[Applause]
if you have in the cinema see a movie
of action you will certainly see some
extraordinary scenes and yet that
you will appear completely credible
[Music]
it’s not always well done but
look carefully at this
extract where the face of the princess
leia and entirely created in images of
synthesis
today the special effects are so
well done that you no longer do the
distinction between the origins
original image and part
fakes makes you bluff of course you
know that all this is not real
since you’re in the cinema if
now I told you that this type of
trickery is becoming so
easy to achieve that will invade our
daily and alter our perception of
the reality is what are you ready to
believe me then I ‘m a creative and a
techie for over 15 years I
works in the field of 3d and
I love to explore the news
technologies to understand their
potential their impact and for the
use my turn is precisely a
some of the most complicated things to do
in computer graphics ie to
make credible
these are the human characters so
when the hollywood cinema we in
watch you must take well
awareness of the means that are deployed
rites behind to generate these
images it takes time and it takes
money
a lot of upstream preparation of
technical means as combinations
of motion capture many of
studios are doing green and also mostly
a lot of people
talented but a new way
to create realistic images is in
emerging
thanks to a branch of intelligence
artificial and I think that’s okay
radically change the way of
manipulate reality then this branch
artificial intelligence is called
dip learning or deep learning
in French so it applies to
IT is basically it’s a
set of two methods that allow
automatic learning
these methods are extremely
powerful to analyze for
recognize to synthesize data
and that’s what I think is really cool
it is that the machine can learn
by herself and be creative at her
turn then lenval started around
in 2010 and these are now
billion who are invested in the
domain the field is progressing very fast
and today we manage to modify
to synthesize
people in video here are some
examples here the face of nicholas cage
has been inserted
instead of the lead actress
is it automatically thanks to a
artificial intelligence the of
researchers show us how he is
possible to transfer our movements from
head and our expressions on another
no one so we finally become
puppeteer and this type of technique
also allows to finally edit the image
from someone in a very intuitive way
so if we could go is what we
can go beyond the face is this
that we can go to the bodies and
goods let’s go
we take a professional dancer
video we ask people to move
no matter how in front of the camera
[Music]
and artificial intelligence their
apply the movements so today
everyone can dance like pros
the righteous ones in video so it’s these
videos we see people who are
modified by artificial intelligence
is called ddip fake it’s contraction
of the word diploma ning so the method
deep learning and fake
the English word which means must
ten fake so what is
artificial intelligence well changed
well what is revolutionary
finally it is that soon you will
all can do this type of trick there
at home contrary to what I
mentioned for the film industry
with these tools you no longer need
of talent you do not need to
technical knowledge in video you
no longer need so much time and
some of its technologies like
the one that allows to swap the faces
are already available they are
free they are open source
you just have to finally have a
computer at home and here’s the proof
so that’s me here’s an example of a
reality alter automatically
thanks to an artificial intelligence
for now neither nothing but that’s my
franck friend whose face comes
replace mine in the video here is a
example of a reality altering so
automatic thanks to an intelligence
artificial
so how did I do that for
make this video
it was pretty easy actually I
downloaded the software and then I asked
in franck to film small videos
five minutes in all sent them I
did the same thing I filmed myself and
that at first it allowed me
to collect a large number of photos
who represents us and who presented
our faces and in a second time
I gave all that to the program that
looked at each photo and who learned to
recognize our faces and synthesize
synthesize also our faces so this
phase there after about
two days is called the phase
of learning and finally I gave in
make a video
entering the program with me
on it and the program has automatically
generated the fake video with franck in
less than two minutes
I could have given any
video of me as input you see it’s
easy that accessible it’s not expensive
so of course there are some flaws
always we can see in these videos
problems with resolution blur
flicker but the field is changing
so fast that soon we will not
will see that fire
so that’s for the video and for the
his is well the advances are major
also today we come to create
synthetic voices that seems so
natural that we can almost no longer
distinguish from those of men and
several companies are working on
technologies so always based on this
deep learning that allow
to imitate to reproduce the voice of
anyone just from a
voice recording
so there we can say but in
combining these two techniques
I can do and make say
anything to almost anyone
so why is it almost because he
need data upstream on the
person for the learning phase
the order of magnitude is thousands
of images and a few minutes
registration before then
thousands of the image so do you
you say it’s good it’s far
to exist
millimages the representative may have
think again because 1000 images
it’s only 40 seconds of video
in other words, it’s easy enough to
get
we will resume in detail what we
can do with these techniques so
acquired a this technique there I can
put someone’s face on the
someone else’s body I can
how to say I can avoid the
movements I can do do some
moves to someone who has never
does I can place a person in a
context has never been and we can even
actually invented what she will say then
I think it’s very
impressive
it may sound a little bit too
scary there is real potential
creative for this technology that’s for sure
for all those who want to seize it
and we can just think of
the movie industry that could
produce more easily and with less
cost
and besides, it could open up
new perspectives
imagine for example with your friends
in a custom movie but what
is a lot more questioning
finally it’s the power of
possible drifts
some see the deep fake as
being the first real threat of
artificial intelligence
the personalities are the main ones
targets for now because it’s
very easy to collect media from
data on it and so the first
really affected people were
actresses and singers whose
face has been inserted into movies
pornographic
and then there were also parodies
staging actors but also
politicians like Vladimir
poutine angela merkel or donald trump
but in politics is it okay
stop at the humor probably that
no you know the information is a
weapon and so the use of the said
done as a propaganda tool can
wreak havoc
here is a video made by the director
john jordan stack just to wake up
consciences are less
impressive then
imagine how we can all be
manipulated if a man’s speech
policy can be created from all
rooms
we can also think that if this type of
videos is spreading in the media but
finally a politician can
deny his words by saying waited for that
it was a fake video so basically
a scandal that really would have
place could be declared as no
having never happened and beyond
the individual there can be also
states that can be targeted
by other states or by
groups that would like to sow the
confusion
one last example take the CEO of a
known company made him do it any
what you stream the video you can
discredit his business or manipulate
the course of action then today
all these scenarios there are possible
are plausible
because the information is true or
false spreads in a massive way
almost instant on the networks
social
since it is sensational and
then admit what about you is this
that you share photos or
personal videos on the networks
social
if so it could actually be the
raw material that could use to
ill- intentioned ends like
blackmail the identity theft the
size of your reputation
by the way on the point on the point
legal or are we ben we are
in the middle of floating is what to do
check dips is illegal is that it’s
moral because we use tools
finally everything depends on what we
done with a hammer you can
to tinker you can use it
as a weapon and if you are
victim that says the law but no
clear answer so far a point
really crucial legal him too
it’s the notion of evidence
today the audio and video files
can be received as coins to
conviction but you saw that can
be falsified or completely created
all rooms so how do we
can still receive them if eventually
we have no guarantee on their
reliability
then how
how to know if a video is
authentic
how to avoid being fooled
how to prevent these videos from getting
shed no one answer
but rather a panel of actions that
finally engages everyone
and here are some axes the first
question we could ask is
can we stop the technology
is what you can ban his
use well prevent its spread
good it’s difficult because some
software is already there already available
and then research advances them
researchers show us new
amazing results he publishes
some take it for good or
bad intentions
in this case what we can not use
technology to help us
finally investigate to detect these types
fake
then the subject is taken very seriously
especially when the faking is actually
perfect for the human eye
researchers working on the
tools that allow to do these say
fake are also those who work on
the means and detect because they are
aware of potential drifts
develops algorithms and among these
algorithms we have rather interesting
some algorithms look at whether the
no one in the video blinks
in a natural way
other algorithms checks who has a
consistency of blood flow between the
part of the cost and part of the face
other algorithms still search it
is an earlier video but that
would be very similar to the one we
studies and finally a kind of method
general is ultimately to lead
simply an intelligence
artificial to detect these ten fake
so where can we deploy these
detection technologies
first of all at home when you are
on the internet a little bit like
anti-virus
we see extensions appear that
do checks on all
media that are present in your page
and who can alert you to
potential falsification but the
the place it seems to me more relevant
to deploy his techniques there it is
where the information is spreading from
massive way ie the
youtube media platforms or the
social networks
so if detect these said does that
should he do what should
delete a priori not because everything says
do it’s not dangerous
but would not it be their
responsibilities ultimately to detect
to detect them
as soon as a video is added to their
servers and just from us
alert by telling us attention this
video is falsified or potentially
falsified states also have a role
important to play because overall
artificial intelligence is going
transform our society a bit in the
extent where internet already transformed
before and the big companies the
know that’s why she’s investing
massively and therefore accordingly
states must also be interested
invest anticipated this technology the
understand to protect us because
that there are not only stakes
economic to intelligence
artificial is specifically the
National security agencies must
to appropriate this subject there this subject of
deep tech
but the central point is finally
we are our critical sense
it’s important to know that these
there exist technologies can
easily be fooled and
today do this tedx allow me to
you talk about it
so in front of information
sensational it is necessary to show
discernment
do not hesitate to doubt sometimes
investigate tried to find the source
some information
good this kind of investigation takes some
time sometimes demands skills
so it would be good for us to go on
sites where information is reliable
reliable it means where the facts were
checked and where the sources are traced
wikipedia is a good example
because the encyclopedia this
encyclopedia lists all
sources and alert us on a page
when to read and missing none
information and unreliable and all this
thanks to hundreds of thousands of
contributors and in 2017 its founder has
announced the creation of forum wiki that
is supposed to transpose this model there to
the news check this check
times by people over
professional cip copywriter or
journalists
we run we all consider
instinctively a video as being
a capture of what really happened
past
and yet we have seen that the video is not
necessarily a reliable witness then
are we just planning a moment in a
time when there will be a very very big
proportion of these fake videos of his
10 sec
well we would not know what is
wrong and so we would no longer be
plus what is true so are we going in
get there
in the history of technologies we see
that all these new technologies
that seem so threatening to us when
it is beyond us are finally good
less alarming camp we tame them
have more advanced technology the more the
tic more moral sense plus vigilance
must prevail
let’s not forget it’s us
decide to believe what we
gives us to see
we will enter an era where
assisted by intelligence
artificial we could all be from
true forgers of reality
so for tomorrow what reality
do we want to manufacture
[Applause]
if you have in the cinema see a movie
of action you will certainly see some
extraordinary scenes and yet that
you will appear completely credible
[Music]
it’s not always well done but
look carefully at this
extract where the face of the princess
leia and entirely created in images of
synthesis
today the special effects are so
well done that you no longer do the
distinction between the origins
original image and part
fakes makes you bluff of course you
know that all this is not real
since you’re in the cinema if
now I told you that this type of
trickery is becoming so
easy to achieve that will invade our
daily and alter our perception of
the reality is what are you ready to
believe me then I ‘m a creative and a
techie for over 15 years I
works in the field of 3d and
I love to explore the news
technologies to understand their
potential their impact and for the
use my turn is precisely a
some of the most complicated things to do
in computer graphics ie to
make credible
these are the human characters so
when the hollywood cinema we in
watch you must take well
awareness of the means that are deployed
rites behind to generate these
images it takes time and it takes
money
a lot of upstream preparation of
technical means as combinations
of motion capture many of
studios are doing green and also mostly
a lot of people
talented but a new way
to create realistic images is in
emerging
thanks to a branch of intelligence
artificial and I think that’s okay
radically change the way of
manipulate reality then this branch
artificial intelligence is called
dip learning or deep learning
in French so it applies to
IT is basically it’s a
set of two methods that allow
automatic learning
these methods are extremely
powerful to analyze for
recognize to synthesize data
and that’s what I think is really cool
it is that the machine can learn
by herself and be creative at her
turn then lenval started around
in 2010 and these are now
billion who are invested in the
domain the field is progressing very fast
and today we manage to modify
to synthesize
people in video here are some
examples here the face of nicholas cage
has been inserted
instead of the lead actress
is it automatically thanks to a
artificial intelligence the of
researchers show us how he is
possible to transfer our movements from
head and our expressions on another
no one so we finally become
puppeteer and this type of technique
also allows to finally edit the image
from someone in a very intuitive way
so if we could go is what we
can go beyond the face is this
that we can go to the bodies and
goods let’s go
we take a professional dancer
video we ask people to move
no matter how in front of the camera
[Music]
and artificial intelligence their
apply the movements so today
everyone can dance like pros
the righteous ones in video so it’s these
videos we see people who are
modified by artificial intelligence
is called ddip fake it’s contraction
of the word diploma ning so the method
deep learning and fake
the English word which means must
ten fake so what is
artificial intelligence well changed
well what is revolutionary
finally it is that soon you will
all can do this type of trick there
at home contrary to what I
mentioned for the film industry
with these tools you no longer need
of talent you do not need to
technical knowledge in video you
no longer need so much time and
some of its technologies like
the one that allows to swap the faces
are already available they are
free they are open source
you just have to finally have a
computer at home and here’s the proof
so that’s me here’s an example of a
reality alter automatically
thanks to an artificial intelligence
for now neither nothing but that’s my
franck friend whose face comes
replace mine in the video here is a
example of a reality altering so
automatic thanks to an intelligence
artificial
so how did I do that for
make this video
it was pretty easy actually I
downloaded the software and then I asked
in franck to film small videos
five minutes in all sent them I
did the same thing I filmed myself and
that at first it allowed me
to collect a large number of photos
who represents us and who presented
our faces and in a second time
I gave all that to the program that
looked at each photo and who learned to
recognize our faces and synthesize
synthesize also our faces so this
phase there after about
two days is called the phase
of learning and finally I gave in
make a video
entering the program with me
on it and the program has automatically
generated the fake video with franck in
less than two minutes
I could have given any
video of me as input you see it’s
easy that accessible it’s not expensive
so of course there are some flaws
always we can see in these videos
problems with resolution blur
flicker but the field is changing
so fast that soon we will not
will see that fire
so that’s for the video and for the
his is well the advances are major
also today we come to create
synthetic voices that seems so
natural that we can almost no longer
distinguish from those of men and
several companies are working on
technologies so always based on this
deep learning that allow
to imitate to reproduce the voice of
anyone just from a
voice recording
so there we can say but in
combining these two techniques
I can do and make say
anything to almost anyone
so why is it almost because he
need data upstream on the
person for the learning phase
the order of magnitude is thousands
of images and a few minutes
registration before then
thousands of the image so do you
you say it’s good it’s far
to exist
millimages the representative may have
think again because 1000 images
it’s only 40 seconds of video
in other words, it’s easy enough to
get
we will resume in detail what we
can do with these techniques so
acquired a this technique there I can
put someone’s face on the
someone else’s body I can
how to say I can avoid the
movements I can do do some
moves to someone who has never
does I can place a person in a
context has never been and we can even
actually invented what she will say then
I think it’s very
impressive
it may sound a little bit too
scary there is real potential
creative for this technology that’s for sure
for all those who want to seize it
and we can just think of
the movie industry that could
produce more easily and with less
cost
and besides, it could open up
new perspectives
imagine for example with your friends
in a custom movie but what
is a lot more questioning
finally it’s the power of
possible drifts
some see the deep fake as
being the first real threat of
artificial intelligence
the personalities are the main ones
targets for now because it’s
very easy to collect media from
data on it and so the first
really affected people were
actresses and singers whose
face has been inserted into movies
pornographic
and then there were also parodies
staging actors but also
politicians like Vladimir
poutine angela merkel or donald trump
but in politics is it okay
stop at the humor probably that
no you know the information is a
weapon and so the use of the said
done as a propaganda tool can
wreak havoc
here is a video made by the director
john jordan stack just to wake up
consciences are less
impressive then
imagine how we can all be
manipulated if a man’s speech
policy can be created from all
rooms
we can also think that if this type of
videos is spreading in the media but
finally a politician can
deny his words by saying waited for that
it was a fake video so basically
a scandal that really would have
place could be declared as no
having never happened and beyond
the individual there can be also
states that can be targeted
by other states or by
groups that would like to sow the
confusion
one last example take the CEO of a
known company made him do it any
what you stream the video you can
discredit his business or manipulate
the course of action then today
all these scenarios there are possible
are plausible
because the information is true or
false spreads in a massive way
almost instant on the networks
social
since it is sensational and
then admit what about you is this
that you share photos or
personal videos on the networks
social
if so it could actually be the
raw material that could use to
ill- intentioned ends like
blackmail the identity theft the
size of your reputation
by the way on the point on the point
legal or are we ben we are
in the middle of floating is what to do
check dips is illegal is that it’s
moral because we use tools
finally everything depends on what we
done with a hammer you can
to tinker you can use it
as a weapon and if you are
victim that says the law but no
clear answer so far a point
really crucial legal him too
it’s the notion of evidence
today the audio and video files
can be received as coins to
conviction but you saw that can
be falsified or completely created
all rooms so how do we
can still receive them if eventually
we have no guarantee on their
reliability
then how
how to know if a video is
authentic
how to avoid being fooled
how to prevent these videos from getting
shed no one answer
but rather a panel of actions that
finally engages everyone
and here are some axes the first
question we could ask is
can we stop the technology
is what you can ban his
use well prevent its spread
good it’s difficult because some
software is already there already available
and then research advances them
researchers show us new
amazing results he publishes
some take it for good or
bad intentions
in this case what we can not use
technology to help us
finally investigate to detect these types
fake
then the subject is taken very seriously
especially when the faking is actually
perfect for the human eye
researchers working on the
tools that allow to do these say
fake are also those who work on
the means and detect because they are
aware of potential drifts
develops algorithms and among these
algorithms we have rather interesting
some algorithms look at whether the
no one in the video blinks
in a natural way
other algorithms checks who has a
consistency of blood flow between the
part of the cost and part of the face
other algorithms still search it
is an earlier video but that
would be very similar to the one we
studies and finally a kind of method
general is ultimately to lead
simply an intelligence
artificial to detect these ten fake
so where can we deploy these
detection technologies
first of all at home when you are
on the internet a little bit like
anti-virus
we see extensions appear that
do checks on all
media that are present in your page
and who can alert you to
potential falsification but the
the place it seems to me more relevant
to deploy his techniques there it is
where the information is spreading from
massive way ie the
youtube media platforms or the
social networks
so if detect these said does that
should he do what should
delete a priori not because everything says
do it’s not dangerous
but would not it be their
responsibilities ultimately to detect
to detect them
as soon as a video is added to their
servers and just from us
alert by telling us attention this
video is falsified or potentially
falsified states also have a role
important to play because overall
artificial intelligence is going
transform our society a bit in the
extent where internet already transformed
before and the big companies the
know that’s why she’s investing
massively and therefore accordingly
states must also be interested
invest anticipated this technology the
understand to protect us because
that there are not only stakes
economic to intelligence
artificial is specifically the
National security agencies must
to appropriate this subject there this subject of
deep tech
but the central point is finally
we are our critical sense
it’s important to know that these
there exist technologies can
easily be fooled and
today do this tedx allow me to
you talk about it
so in front of information
sensational it is necessary to show
discernment
do not hesitate to doubt sometimes
investigate tried to find the source
some information
good this kind of investigation takes some
time sometimes demands skills
so it would be good for us to go on
sites where information is reliable
reliable it means where the facts were
checked and where the sources are traced
wikipedia is a good example
because the encyclopedia this
encyclopedia lists all
sources and alert us on a page
when to read and missing none
information and unreliable and all this
thanks to hundreds of thousands of
contributors and in 2017 its founder has
announced the creation of forum wiki that
is supposed to transpose this model there to
the news check this check
times by people over
professional cip copywriter or
journalists
we run we all consider
instinctively a video as being
a capture of what really happened
past
and yet we have seen that the video is not
necessarily a reliable witness then
are we just planning a moment in a
time when there will be a very very big
proportion of these fake videos of his
10 sec
well we would not know what is
wrong and so we would no longer be
plus what is true so are we going in
get there
in the history of technologies we see
that all these new technologies
that seem so threatening to us when
it is beyond us are finally good
less alarming camp we tame them
have more advanced technology the more the
tic more moral sense plus vigilance
must prevail
let’s not forget it’s us
decide to believe what we
gives us to see
we will enter an era where
assisted by intelligence
artificial we could all be from
true forgers of reality
so for tomorrow what reality
do we want to manufacture
[Applause]
Please follow and like us: