
fashioned a I people so when I started
to do research in artificial
intelligence we were really about doing
this knowledge representation so I
actually come from the knowledge
representation kind of laboratory at the
in Chipping University so I actually
would argue that we have a lot to to
learn from this traditional techniques
to artificial intelligence as as well
but of course today we have a lot of
machine learning to assist and to
improve on those technologies so what
I’m interested in in sharing here the
idea that I think is worth cheering here
today is that how do we go from all this
exciting and very fascinating results
that we see in artificial intelligence
which is really about understanding how
can mechanize intelligence or thought
processes cognitions and so on and how
do we leverage that when it comes to our
own probably problem-solving
capabilities and abilities and I would
argue that we can actually both learn a
lot from artificial intelligence but
also take advantage of the fact that we
have extremely powerful computers when
we solve problems so what we really need
to do is to develop and cultivate our
computational thinking and leveraging
and this computational power that exists
while solving problems so I mean what is
artificial intelligence so I asked the
president of Swedish area society I do
get asked this question quite a lot but
I think from the purpose of this talk at
least I would just say that artificial
intelligence is really kind of the study
of the algorithmic mechanization or
cognition where cognition is how do we
think how do we solve problems how do we
plan how do we reason and so on so how
do we get understand it well enough
and our being able to write algorithms
that can actually carry
the same kind of problem solving
processes for example so we have seen
that in in the previous examples that we
have made quite a good progress and we
have come quite far so I mean one
question we can ask is why now why are
we here today showing these kind of
advances well first of all software is
eating the world so basically software
is taking over where we previously had
hardware for example one just bring in
one example is the Ericsson telecom
networks where previously you had
specialized hardware and specialized the
base stations and if you want to change
the feature of your telecom network
someone has to physically go there
replace the hardware today they just
have a standard computer and they just
change the software so if you want a new
feature in your telecom network you just
change the software and everything works
the same and the same is for cars where
you can have the same engine in several
models of the car the only difference is
that if you pay a hundred thousand more
you get the sports Edition which means
that it has a different software than in
the standard edition so software is
continually eating the world and it’s
also the fact it’s not only the digital
world but those are the analog word are
actually merging so what we see is that
things that we do in the digital world
has consequences in the analog world of
the physical world and of course the
other way around that the things that we
do in the physical world have
consequences in the digital world that’s
kind of easier to see but this actually
been the case that there was a reporter
who had epilepsy and there was someone
who was angry at him they didn’t like
what he’s writing and actually sent him
a file that they opened his file and he
started to blink and it’s like a
stroboscope effect and they actually
caused a seizure for him so it’s
actually a way of attacking a person
through a digital media so the digital
world and the analog world are truly
becoming one so that’s another
interesting fact of course then we also
have this development computational
powers by geographical processing units
so
thanks to all the kids playing a lot of
video games at home we today have the
massive computational powers that we
have today that we need for all this
machine learning and so on we also have
a lot of labeled data we actually have
examples so again it’s basically us here
using our mobile devices or other things
to collect data continuously that allows
these machine learning algorithms to to
learn from the data that we have
available and finally of course it’s
also improvement in algorithms due to
research but it’s interesting that what
we see is this kind of convergence of a
number of different factors that makes
all of this possible so why is this
important them why should we care well
first of all what we’ve seen is this
machine learning can actually find these
complex patterns and extract them so
previously we had to spend a lot of
effort in in defining these ourselves
now they can be extracted much more
efficiently and at scale and it also
allows for adaptation so it’s not a
fixed system but rather it can change
and evolve and adapt at runtime so it
becomes much more flexible and of course
then much more useful but we also see is
that the interface to computers is
becoming much more based on natural
language so I mean in some sense you can
see it I mean the Google search box is
kind of interest it’s just it’s just a
box enter anything you like and it will
answer any kind of questions you have
and what we see today is that of course
that we are evolving this so previously
it was this kind of one question system
or question answering system of course
Google is one example this IBM Watson is
a different example you you ask you the
question it comes back with an answer so
what we’re starting to see now is
dialogues so you actually have a
conversation with your computer or your
your system and it understands you and
keeps a track of history and can respond
in a meaningful manner so the the kind
of low-hanging fruit there is all
customer support functions that
companies
this is very much in the process of
being taken over by these kind of
avatars but then if we look at more
physical entities like robots so what I
would say is the really interesting
thing with a robot it’s that it allows
physical presence anywhere at any time
that I mean previously you had to be
somewhere to physically interact with
that particular location now by having a
robot you can actually be at any place
any place at least that where there is a
robot and you can be at many places
physical places at the same time so I
think that or those things are what’s
really interesting with robotics and I
mean taking all this together is that
what we have seen previously is like
relatively what is a simple or
repetitive repetitive work being taking
over by automation but what we see today
especially if you look at systems like
IBM Watson is that knowledge work it’s
very much right for automation and it’s
actually the kind of professional work
which requires a lot of knowledge a
person to gather a lot a lot of
knowledge and then being able to ask
questions do diagnosis or suggest
treatments and so on those are kind of
jobs which are currently being targeted
by these kind of systems so think for
example lawyers they’re currently
systems that are giving legal advice
much more efficient much faster much
cheaper and also with higher accuracy
than than humans and it’s also the fact
that you have again if you take the Iban
what’s example what they’re trying to do
now is of course go into the medical
business as supporting doctors in making
medical diagnosis so it’s actually very
interesting to see what are the
consequences and then it is important
for all of us so on this note I think
it’s very interesting to look at the
game of chess
so the game of chess has been a game
where we humans don’t have a chance
so since deep blue beat Garry Kasparov
in 1997 I believe it was actually 20
years ago people haven’t had a chance in
a game of chess but what’s interesting
is that even if humans cannot be
computers it’s actually the fact that
humans are becoming increasingly much
better at chess and the reason they are
becoming better is because they are
using the chess programs to train so
they get more effective training through
these computerized chess players and the
second is that the interest per purchase
is still very high so I mean now that –
Carlson is playing and so on that
generates tremendous interest even
though that humans are not the best but
to kind of make the third point which is
the most interesting one is that even
though computers are better than humans
if you combine humans and computers
together in teams which to do in
something called freestyle chess where
you have one or more humans playing
together with one or more chess
computers and they are better than both
the humans and the computers so of
course there is very interesting that we
complement each other that humans have
some strengths and the computers have
some strengths let’s leverage that and
become even better together so in my in
my mind it’s not either computers or
humans it’s more both it’s computers and
humans together collaborating and since
since I think this is very important I
have also been heavily involved in
education so how do we take this
understanding and move it into education
so the concept that I have been
promoting is the concept of
computational thinking which is
basically a problem-solving process to
describe analyze and solve problems in
such a way that computers can assist and
often we use techniques from computer
science because in my mind
computer science is really about how do
we mechanize problem-solving I mean
already from start when Alan Turing
basically invented computer science
during during the Second World War
his goal was we need to be better at
solving this problem namely cracking the
determined code and we need help with
that and therefore he invented this and
we have been studying the computational
properties or problem solving or
computational computation in general so
I will really argue that this is the
kind of skills that people need to
develop so what is it I mean what are
the actual kind of concrete components
of course it’s giving step-by-step
instruction actually instructing a
computer or a person to do things it’s
decomposing problems into smaller parts
breaking problems down finding patterns
again this is what machine learning is
very good at and how do we then taking
these patterns and creating abstractions
or more general models from the
specifics and finally designing
algorithms to actually mechanize parts
or all of the problem-solving process so
these are usually the basic components
of computational thinking so to take a
concrete example so since I’ve worked
with the National Agency for Education
we actually have the government just
decided a or had made a new curriculum
for the K to 9 education here in Sweden
which now involves programming and as
I’ve been part of that process in
developing this new curriculum I’ve been
using this example to to show what we
mean so assume you want to teach the
kids to do regular polygons so in this
case do a square and one way to free
kids to show that to understand what a
square is make a program that draws a
square and in one example would be to
use scratch which is one particular
programming language for doing that and
to do that in scratch you paste
would write a program like this you you
you have to understand that the square
consists of four sides and to draw a
side you move 100 step and then you turn
90 degrees and you’d go 100 F you turn
90 degrees it go 1 and step turn 90
degrees go one on the steps and then you
have done your your square but then you
will what is the commonality here you
want to find the patterns so of course
the pattern is you repeat something four
times so you do draw the line four times
and you turn and you repeat that four
times then you say all great now you’re
down to square how can you do a Pentagon
of course what do you need to change
well you have to change the four into a
five everyone can understand that but
what happens if you change the 100 into
200 well you get the square with twice
as long sides and four times as much
area so you can talk about the relation
between circumference and an area so
what we need to change this course how
much we actually turn and then idea can
be like me you can compute it or it can
be lazy like me and just say oh well you
have to do 360 degrees well you did four
times ninth in the first case now you
have to do four times five times what is
the seventh 2.5 or something
in order to get the full square and now
you can generalize this even further and
get construct an algorithm for drawing
an arbitrary polygon basically just
asking for how many sides and repeating
this so in this particular simple
example we’re going through all these
different steps we have decomposed we
have found a patterns we have
generalized and we have designed an
algorithm for solving this but it’s not
only about skills it’s also about
attitudes like confidence in dealing
with complexity problems with many
different components working with
difficult problems and persistence I
mean this is one of major challenges
persistence how to get the students to
not give up but actually continue in
solving these problems tolerance for
ambiguity the ability to deal with
open-ended problems how do we take an
open and a problem and narrow 8,000
something that we can actually solve and
maybe we can even just by narrowing it
down and juicing this factor we have the
computational tools to our advantage
and finally also the ability to
communicate and work with others to
achieve a common goal so it’s really
about communicator and collaborating you
cannot do things by your own you have to
work together whether you work together
with people or with computers that does
not matter that much I mean you have to
collaborate because the interesting
problems that we need to solve these
days are not promised at a single
individual can solve by the or by
themselves so therefore I think that
computational thing it really captures
what we need to be good at in order to
leverage all these artificial
intelligence and other computational
tools that we have so what’s interesting
here is that as we learn more about
artificial intelligence and as we
develop these tools based on for example
machine learning we need to be able to
leverage them through our computational
thinking by being able to understand how
the computational processes work and how
we can benefit from those when we saw
problems so what we see today is
actually a number of interesting cases
where people use machine learning or
other tools as part of their own
problem-solving process and to just give
the single concrete example was this
writing these reports on what jobs are
likely to be automated by actually
learning and model of what are the
factors that influence the risk of being
taken over by automation and of course
as we are getting better at this we can
of course develop better artificial
intelligence tools so by becoming better
at using the tools we also become better
at developing them further so I think
this is really interesting
interconnection between these two and
also the fact that we as people can
benefit from having this tool at our
disposal and becoming better at solving
problems so kind of to summarize I think
it’s it’s not either/or it’s together
together with the machines and together
with these very sophisticated tools we
can solve the problems that we face as a
society and maybe you can see this as
some next step in evolution
taking this synergy
between man and artificial intelligence thank you very much