our world is changing in many ways and one of the things which is going to have
one of the things which is going to have
one of the things which is going to have a huge impact on our future is
a huge impact on our future is
a huge impact on our future is artificial intelligence AI bringing
artificial intelligence AI bringing
artificial intelligence AI bringing another Industrial Revolution previous
another Industrial Revolution previous
another Industrial Revolution previous industrial revolutions expended humans
industrial revolutions expended humans
industrial revolutions expended humans mechanical power this new revolution
mechanical power this new revolution
mechanical power this new revolution this second Machine Age is going to
this second Machine Age is going to
this second Machine Age is going to expand our cognitive abilities our
expand our cognitive abilities our
expand our cognitive abilities our mental power computers are not just
mental power computers are not just
mental power computers are not just going to replace manual labor but also
going to replace manual labor but also
going to replace manual labor but also mental labor so where do we stand today
mental labor so where do we stand today
mental labor so where do we stand today you may have heard about what happened
you may have heard about what happened
you may have heard about what happened last March when a machine learning
last March when a machine learning
last March when a machine learning system called alphago used deep learning
system called alphago used deep learning
system called alphago used deep learning to beat the world champion at the game
to beat the world champion at the game
to beat the world champion at the game of Go Go is an ancient Chinese game
of Go Go is an ancient Chinese game
of Go Go is an ancient Chinese game which had been much more difficult for
which had been much more difficult for
which had been much more difficult for computers to master than the game of
computers to master than the game of
computers to master than the game of chess how did we succeed now after
chess how did we succeed now after
chess how did we succeed now after decades of AI research alphago was
decades of AI research alphago was
decades of AI research alphago was trained to play go first by watching
trained to play go first by watching
trained to play go first by watching over and over tens of millions of moves
over and over tens of millions of moves
over and over tens of millions of moves made by very strong human players then
made by very strong human players then
made by very strong human players then by playing against itself millions of
by playing against itself millions of
by playing against itself millions of games machine learning allows computers
games machine learning allows computers
games machine learning allows computers to learn from examples to learn from
to learn from examples to learn from
to learn from examples to learn from data machine learning has turned out to
data machine learning has turned out to
data machine learning has turned out to be a key to cram knowledge into
be a key to cram knowledge into
be a key to cram knowledge into computers and this is important
computers and this is important
computers and this is important because knowledge is what enables
because knowledge is what enables
because knowledge is what enables intelligence putting knowledge into
intelligence putting knowledge into
intelligence putting knowledge into computers had been a challenge for
computers had been a challenge for
computers had been a challenge for previous approaches to AI why there are
previous approaches to AI why there are
previous approaches to AI why there are many things which we know intuitively so
many things which we know intuitively so
many things which we know intuitively so we cannot communicate them verbally we
we cannot communicate them verbally we
we cannot communicate them verbally we do not have conscious access to that
do not have conscious access to that
do not have conscious access to that intuitive knowledge how can we program
intuitive knowledge how can we program
intuitive knowledge how can we program computers with that knowledge what’s the
computers with that knowledge what’s the
computers with that knowledge what’s the solution the solution is for machines to
solution the solution is for machines to
solution the solution is for machines to learn that knowledge by themselves just
learn that knowledge by themselves just
learn that knowledge by themselves just as we do this is important because
as we do this is important because
as we do this is important because knowledge is what enables intelligence
knowledge is what enables intelligence
knowledge is what enables intelligence my mission has been to contribute to
my mission has been to contribute to
my mission has been to contribute to discover and understand principles of
discover and understand principles of
discover and understand principles of intelligence through learning whether
intelligence through learning whether
intelligence through learning whether animal or human or machine learning I
animal or human or machine learning I
animal or human or machine learning I and others believe that there are a few
and others believe that there are a few
and others believe that there are a few key principles just like the laws of
key principles just like the laws of
key principles just like the laws of physics simple principles which could
physics simple principles which could
physics simple principles which could explain our own intelligence and help us
explain our own intelligence and help us
explain our own intelligence and help us build intelligent machines for example
build intelligent machines for example
build intelligent machines for example think about the laws of aerodynamics
think about the laws of aerodynamics
think about the laws of aerodynamics which are general enough to explain the
which are general enough to explain the
which are general enough to explain the flight of both birds and planes wouldn’t
flight of both birds and planes wouldn’t
flight of both birds and planes wouldn’t it be amazing to discover such simple
it be amazing to discover such simple
it be amazing to discover such simple but powerful principles that would
but powerful principles that would
but powerful principles that would explain intelligence itself well we’ve
explain intelligence itself well we’ve
explain intelligence itself well we’ve made some progress my collaborators and
made some progress my collaborators and
made some progress my collaborators and I have contributed in recent years in a
I have contributed in recent years in a
I have contributed in recent years in a revolution in AI with our research on
revolution in AI with our research on
revolution in AI with our research on neural networks and deep learning an
neural networks and deep learning an
neural networks and deep learning an approach to machine learning which is
approach to machine learning which is
approach to machine learning which is inspired by the brain
inspired by the brain
inspired by the brain it started with speech recognition
it started with speech recognition
it started with speech recognition on your phones with neural networks
on your phones with neural networks
on your phones with neural networks since 2012 shortly after came a
since 2012 shortly after came a
since 2012 shortly after came a breakthrough in computer vision
breakthrough in computer vision
breakthrough in computer vision computers can now do a pretty good job
computers can now do a pretty good job
computers can now do a pretty good job of recognizing the content of images in
of recognizing the content of images in
of recognizing the content of images in fact approaching human performance on
fact approaching human performance on
fact approaching human performance on some benchmarks over the last five years
some benchmarks over the last five years
some benchmarks over the last five years a computer can now get an intuitive
a computer can now get an intuitive
a computer can now get an intuitive understanding of the visual appearance
understanding of the visual appearance
understanding of the visual appearance of a go board that is comparable to that
of a go board that is comparable to that
of a go board that is comparable to that of the best human players more recently
of the best human players more recently
of the best human players more recently following some discoveries made in my
following some discoveries made in my
following some discoveries made in my lab deep learning has been used to
lab deep learning has been used to
lab deep learning has been used to translate from one language to another
translate from one language to another
translate from one language to another and you know start seeing this in Google
and you know start seeing this in Google
and you know start seeing this in Google Translate this is expanding the
Translate this is expanding the
Translate this is expanding the computer’s ability to understand and
computer’s ability to understand and
computer’s ability to understand and generate natural language but don’t be
generate natural language but don’t be
generate natural language but don’t be fooled we are still very very far from a
fooled we are still very very far from a
fooled we are still very very far from a machine that would be as able as humans
machine that would be as able as humans
machine that would be as able as humans to learn to master many aspects of our
to learn to master many aspects of our
to learn to master many aspects of our world so let’s take an example even a
world so let’s take an example even a
world so let’s take an example even a two-year-old child is able to learn
two-year-old child is able to learn
two-year-old child is able to learn things in a way that computers are not
things in a way that computers are not
things in a way that computers are not able to do right now a two-year-old
able to do right now a two-year-old
able to do right now a two-year-old child actually masters intuitive physics
child actually masters intuitive physics
child actually masters intuitive physics she knows when she drops a ball that it
she knows when she drops a ball that it
she knows when she drops a ball that it is going to fall down when she spilled
is going to fall down when she spilled
is going to fall down when she spilled some liquids she expects the resulting
some liquids she expects the resulting
some liquids she expects the resulting mess her parents do not need to teach
mess her parents do not need to teach
mess her parents do not need to teach her about Newton’s laws or differential
her about Newton’s laws or differential
her about Newton’s laws or differential equations she discovers all these things
equations she discovers all these things
equations she discovers all these things by herself in an unsupervised way own
by herself in an unsupervised way own
by herself in an unsupervised way own supervised learning actually remains one
supervised learning actually remains one
supervised learning actually remains one of the key challenges for AI and it may
of the key challenges for AI and it may
of the key challenges for AI and it may take several more decades of fundamental
take several more decades of fundamental
take several more decades of fundamental research to crack that not unsupervised
research to crack that not unsupervised
research to crack that not unsupervised learning is actually trying
learning is actually trying
learning is actually trying to discover representations of the data
to discover representations of the data
to discover representations of the data let me show you an example consider a
let me show you an example consider a
let me show you an example consider a page on the screen that you’re seeing
page on the screen that you’re seeing
page on the screen that you’re seeing with your eyes or that the computer is
with your eyes or that the computer is
with your eyes or that the computer is seeing as an image a bunch of pixels in
seeing as an image a bunch of pixels in
seeing as an image a bunch of pixels in order to answer a question about the
order to answer a question about the
order to answer a question about the content of the image you need to
content of the image you need to
content of the image you need to understand its high-level meaning and
understand its high-level meaning and
understand its high-level meaning and this high level meaning corresponds to
this high level meaning corresponds to
this high level meaning corresponds to the highest level of representation in
the highest level of representation in
the highest level of representation in your brain lower down you have the
your brain lower down you have the
your brain lower down you have the individual meaning of words and even
individual meaning of words and even
individual meaning of words and even lower down you have characters which
lower down you have characters which
lower down you have characters which make up the words and those characters
make up the words and those characters
make up the words and those characters could be rendered in different ways with
could be rendered in different ways with
could be rendered in different ways with different strokes that make up the
different strokes that make up the
different strokes that make up the characters and those strokes are made up
characters and those strokes are made up
characters and those strokes are made up of edges and those edges are made up of
of edges and those edges are made up of
of edges and those edges are made up of pixels so these are different levels of
pixels so these are different levels of
pixels so these are different levels of representation but the pixels are not
representation but the pixels are not
representation but the pixels are not sufficient by themselves to make sense
sufficient by themselves to make sense
sufficient by themselves to make sense of the image to answer a high-level
of the image to answer a high-level
of the image to answer a high-level question about the content of the page
question about the content of the page
question about the content of the page your brain actually has these different
your brain actually has these different
your brain actually has these different levels of representation starting with
levels of representation starting with
levels of representation starting with neurons in the first visual area of
neurons in the first visual area of
neurons in the first visual area of cortex v1 which recognize edges and then
cortex v1 which recognize edges and then
cortex v1 which recognize edges and then neurons in the second visual area of
neurons in the second visual area of
neurons in the second visual area of cortex v2 which recognize strokes and
cortex v2 which recognize strokes and
cortex v2 which recognize strokes and small shapes higher up you have neurons
small shapes higher up you have neurons
small shapes higher up you have neurons which detect parts of objects and then
which detect parts of objects and then
which detect parts of objects and then objects and full scenes mule networks
objects and full scenes mule networks
objects and full scenes mule networks when they are trained with images can
when they are trained with images can
when they are trained with images can actually discover these types of levels
actually discover these types of levels
actually discover these types of levels of representation that match pretty well
of representation that match pretty well
of representation that match pretty well what we observe in the brain both
what we observe in the brain both
what we observe in the brain both biological neural networks which are
biological neural networks which are
biological neural networks which are what you have in your brain and the deep
what you have in your brain and the deep
what you have in your brain and the deep neural networks that we train on our
neural networks that we train on our
neural networks that we train on our machines can learn to transform from one
machines can learn to transform from one
machines can learn to transform from one level of representation to the next with
level of representation to the next with
level of representation to the next with the higher levels corresponding to
the higher levels corresponding to
the higher levels corresponding to more abstract notions for example the
more abstract notions for example the
more abstract notions for example the abstract notion of the character a can
abstract notion of the character a can
abstract notion of the character a can be rendered in many different ways at
be rendered in many different ways at
be rendered in many different ways at the lowest levels as many different
the lowest levels as many different
the lowest levels as many different configurations of pixels depending on
configurations of pixels depending on
configurations of pixels depending on your the position rotation font and so
your the position rotation font and so
your the position rotation font and so on so how do we learn these high levels
on so how do we learn these high levels
on so how do we learn these high levels of representations one thing that has
of representations one thing that has
of representations one thing that has been very successful up to now in the
been very successful up to now in the
been very successful up to now in the applications of deep learning is what we
applications of deep learning is what we
applications of deep learning is what we call supervised learning with supervised
call supervised learning with supervised
call supervised learning with supervised learning the computer needs to be taken
learning the computer needs to be taken
learning the computer needs to be taken by the hand and humans have to tell the
by the hand and humans have to tell the
by the hand and humans have to tell the computer the answer to many questions
computer the answer to many questions
computer the answer to many questions for example on millions and millions of
for example on millions and millions of
for example on millions and millions of images humans have to tell the Machine
images humans have to tell the Machine
images humans have to tell the Machine well for this image it is a cat for this
well for this image it is a cat for this
well for this image it is a cat for this image it is a dog for this image it is a
image it is a dog for this image it is a
image it is a dog for this image it is a laptop for this image is the keyboard
laptop for this image is the keyboard
laptop for this image is the keyboard and so on and so on millions of times
and so on and so on millions of times
and so on and so on millions of times this is very painful and we use
this is very painful and we use
this is very painful and we use crowdsourcing to manage to do that
crowdsourcing to manage to do that
crowdsourcing to manage to do that although this is very powerful and we
although this is very powerful and we
although this is very powerful and we are already able to solve very
are already able to solve very
are already able to solve very interesting problems humans are much
interesting problems humans are much
interesting problems humans are much stronger and they can learn over many
stronger and they can learn over many
stronger and they can learn over many more different aspects of the world in a
more different aspects of the world in a
more different aspects of the world in a much more autonomous way just as we’ve
much more autonomous way just as we’ve
much more autonomous way just as we’ve seen with the two-year-old child
seen with the two-year-old child
seen with the two-year-old child learning about intuitive physics
learning about intuitive physics
learning about intuitive physics unsupervised learning could also help us
unsupervised learning could also help us
unsupervised learning could also help us deal with self-driving cars let me
deal with self-driving cars let me
deal with self-driving cars let me explain what I mean
explain what I mean
explain what I mean unsupervised learning allows computers
unsupervised learning allows computers
unsupervised learning allows computers to project themselves into the future to
to project themselves into the future to
to project themselves into the future to generate plausible futures conditioned
generate plausible futures conditioned
generate plausible futures conditioned on the current situation and that allows
on the current situation and that allows
on the current situation and that allows computers to reason and to plan ahead
computers to reason and to plan ahead
computers to reason and to plan ahead even for circumstances that they have
even for circumstances that they have
even for circumstances that they have not been trained on this is important
not been trained on this is important
not been trained on this is important because if we use supervised learning we
because if we use supervised learning we
because if we use supervised learning we would have to tell the computers about
would have to tell the computers about
would have to tell the computers about all the circumstances where the car
all the circumstances where the car
all the circumstances where the car could be and that you know how human
could be and that you know how human
could be and that you know how human would react in that situation how did I
would react in that situation how did I
would react in that situation how did I learn to avoid dangerous
learn to avoid dangerous
learn to avoid dangerous fighting behavior did I have to die a
fighting behavior did I have to die a
fighting behavior did I have to die a thousand times in an accident well
thousand times in an accident well
thousand times in an accident well that’s the way we’re trying machines
that’s the way we’re trying machines
that’s the way we’re trying machines right now so it’s not going to fly or at
right now so it’s not going to fly or at
right now so it’s not going to fly or at least not to drive so what we need is to
least not to drive so what we need is to
least not to drive so what we need is to train our models to be able to generate
train our models to be able to generate
train our models to be able to generate plausible images plausible futures be
plausible images plausible futures be
plausible images plausible futures be creative and we’re making progress with
creative and we’re making progress with
creative and we’re making progress with that so we are training these deep
that so we are training these deep
that so we are training these deep neural networks to go from high-level
neural networks to go from high-level
neural networks to go from high-level meaning to pixels rather than from
meaning to pixels rather than from
meaning to pixels rather than from pixels to high-level of meaning so going
pixels to high-level of meaning so going
pixels to high-level of meaning so going in the other direction through the
in the other direction through the
in the other direction through the levels of representation in this way the
levels of representation in this way the
levels of representation in this way the computer can generate images there are
computer can generate images there are
computer can generate images there are new images different from what the
new images different from what the
new images different from what the computer has seen while it was trained
computer has seen while it was trained
computer has seen while it was trained but are plausible that looked like
but are plausible that looked like
but are plausible that looked like natural images we can also use these
natural images we can also use these
natural images we can also use these models to dream up strange sometimes
models to dream up strange sometimes
models to dream up strange sometimes scary images just like our dreams and
scary images just like our dreams and
scary images just like our dreams and nightmares here are some images that
nightmares here are some images that
nightmares here are some images that were synthesized by the computer using
were synthesized by the computer using
were synthesized by the computer using these deep genitive models they look
these deep genitive models they look
these deep genitive models they look like natural images but if you look
like natural images but if you look
like natural images but if you look closely you’ll see there are different
closely you’ll see there are different
closely you’ll see there are different and they’re still missing some of the
and they’re still missing some of the
and they’re still missing some of the important details that we would
important details that we would
important details that we would recognize as natural about ten years ago
recognize as natural about ten years ago
recognize as natural about ten years ago unsupervised learning has been a key to
unsupervised learning has been a key to
unsupervised learning has been a key to the breakthrough that we obtained
the breakthrough that we obtained
the breakthrough that we obtained discovering deep learning this was
discovering deep learning this was
discovering deep learning this was happening in just a few labs including
happening in just a few labs including
happening in just a few labs including mine at a time when neural networks were
mine at a time when neural networks were
mine at a time when neural networks were not popular
not popular
not popular they were almost abandoned by the
they were almost abandoned by the
they were almost abandoned by the scientific community now things have
scientific community now things have
scientific community now things have changed a lot it has become a very hot
changed a lot it has become a very hot
changed a lot it has become a very hot field there are now hundreds of students
field there are now hundreds of students
field there are now hundreds of students every year applying for graduate studies
every year applying for graduate studies
every year applying for graduate studies at my lab with my collaborators Montreal
at my lab with my collaborators Montreal
at my lab with my collaborators Montreal has become the largest academic
has become the largest academic
has become the largest academic concentration of deep learning
concentration of deep learning
concentration of deep learning researchers in the world
researchers in the world
researchers in the world we just received a huge research grant
we just received a huge research grant
we just received a huge research grant of 94 million dollars to push the
of 94 million dollars to push the
of 94 million dollars to push the boundaries of AI and data science and
boundaries of AI and data science and
boundaries of AI and data science and also to transfer technology of deep
also to transfer technology of deep
also to transfer technology of deep learning and data science to industry
learning and data science to industry
learning and data science to industry businesspeople stimulated by all this
businesspeople stimulated by all this
businesspeople stimulated by all this are creating startups industrial labs
are creating startups industrial labs
are creating startups industrial labs many of which near the universities for
many of which near the universities for
many of which near the universities for example just a few weeks ago we
example just a few weeks ago we
example just a few weeks ago we announced the launch of a startup
announced the launch of a startup
announced the launch of a startup factory called element AI which is going
factory called element AI which is going
factory called element AI which is going to focus on deep learning applications
there is just not enough deep learning
there is just not enough deep learning experts so they’re getting paid crazy
experts so they’re getting paid crazy
experts so they’re getting paid crazy salaries and many of my former academic
salaries and many of my former academic
salaries and many of my former academic colleagues have accepted generous deals
colleagues have accepted generous deals
colleagues have accepted generous deals from companies to work in industrial
from companies to work in industrial
from companies to work in industrial labs I for myself have chosen to stay in
labs I for myself have chosen to stay in
labs I for myself have chosen to stay in university to work for the public good
university to work for the public good
university to work for the public good to work with students to remain
to work with students to remain
to work with students to remain independent to guide the next generation
independent to guide the next generation
independent to guide the next generation of deep learning experts one thing that
of deep learning experts one thing that
of deep learning experts one thing that we’re doing beyond commercial value is
we’re doing beyond commercial value is
we’re doing beyond commercial value is thinking about the social implications
thinking about the social implications
thinking about the social implications of AI many of us are now starting to
of AI many of us are now starting to
of AI many of us are now starting to turn our eyes towards social value added
turn our eyes towards social value added
turn our eyes towards social value added applications like health we think that
applications like health we think that
applications like health we think that we can use deep learning to improve
we can use deep learning to improve
we can use deep learning to improve treatment with personalized medicine I
treatment with personalized medicine I
treatment with personalized medicine I believe that in the future as we collect
believe that in the future as we collect
believe that in the future as we collect more data from millions and billions of
more data from millions and billions of
more data from millions and billions of people around the earth we’ll be able to
people around the earth we’ll be able to
people around the earth we’ll be able to provide medical advice to billions of
provide medical advice to billions of
provide medical advice to billions of people who don’t have access to it right
people who don’t have access to it right
people who don’t have access to it right now and we can imagine many other
now and we can imagine many other
now and we can imagine many other applications for social value of AI for
applications for social value of AI for
applications for social value of AI for example something that will come out of
example something that will come out of
example something that will come out of our research on natural language
our research on natural language
our research on natural language understanding is providing all kinds of
understanding is providing all kinds of
understanding is providing all kinds of services like legal services to
services like legal services to
services like legal services to who can’t afford them we are now turning
who can’t afford them we are now turning
who can’t afford them we are now turning our eyes also towards the social
our eyes also towards the social
our eyes also towards the social implications of AI in my community but
implications of AI in my community but
implications of AI in my community but it’s not just for experts to think about
it’s not just for experts to think about
it’s not just for experts to think about this
this
this I believe that beyond the math and the
I believe that beyond the math and the
I believe that beyond the math and the jargon ordinary people can get a sense
jargon ordinary people can get a sense
jargon ordinary people can get a sense of what goes on under the hood enough to
of what goes on under the hood enough to
of what goes on under the hood enough to participate in the important decisions
participate in the important decisions
participate in the important decisions that will take place in the next few
that will take place in the next few
that will take place in the next few years and decades about AI so please set
years and decades about AI so please set
years and decades about AI so please set aside your fees and give yourself some
aside your fees and give yourself some
aside your fees and give yourself some space to learn about it my collaborators
space to learn about it my collaborators
space to learn about it my collaborators and I have written several introductory
and I have written several introductory
and I have written several introductory papers and a book entitled deep learning
papers and a book entitled deep learning
papers and a book entitled deep learning to help students and engineers jump into
to help students and engineers jump into
to help students and engineers jump into this exciting field there are also many
this exciting field there are also many
this exciting field there are also many online resources software tutorials
online resources software tutorials
online resources software tutorials videos and many undergraduate students
videos and many undergraduate students
videos and many undergraduate students are learning a lot of this about
are learning a lot of this about
are learning a lot of this about research and deep learning by themselves
research and deep learning by themselves
research and deep learning by themselves to later join the ranks of labs like
to later join the ranks of labs like
to later join the ranks of labs like mine ai is going to have a profound
mine ai is going to have a profound
mine ai is going to have a profound impact on our society so it’s important
impact on our society so it’s important
impact on our society so it’s important to ask how are we going to use it
immense positives may come along with
immense positives may come along with negatives such as military use or rapid
negatives such as military use or rapid
negatives such as military use or rapid disruptive changes in the job market to
disruptive changes in the job market to
disruptive changes in the job market to make sure the collective choices that
make sure the collective choices that
make sure the collective choices that will be made about AI the next few years
will be made about AI the next few years
will be made about AI the next few years will be for the benefit of all every
will be for the benefit of all every
will be for the benefit of all every citizen should take an active role in
citizen should take an active role in
citizen should take an active role in defining how a I will shape our future
defining how a I will shape our future
defining how a I will shape our future thank
”
Be First to Comment