I only have one goal tonight and that is
to convince you that data and data
science is very important to innovation
it’s crucial that data and data science
are within a company to leverage all the
important aspects of data and also the
three symbols on my slide will make
sense to you at the end of this talk so
being a theoretical computer scientists
I like to define things first before I
actually use that so let’s first start
with innovation what is innovation
innovation means that we want to take
things that already exist change them
using well-known methods and make them
better we want to take them to the next
level
that’s what innovation is about what is
data
well data is something that’s everywhere
you have tiny bits of information that
are across the world across the planet
and we want to let them again to use for
innovation so let’s start with this
quote it’s a quote from David McCandless
as that data is a new soil what does it
mean it means that when you have data
you take it as a your garden you want to
do something with it you want to make it
profitable so let’s see how we can do
that who of you knows the HBO show
Silicon Valley in the audience
seafood app so what the seafood app is
about it’s on the show it’s an app that
can recognize your food so you point the
camera to your plate and says well this
is the thing you’re eating only the gouf
of the app is that it can only recognize
that it’s in hot up or not and hotdogs
so there’s only two options but the
amazing thing here is that this app is
actually in the App Store you can just
download it and it works so the
producers of the show they actually
hired someone to take all the hotdog
images from the planet and train a deep
Learning Network to recognize hot dogs
that’s great I guess
but anyway what’s the main point here
is that one week later on Pinterest the
following shows up you can now upload
your pictures of your own dishes and
then you can just see what’s on there
you can see that there’s a hot dog on
your plate or anything else but you can
also search for recipes so we have
created an actual use case by using all
this data the whole point here is that
we want to take a dream and from the
dream we want to build something that
can improve the world and also the thing
itself we can improve and we can create
a new dream and we want to create a spy
file or we can keep improving ourselves
using data and I am convinced that this
can be done by using three things that
is data that is tools we need tools to
leverage the data and we need someone to
operate the tools we need someone like
this guy here
to do something with the tools and have
the right knowledge to take them to the
next level that is why we have the three
images we have the brains we have the
tools and we have the data we just need
to use them in the right way so that’s
why I’d like to do an experiment with
you guys I hope you’re up for it I don’t
like to do these things when I’m in the
audience but now I have choice so you
have a very simple job the only thing
that you need to do is remain seated at
all costs so you need to look at a
person to the right of you you need to
ask what is the battery level of your
phone and then you need to report it to
my data scientist team please stand up
data scientists team I myself am part of
the data scientist teams so you can also
use me you can spread across the room a
bit so there a notebook and they will
note down all the percentages of all
your phone so we can map it and we can
analyze all the data your job is to
think of a new way of getting the data
there or an existing way you can do
whatever you want
you just remain seated again remain
seated and you have one minute to
so data Sciences team start analyzing
police you can go back to your seats and
if you’re done you can report the
results to me and then we’ll go figure
something out so I’m glad that these
things didn’t happen not none of the
people got mad I think so my colleague
Ronnie here was a bit less lucky but
what it happened now I think we can all
agree it was kind of chaos so but
sometimes chaos is good so anyway so the
first thing we noticed that there was a
lot of information coming through the
data scientists is it processed yes yet
by the way how you got a you got an
airplane great so there was too much
information to deal with at the moment
itself that’s the first thing we can see
and then when the time got to 60 minutes
I predicted this we can see that there
was even more formation coming in so it
was at the beginning they were just
standing there and to the end they were
just grabbing information all over the
place so this was not quite ideal and
then finally there were a lot of methods
we saw an airplane a paper airplane
airplane that was true we’ve seen people
shouting we’ve seen people writing
something down so there were a lot of
different ways that the information got
to the data scientists so they still
have to figure out your handwriting even
so because it’s all different so these
are the main problems that are core to
big data that is we have volume velocity
and variety too much data too fast data
and to different data and every year or
so we make up another V so we get a new
characteristic if the data becomes even
more complex to deal with so this is why
it’s quite important to actually know
the tools you’re using how can we
leverage this how can we fix this for
the poor data scientists that are still
working on it and the answer is at some
point like a decade ago we got to Hadoop
who has heard of a dupe Hadoop is
actually a technology that allowed to
deal with this V problem in deal with
large amounts of data it can offer
processing on top of this data so it can
help the data scientist to
something out of there or to look into
the data and since then since the
origins of Hadoop we have a total data
landscape and every thing here is a new
technology that aims to solve one of
these new problems so the data scientist
has to know the tools and that’s what I
said before there are tools but you need
to know them so it’s very important so
why this as this experiment fail because
we did not use the right to I think
that’s quite obvious
so let’s retry with a new tool yet
another experiments so now take your
phones and you get the following job you
can go to Monte comm and you can fill in
the code you see on top that’s five
seven six nine four three and you can
just answer the basic question of your
phones very percentage and the brand and
if all goes well I start analyzing this
data immediately Oh first ones are
coming in I see Android is ahead Oh
things are changing the advantage of
using this – I think it’s quite obvious
we can immediately start analyzing and
we can see that there are a lot of more
Android users than there are Apple users
and we can also see that people’s
batteries are fine so I think the need
for the extra battery packs that we
bought as gifts just put them away not
necessary we can see all the complaints
about the iPhone battery being not so
good is also ok so well if I was a
battery manufacturer I’ll be quite happy
with this set and also TEDx I think you
can do something with this data so to
maybe target some people for that use
androids or something like that but
anyway so the main point of my
presentation is that now we use the
right tool for the job we used something
and I was not responsible for collecting
all the data and figuring stuff out the
tool did that for me it presented it in
a uniform way to me and I could
immediately start drawing conclusions
and gaining insight into what was
happening so this time everything was ok
we only needed one data scientist
instead of they said the team of three
so that’s also an advantage I guess but
what does this actually mean for you or
for an organization well let’s see what
an organization typically means or what
it does with data the first thing is we
gather data that you can see on the left
side we got our data a lot of different
resources we got our data from the web
we got our data from sensors from paper
even or from people directly and we need
to capture all of these things so we
could output the data to the users or to
other technologies that hook up to the
data and what we have in between is very
important there are two ways you
typically do data processing the first
way is that you every week you want the
report or every month or every day so
that is batch processing at one point in
time you go through all your data and
you just do something with it and then
they’re streaming processing and that
means you get live insights to your data
for example if we look at what Twitter
is doing you can see a trending feed so
you can see live what people in Belgium
are looking for currently this is an
example of streaming or real-time
processing data and then we have of
course the layer on top we have the data
exploration layer and also the data
science layer that’s a layer where the
date of scientists lives he lives
between the tools and he needs to be
able to capture all of this data and to
hook himself to all of these these
pipelines and this is a very important
thing so if we simplify the image you
get again the same three images you get
data coming in you get data coming out
then between it’s all made up of tools
so you have a pipeline in a data pipe
that consists of well-known things and a
data scientist lives there he can use
all the data that’s available but
there’s one important thing he has to do
and that’s know the data landscape of
course you don’t need to know all the
technologies but you need to know the
basics when do I use which to what type
of tools the tools are out there so
that’s the thing the dead scientist
needs to master how do you build your
data how do you build your
infrastructure and this is a very
well-known image that represents this
first
of course you need to be able to collect
data but data is everywhere so
collecting it as easy the difficult
thing is storing it you need a lot of
interest infrastructure in place to
capture it and to keep it there and the
next thing you need to do once you have
the data you want to use it you want to
adjust that you want to do operations on
it so you can actually leverage it and
then once you have that you can go to
the next level start exploring use cases
and with these use cases you can for
example go to the top of the pipeline or
of the pyramid and use AI everybody
wants to do AI of course because it’s a
buzzword but we want to solve real
problems and you can only discover the
problems when your base is solid you
want to have your infrastructure in
place so you get reliable data you get
clean data out of it and you can use it
for anything you want coming back to my
original image where we dream built and
improve our idea we want to do this in
an organization but what typically
happens is that we don’t just do this
but we just have a problem we solve the
problem and that’s it we have created
some value and we stop there but we want
to keep iterating over this so I propose
a different way of doing this so instead
of going directly from the problem to
the solution what we actually want to do
is hook up a data science pipeline we
want to capture the data research it and
understand the problem thoroughly so
that the next step is that we can create
new ideas that we can gain new insights
which can be fueled again to new
problems and this is the dream site or
we can create using the existing
pipeline that’s already in place in the
company and yet again we can find the
three things we need the data itself we
also need the tools to do research and
to make it easier for a data scientist
and we need two brains the ones that
know the tools that know the data that
know the domain so they can actually
leverage all the information that’s
there so coming back to the original
quote the data is a new soil I think we
can adjust this one
seeing as data is so important for
innovation for making decisions for
creating value I think you can adjust it
to the following data is not a new soil
it is the best soil we have for
innovation thank you [Applause]