat
the world’s champions of chess go and
jeopardy ibm’s watson is now or soon
will be the world’s leading oncology
expert and the new AI is fueling the
robot revolution in my own work i’ve
been able to contribute to this
revolution in areas including
manufacturing we’re feeling is required
for a mechanical assembly or grinding
and in autonomous vehicles which will
soon be commercial products offering us
greater safety and convenience in
machines that sense and interpret their
environments with applications such as
voice controlled smart wheelchairs and
in robotic li assisted minimally
invasive surgery with the promise of
faster more precise operations with
lower trauma and less risk of infection
as well as service robots such as tour
guides and sentries and in humanoid
robots that can operate in domestic
environments and can navigate over
difficult terrains and can operate
controls like bowels levers switches and
buttons or operate power tools designed
for humans or behaved as co-workers
collaborating with humans these results
from my lab and many others around the
world are now emerging into new robots
and emerging into our lives of course
there are applications in dull
monotonous and repetitive tasks but
there are also new applications in tasks
that require superhuman strength speed
precision reliability Total Recall or
unflinching focus of attention why is
this happening why now and how will it
change our lives as a career robot
assist I’ve lived through the robot
depression I used to advise my students
don’t call yourself a robot assist you
won’t get a job but in recent years I’ve
seen that turned upside down my students
are getting great jobs specifically for
the robotics expertise from my
perspective this has come about from two
important changes one
is the evolution of the new AI
specifically deep learning and the
second is a fundamental change in the
way that we program robots let me first
give you a little background on deep
learning what is it where did it come
from the foundations of deep learning
come from brain science we all know that
brains are made up of neurons
interconnected with synapses and from
there we generate all thought action
memories perception emotions feelings
but exactly how this is done has been a
long-standing challenge neuroscientists
has studied neurons down to the
molecular level and we understand in
detail how they communicate with release
of neurotransmitters and responsiveness
of ion channels scientists have even
demystified the act of learning by
tagging ion channels and watching their
migration through dendrites in an
installation into synapses this is
watching learning under a microscope the
physiological manifestation of learning
a biological change you by listening to
this talk are biologically changed and
we take our knowledge of these and we
wrap them up into mathematical models
and we program these into computer
simulations and so neurons are the
building blocks of brains and simulated
neurons are the building blocks of the
new AI but it’s not enough to understand
how neurons work we have to understand
how they’re interconnected how are their
synapses formed and for this we get
knowledge from biology for example the
horseshoe crab the horseshoe crab is an
ancient creature with primitive eyes and
scientists have been able to disentangle
the axons from their primitive light
sensors and found that they’re connected
in patterns creating the equivalent of
computing tissue that performs low-level
image processing these synapses are not
an accident or random these are dictated
by the DNA of the creature because these
synaptic values perform useful
computations and we learned more from
the Nobel prize-winning work of Hubel
and Wiesel studying the visual cortex of
the cat they found more of these innate
connections among neurons notably there
are collections of neurons in the retina
that can detect light dark changes like
the horseshoe crab and these bits
project onto deeper neurons that
assemble these into line segments at
different displacements and angles and
these project onto deep
so to assemble them into geometric
patterns and so we see not only are
there patterns of connections at
synapses but there’s also an
organization by layers of abstraction so
we’ve learned more about how at least
ensembles of neurons are connected
together but are we ready to build
brains maybe we get insight and
inspiration from biologists studying the
behavior of creatures an example is the
ghost crab the ghost crab has a behavior
in that it lives in a burrow in the sand
and emerges periodically to enter the
sea if you chase it it runs quickly to
the sanctuary of its burrow it appears
to have a strong reflex for
self-preservation and good strategies
for survival but if while the crab is
out you cover its burrow at the metal
plate and when it really chase it it
runs towards its burrow and it stops on
top of the plate and at this point you
can just walk over and pick it up and so
although the ghost crab has a biological
brain it reveals itself to be
indistinguishable from a robot it has a
program which is normally effective but
it has no room for creativity or
innovation and so we’re embolden to try
to build our primitive brains and we put
together what we’ve learned our
knowledge of neurons in simulations
patterns of interconnections in synapses
and layers of abstraction and these form
what are called
deep neural networks now the trick to
getting useful computations have it out
of a deep neural network is to discover
good synaptic weights like the horseshoe
crab the process of this discovery is
called deep learning now we knew about
many of these discoveries much earlier
but they were never successful until
recently and part of the reason is this
discovery of good synaptic values
requires intensive computations and
we’ve been helped recently by the
emergence of new computing hardware
notably the graphical processing unit or
GPU these were originally developed for
computer gaming but they were found to
be effective for the intensive
computations for discovery of synapses a
second critically important development
was the emergence of big data we all
know that companies such as Facebook
Amazon and Google are amassing huge
repositories of labeled information and
the ease of gathering data over the
Internet
leads to the creation of new
repositories springing up around the
world and this is essential for training
the new networks because they still
learn very slowly we require both
intensive computations and literally
millions of examples but the results
have been stunning now computers can
interpret images semantically they can
say that as a pedestrian that is a
bicyclist that is a car that is my Road
and this is necessary for effective
autonomous vehicles or for robots
behaving intelligently in factories or
in homes applications of deep learning
are springing up everywhere facial
recognition recognition of emotions
interpretation of natural languages even
patent searches and an avalanche of new
applications on the horizon and these
dramatically better brains are spurring
the rise of the robots the second
important development for the rise of
robots was a fundamental change in how
we program them as built into the robot
operating system or Ross in my career
I’ve seen decades of robot projects that
started from scratch and invested all of
their resources only to recreate results
of the past we were not building on
foundations we were not making progress
by reinventing the wheel we’re not
protecting our advances and reusing them
at integrating them with others and Ross
has changed all of this by bringing
modern software engineering techniques
to robots ross was innovated at
Stanford’s artificial intelligence lab
and was nurtured for nine years by
Google and it’s now emerged as the de
facto standard robot programming
environment around the world I was not a
founder of Ross but I am a convert I’m a
user a contributor a teacher an
evangelist most recently I’ve also
written a textbook in pursuit of my
desire to train and recruit the next
generation of robot assists the number
of robots the variety of robots using
Ross is exploding old robots are being
converted to Ross new robots are born
with Ross the existence of Ross has led
to the creation of new robot startup
companies and Ross has spread throughout
the world in academia in industry in
research labs and now we’re seeing
experts in Tokyo in Edinburgh in
Cleveland who
contributing their expertise in modules
like Lego blocks that can be assembled
in new novel systems in California in
Australia and in China and for the first
time we’re able to create robots that
are more sophisticated than any one
person or anyone research group could do
in a lifetime
dramatically better robots are coming
out and we can expect dramatic changes
in our life one of the things we expect
will make a difference is jobs a recent
McKinsey study estimated that half of
all human labor can be replaced by a
robot or machine enabled with artificial
intelligence they also identified over
70 professions in which 90% of human
content can be eliminated ranging from
stock pickers to truck drivers a recent
study showed that loss of manufacturing
jobs in the United States was
attributable to automation not
outsourcing robots are taking our
manufacturing jobs and it’s not just in
manufacturing it’s across the spectrum
for example in logistics like Amazon
automated distribution centers or in
design with 3d printers in home
automation in security and defense
agriculture and food health care energy
Elon Musk warns that there will be fewer
and fewer jobs that a robot cannot do
better than he human and Stephen Hawking
worries that there will only be jobs for
the most exceptional humans but
throughout modern history has been
predicted the technological advances
would lead to massive job layoffs and
historically this has not been true and
the 1800s the Luddites protested
automation in the textile industry by
smashing weaving machines but ultimately
these machines increased employment in
the industry and increased wages overall
the Industrial Revolution elevated the
lives of the common worker in the 20th
century United States we began with 40%
of our workforce involved in farming and
with introduction of technology this
fell to 2% by the end of the century but
there was not a massive jobs collapse
instead there was jobs growth there was
wealth increases we largely moved into
manufacturing including in sectors that
had been undreamed-of in the past and
the introduction of computer into
business was predicted to lead to
massively
and certainly businesses became more
efficient but there were not massive
layoffs there was increase in employment
and increased salaries today you can’t
run a business without a computer and
the existence of the computer is made
startups easier and as we lost our
manufacturing jobs we gained new jobs in
the service sector to a large extent we
gave up dull dirty and dangerous
manufacturing jobs for knowledge workers
and so we have reason to believe there
will not be a robot apocalypse but a
robot utopia a world where robots do all
of the dirty dull and dangerous work for
us and that humans working together with
robots increase the value of human
labour increased productivity and
increased wealth and we will still need
doctors therapists poets writers
researchers teachers musicians composers
comedians actors athletes dancers and
artists and countless others who elevate
and enrich our lives that we can expect
with our newfound wealth we’ll be able
to enjoy these arts and entertainment
and travel and leisure robots will
become our laborers and also our
assistants and our advisors our servants
and our Guardians
and our companions robots will become
commonplace and indispensable in our
lives and this will give birth to a new
basic industry and generate millions of
jobs involved in the creation of ever
more capable machines tomorrow’s robot
assists will face a fundamental grand
challenge to create robots in our image
but that ultimately surpass human
strength speed precision intelligence
and most importantly compassion and
ethics addressing this grand challenge
will require a collaboration among
contributors with greater
multidisciplinary breadth than any time
in history of course we’ll need more
engineers and computer scientists but
we’ll also need philosophers linguists
psychologists psychiatrists neuro
scientists biologists skill experts and
countless more knowledge workers the
result of this grand challenge will be a
transformational elevation of human life
and so yes the robot revolution is here
but it’s not to be feared but to be
embraced as our hope for the future
thank you
[Applause]
Please follow and like us: