[Applause]
so after I’ve received this email saying
you would like to give a TED talk here
in Freiburg I was like whoa what an
honor but then I was like thinking about
the topic and you know they asked me to
give a talk about big data today and I
was like yay big data yeah that’s my
favorite topic and you know we are
working for some very large companies
and the first meetings they’re like
always the same they are like hey we are
generating these huge amounts of data
please can you take it and solve all our
problems and you know sometimes it
really feels like you want to stop
poverty and hunger in the world use big
data you want to prevent terrorism
attacks use machine learning you want to
replace a president hmm I have an idea
what about if we built a deep neural
network for that and you know after a
while like everybody is like thinking
the same question however no one dares
to ask it and it’s what exactly is a
neural network and to give you some more
insights on that topic I made a short
video so it’s about me using a neural
network while I like wise generating
data to improve it so watch this it’s a
game it’s called quick draw of course
it’s from Google and my task is to draw
see circle yeah I see blueberry oh I
know it’s baseball
very good so next one is castle so watch
this I see hockey stick or stairs or
skyline oh I know it’s castle yeah very
good so this is the last one fence so
watch this
I see gear
or pants or lighthouse no it’s not a
lighthouse or jail I see stitches sorry
or swingset or camouflage or crayons
I see oven yeah
and obviously it’s also another Nova so
what you can see here is like two
different things so first of all you can
build a machine that is like very
powerful and it’s able to identify my
poor drawings and second of all after I
draw something completely different the
machine is lost
okay so completely lost and before I
tell you like the logic behind that I
brought a little picture with me and I
would like to give you the chance to try
it by your own of course
so it’s this I’m one of the co artists
of this beauty I did this incorporation
with the young lady so she is the
daughter of a good friend of mine and
the question is what can you see here on
that picture very good
so very good so first thing I’m looking
for is like in this area here and
believe it or not but this is an
elephant and this supposed to be a pig
and you know this is a carrot the
question is why do you know that this is
the carrot okay so why do you know that
this is a carrot why is I mean it can be
like a cat or let’s say a bear trying to
attack you well it’s because when you
are like a little child you know like
your mother told you this is a carrot
and it’s orange and it has like this
shape and you can chop it in pieces and
eat it because it’s like healthy food
it’s good for you it makes you like big
and strong and after a while after you
are collecting more data meaning you saw
a couple of more carrots your brain is
finally able to differentiate cats from
carrots and this is exactly how our
neural network works so think about your
brain is like a huge network of
different neurons and they are like all
connecting to each other okay so all the
guys from Google’s if they if they want
to like build a machine to recognize
care
of course any data and this is where the
big data comes into play
so think about millions of pictures and
the only thing you need to tell the
Machine is in some pictures are carrots
and in others there’s like other stuff
like all the elephants and penguins and
tag at TEDx letters and so on now of
course you need to transfer these
pictures into a machine readable
language and then you feed it in and
then there’s something happening and I
call it the magic in-between and this is
like exactly when like you’re your
mother is telling you hey little friend
this is not a carrot this is an Apple
the machine learns that in some pictures
there’s like a specific structure a
specific shape and this is our carrot
and whenever you feed in a carrot
only the lower output neuron is like
firing and whenever you feed in let’s
say an elephant then the upper neuron is
firing and that’s it so basically that’s
all the magic behind neural networks and
yeah question is why is this such a hype
topic why are so many famous people
claiming that this is the next trillion
dollar market so the thing is if you
have a lot of data you can build even
deeper and deeper in rural networks and
thereby you can recognize even more
complex structures okay so think about a
car that is like able to identify and
react on like crossing pedestrians or
let’s say are the bicycles on the road
of other cars so this will make driving
way more safe right this will literally
reduce accidents to a minimum and this
is good for us so that’s very good
however the general functionality is not
changing so it’s still a car it’s still
something that takes you from one
direction to another okay so here’s
another example so think about a doctor
that is like scanning a patient’s body
to look for a tumor let’s say so the
doctor needs to be very highly educated
and he needs to view all these Mr images
to see if there is a tumor or not so now
you can use like millions of Mr images
and basically you can teach a machine to
do exactly the same job and you know
don’t get me wrong that’s really great
that’s it’s a beautiful technology and
having all of that data in the world you
can do that the problem is that the real
problem doesn’t get solved so there are
still people suffering from disease
there are still people out there dying
from cancer you know you are using
artificial intelligence like every day
like every day in your life
so for example like your phone like the
voice the voice of Siri this is like the
product of a neural network or let’s say
if you ever use Google Translate this is
a neural network and they use all the
data you are generating to improve it
however it remains a tool to transfer
language right so it remains something
that you can like feed in a sentence and
it will translate the sentence from like
one language to another
so one last statement for today and
unfortunately even with all ADA in the
world I don’t believe that we can build
powerful enough deep neural network to
replace presidents however there’s hope
I am pretty sure that we can use this
beautiful technology to solve a lot of
problems in the future thank you very
much [Applause]
so after I’ve received this email saying
you would like to give a TED talk here
in Freiburg I was like whoa what an
honor but then I was like thinking about
the topic and you know they asked me to
give a talk about big data today and I
was like yay big data yeah that’s my
favorite topic and you know we are
working for some very large companies
and the first meetings they’re like
always the same they are like hey we are
generating these huge amounts of data
please can you take it and solve all our
problems and you know sometimes it
really feels like you want to stop
poverty and hunger in the world use big
data you want to prevent terrorism
attacks use machine learning you want to
replace a president hmm I have an idea
what about if we built a deep neural
network for that and you know after a
while like everybody is like thinking
the same question however no one dares
to ask it and it’s what exactly is a
neural network and to give you some more
insights on that topic I made a short
video so it’s about me using a neural
network while I like wise generating
data to improve it so watch this it’s a
game it’s called quick draw of course
it’s from Google and my task is to draw
see circle yeah I see blueberry oh I
know it’s baseball
very good so next one is castle so watch
this I see hockey stick or stairs or
skyline oh I know it’s castle yeah very
good so this is the last one fence so
watch this
I see gear
or pants or lighthouse no it’s not a
lighthouse or jail I see stitches sorry
or swingset or camouflage or crayons
I see oven yeah
and obviously it’s also another Nova so
what you can see here is like two
different things so first of all you can
build a machine that is like very
powerful and it’s able to identify my
poor drawings and second of all after I
draw something completely different the
machine is lost
okay so completely lost and before I
tell you like the logic behind that I
brought a little picture with me and I
would like to give you the chance to try
it by your own of course
so it’s this I’m one of the co artists
of this beauty I did this incorporation
with the young lady so she is the
daughter of a good friend of mine and
the question is what can you see here on
that picture very good
so very good so first thing I’m looking
for is like in this area here and
believe it or not but this is an
elephant and this supposed to be a pig
and you know this is a carrot the
question is why do you know that this is
the carrot okay so why do you know that
this is a carrot why is I mean it can be
like a cat or let’s say a bear trying to
attack you well it’s because when you
are like a little child you know like
your mother told you this is a carrot
and it’s orange and it has like this
shape and you can chop it in pieces and
eat it because it’s like healthy food
it’s good for you it makes you like big
and strong and after a while after you
are collecting more data meaning you saw
a couple of more carrots your brain is
finally able to differentiate cats from
carrots and this is exactly how our
neural network works so think about your
brain is like a huge network of
different neurons and they are like all
connecting to each other okay so all the
guys from Google’s if they if they want
to like build a machine to recognize
care
of course any data and this is where the
big data comes into play
so think about millions of pictures and
the only thing you need to tell the
Machine is in some pictures are carrots
and in others there’s like other stuff
like all the elephants and penguins and
tag at TEDx letters and so on now of
course you need to transfer these
pictures into a machine readable
language and then you feed it in and
then there’s something happening and I
call it the magic in-between and this is
like exactly when like you’re your
mother is telling you hey little friend
this is not a carrot this is an Apple
the machine learns that in some pictures
there’s like a specific structure a
specific shape and this is our carrot
and whenever you feed in a carrot
only the lower output neuron is like
firing and whenever you feed in let’s
say an elephant then the upper neuron is
firing and that’s it so basically that’s
all the magic behind neural networks and
yeah question is why is this such a hype
topic why are so many famous people
claiming that this is the next trillion
dollar market so the thing is if you
have a lot of data you can build even
deeper and deeper in rural networks and
thereby you can recognize even more
complex structures okay so think about a
car that is like able to identify and
react on like crossing pedestrians or
let’s say are the bicycles on the road
of other cars so this will make driving
way more safe right this will literally
reduce accidents to a minimum and this
is good for us so that’s very good
however the general functionality is not
changing so it’s still a car it’s still
something that takes you from one
direction to another okay so here’s
another example so think about a doctor
that is like scanning a patient’s body
to look for a tumor let’s say so the
doctor needs to be very highly educated
and he needs to view all these Mr images
to see if there is a tumor or not so now
you can use like millions of Mr images
and basically you can teach a machine to
do exactly the same job and you know
don’t get me wrong that’s really great
that’s it’s a beautiful technology and
having all of that data in the world you
can do that the problem is that the real
problem doesn’t get solved so there are
still people suffering from disease
there are still people out there dying
from cancer you know you are using
artificial intelligence like every day
like every day in your life
so for example like your phone like the
voice the voice of Siri this is like the
product of a neural network or let’s say
if you ever use Google Translate this is
a neural network and they use all the
data you are generating to improve it
however it remains a tool to transfer
language right so it remains something
that you can like feed in a sentence and
it will translate the sentence from like
one language to another
so one last statement for today and
unfortunately even with all ADA in the
world I don’t believe that we can build
powerful enough deep neural network to
replace presidents however there’s hope
I am pretty sure that we can use this
beautiful technology to solve a lot of
problems in the future thank you very
much [Applause]
Please follow and like us: