Hello who are you my name is Jan Verhasselt
i'm the Managing Director of
artificial intelligence company called
Yazzoom and I'm here at the expo in
Amsterdam to demonstrate our product
called Yanomaly that you see behind me
and is used to monitor continuously all
kinds of machines and production lines
and detective something is going wrong
in that much machine and flag it and
alert the users that I should go and
inspect a certain sensor or a certain
part of the Machine and this purely
automatically based on artificial
intelligence what are your company
exactly exactly doing what is what are
your company up to yeah so we are
helping our customers which are mostly
an industry
this can be process industry like in
Finland very popular belt and paper
industry but can be also discrete
manufacturing like making components for
cars or can be batch chemical processes
or can be also machine builders so like
companies building weaving machines and
selling them around the world all these
are customers of ours that and we help
them with artificial intelligence to
either make their production better or
to create new innovative concepts into
their products and create maybe new
business opportunities by making
innovations inside their products what
is happening here what we are seeing
here is a small factory it is driven by
this steam engine over here and that
generates this motion inside the factory
we see a small stamping machine over
there we see a sledgehammer over here
and we have equipped this small factory
with a number of sensors which you see
behind me on the screen so for example
here on top of the steam boiler is a
temperature sensor and you see the
temperature measurement over here and
our technology using unsupervised
machine learning which is a sub area of
artificial intelligence automatically
learns for each of the signals normal
patterns normal values and normal
statistical distributions and if now
something happens which is out of the
ordinary
it will race here a high anomaly score
and also inform the user why it has a
high anomaly score so that they can go
and inspect either that sensor or the
equipment behind the sensor to check
whether something is going wrong and I
can illustrate this if you want well why
not here under this wheel is a magnetic
flux meter and there is a small magnet
attached to the wheel so each time then
the wheel turns you will see a peak in
this magnetic flux meter because the
magnet is getting close to the sensor
now I have here a small additional
magnet and if I bring that in the
neighborhood of the sensor it will
create another magnetic field and the
sensor will pick that up and hopefully
the anomaly will also show an anomaly
score because something is disturbing
the magnetic sensor so let me illustrate
that now you saw me bring the magnet
close to the and here you see how the
magnetic field was influenced and there
you see a peak in the anomaly score
which is also shown shown here in the
very red warning with an anomaly score
of one so this is an example of what is
called univariate anomaly detection
meaning you detect an anomaly on one
single sensor at a time another cool
thing I can demonstrate is multivariate
anomaly detection this means that the
artificial intelligence will learn the
normal relationships between the
different sensors and if something
changes in that relationship it will
raise an alarm alright so and I can
illustrate that please be my guest here
what you see here is a steam valve in
between the steam boiler and the Pistons
that generate a motion so and our
artificial intelligence has been trained
in normal operation in normal operation
this machine this steam valve is always
open now if I close it as you will see
I will slow down the turning I don't
want to let it stop so let me carefully
monitor that it doesn't stop completely
so you see the voltage generated by that
power generator over there drop whereas
the temperature is still pretty high and
it says there is something wrong it's
not normal that the temperature is one
hundred and seven and a half degrees and
that there is only that much that little
power which is being generated by that
dynamo which is proportional to the
speed of the rotation also the spacing
between these Peaks is now wider than it
was before and the algorithm learns that
there is something wrong because
normally if the temperature is this high
it has learned the relationship between
temperature and the spacing in between
the individual Peaks which are
indicative of the turning speed of the
wheel and now it will little by little
get back to normal because because I
have closed the valve temperature is
rising further pressure is building up
and it needs to release the steam anyway
so it will just use higher pressure to
just to force this the steam through the
small opening and it will again
accelerate and that's the reason why you
see the anomaly score dropping again now
if I open it and put it again in the
completely normal position here you will
see that very quickly the anomaly score
will drop to a very low value meaning
everything is back to normal
so problem is solved problem solved a
happy ending that's a happy ending
indeed yeah artificial intelligence in
general is computer algorithms that
exhibit some behavior which we would
call intelligent behavior right which is
a quite broad term and one sub domain
within artificial intelligence is
machine learning which is the ability
for a computer to learn from examples
right as we learn when we are small kids
we learn from a hearing people talk and
that this is how we acquire language so
that's learning form examples now the
science of
figuring out algorithms that can let
computers learn from examples is called
machine learning
now machine learning is only one sub
domain
of artificial intelligence there is
another sub domain that we also use
within our company which is optimization
so optimization means if you have a
certain quantity like a euro amount of
profits for example that you would like
to maximize and it is influenced by many
different factors but in in a very
complicated ways and you want to find
out what is the best way to make the
biggest possible profit and algorithms
that find in this many different options
the best choice this is called
optimization algorithms which is also a
field within artificial intelligence
because this kind of smart
decision-making is also something when
we see a human do it that we would call
out this is an intelligent intelligent
person he make smart decisions all right
so this is another example and a third
category is search algorithms so
algorithms to find useful information
inside in for inside raw data and the
algorithms that everybody knows is
Google right so this is also artificial
intelligence but yet another sub domain
of artificial intelligence I hope that
clarified and gave a no it definitely
does such an intelligent stuff what do
you think what kind of professionals we
must educate in the future if we think
about school education what kind of
professionals the world will need to
tomorrow and in the future you mean in
the field of artificial intelligence or
in general in the field of artificial
intelligence let's put it that way ok so
I think the ideal education for doing
artificial intelligence is a computer
science as a basic education and then
specialize in the domain of artificial
intelligence that is ideal but inside my
company I have also people educated in
exact Sciences that are different from
computer science like myself my first
engineering study was automation my
second engineering study was physics and
only then in my PhD I did artificial
intelligence and I have several people
that had done physics as a study
had done electro-mechanics as a study
and then later throughout a PhD or
through some post master education they
acquired skills in programming
specifically in the sub domain of
artificial intelligence so I think it's
important to be analytical to love let's
say exact sciences and dealing with
numbers and working with computers and
analyzing problems and being prepared
to to keep looking if it's not clear at
first sight and not very obvious but you
need to have this kind of investigation
spirit to find the truth in this massive
amount of data and think about the best
ways of assessing the performance of
what you're doing all right so that's
this the kind of skills you need and I
think there are several pre-studies
which are useful but the area of
artificial intelligence in itself is
worth of at least one year of dedicated
study of the different algorithms and
the different tools that exist in this
field and then afterwards like in any
other ICT so information technology
sector it's evolving very quickly so
that means that you need to love you
know being up-to-date and retrain
yourself continuously on what is
changing in the world in this area and
you have small additional studies and
and and Lao and love this right I mean
it's not a burden it we love to be on
top of what is going on in the world and
be one of the people that that or one of
the companies that can really set the
barriers of what is possible with
artificial intelligence and use that to
help the world get a little bit better
right love is the key to everything
indeed yes summarize your message about
artificial intelligence to the world
what do you want to say to the world
about artificial intelligence I would
say don't be afraid of artificial
intelligence because some people think
it will kill our jobs and it will
whatever I think like any other
technology it can be used in bad ways
but also in good ways
and I think artificial intelligence will
really improve our lives a lot in the
next 10 years to come I think in in 10
years from now it will be hard to
imagine that you and I are standing here and
maybe we have some kind of a disease and
but there is no continuous monitoring of
our health maybe in 10 years we will
look back and say oh how primitive was
the world in 2018 because we were having
these health risks and not being aware
that we were slowly getting some disease
right whereas early warning systems like
what we are showing here for industrial
applications can also be applied of
course to human health and animal health
and so on so this is one area where I
think it will help humanity a lot and so
I would say if you love to be on working
in a high-tech technology and having an
impact on the future lives of our
children then doing the study in
artificial intelligence is definitely a
good one because it will enable you to
do things which were not impossible
before
ok wise words thank you Jan ok you're
welcome
Không có nhận xét nào:
Đăng nhận xét