
let’s talk about artificial intelligence
and before most of us were born there
was already a good part of literature
about science fiction and artificial
intelligence for example Isaac Asimov
has written a book about a robot who can
think like a human and probably you’ve
seen the film with this i robot with
will smith and at the same time Philip
dick wrote a novel story about a robot
who was thinking he was a human and this
story was the background for blade
runner you have seen this film perhaps
this movie so the idea of artificial
intelligence that we can create
computers that can really think like
humans has been around for a very long
time and i would like to combine it
today with two different passions of
mind behind science fiction literature
so my second passion i would like to
share with you it’s machine learning
machine learning is this kind of pattern
recognition that makes everything smart
from your smartphone to ultra nomas cars
and i’m teaching machine learning at the
university and doing a lot of machine
learning and from time to time i even
dream machine learning and this sounds
weird but it even gets worse because i’m
a big fan of the philosophy of Hegel you
know geography pleased with him he the
old German philosopher was famous for
his dialectics and what I would like to
show to you is if you put these three
things together artificial intelligence
machine learning and the philosophy of
legal you got a combination that is like
gasoline and matches I think that we
will see in the next five years
something like artificial spirits with
this I mean computer programs that are
really able to
you make operation similar to human
thinking but without the complexity of
the human mind but if we can reach this
step then soon after perhaps even 2025
we might reach a point where we have an
artificial mind the thing all this
science fiction literature is dreaming
about a computer that is able to think
and has a consciousness and then if this
would happen shortly after we probably
will have super intelligence the
computer that is superior in all kind of
cognitive tasks compared with any human
well I know that at least ninety percent
of AI researchers will strongly disagree
with me about this time scale but let’s
have a look what a is and then I will
give it a dialectic turn because i would
like to convince you first that real AI
a computer that things like a human is
not possible and then i will convince
you that this was wrong so this is my
plan i I can be defined as intelligence
exhibited by machines and this was the
big program early AI researchers were
trying to find from the beginning
touring was dreaming about AI everyone
was very excited in the 50s but what we
have today is something else if you look
at them state of the art in AI research
we have mostly machine learning and
Andrew entering one of the most
important researchers in the field of AI
he has said it like this almost all of a
ice recent progress is through one type
and in which some input data a is used
to quickly generate some response be and
this is for example when you think about
your smartphone and theory or the google
assistance you can ask any kind of
question and this
language is the data a and then there is
a mathematical function you don’t have
to care about they transfer this input
data a into an output a good answer to
your question that is B but this output
is already known before you are asking
your questions so what the computer only
can do is to fit with mathematical terms
the input a to an already known be and
because of this John Searle was arguing
that artificial intelligence is not
possible and he used the thought
experiment to do this it’s the famous
Chinese room officer the experiment
thought experiment goes like this Searle
is sitting in a room and someone is
handing in papers with Chinese symbols
on the papers and Searle has a big book
with a lot of rules a kind of algorithm
and uses this rules to transfer the
symbols to other Chinese symbols and
hence them out and from the outside it
might look as if so was able to speak
perfectly Chinese because it does
exactly what zelia might do on the
smartphone all answers are right
everything makes sense but so in this
room will never ever understand a word
of Chinese he has no consciousness about
what he’s doing he has no clue what he’s
doing and this is a metaphor for what is
going on in computers and as long as we
try to teach computers this way to go
from A to B with a mathematical function
there is really no way that they can
understand what they are doing and they
won’t get intelligent and theory won’t
become your best friend said news but
what is wrong about this whole picture
is not the definition of what a is doing
in this room but what thinking is if you
think about we’re thinking is thinking
is definitely something else than just
having a known a and a known be and
tried
get from A to B but if you start to
think you don’t know where you end up
you don’t know the be so inserts Chinese
room there is really no room for a
thought because everything is given up
front and here comes higley so imagine
the door of the Chinese room opens higa
comes in and says lay aside your book so
I gonna teach you how to think because
this was what he was really deeply
involved he said there are rules how we
think and these rules are called logic
well the logical figure is very
complicated but I think you can simplify
it to three simple rules and I hope
there are no philosophers here because
otherwise I get shot for this one first
rule is no second rule is bad and the
third rule is reflection or I give it to
you in the words of Hegel it would be
omnis determine a cos Nicasio which is
latin and means everything that is
dieter mind is the negation the two
moments of the difference our identity
and difference and finally then essence
is the move is the movement from nothing
to nothing and thereby back to itself
well I know this doesn’t make really
sense for you right now but I will
explain it on example so if we take the
sentence from the beginning AI is
intelligence exhibited by machines then
we can start to negate the sentence to
make it more interesting AI is not
intelligence ok now you couldn’t get it
again a is intelligence no it’s not yes
no yes no it doesn’t lead you anywhere
this is what he called a bet infinity so
you have to find another way of negating
it it’s the negation of the negation and
we could do this by finding a different
word that means the engagement of
intelligence something like stupidity so
we would end up with the first sentence
like a
I is intelligence and stupidity and now
we have a contradiction and
contradictions are always very
fascinating if you want to think about
them more now comes the second rule we
got some more words left in the sentence
we got artificial and we got exhibited
by machines and these words construct
the difference but there is an identity
within this difference because if you
think about words like artificial
exhibited machines you come up with an
idea of something unnatural something
simulated perhaps you come up with the
idea that is about computers so we might
end up with a sentence like AI is
intelligence and stupidity because it is
not real intelligence but fake
intelligence simulated in a computer
model and this is more or less the whole
idea of Searle what he was arguing in
the Chinese room but now comes the big
surprise i created a computer model that
uses exactly these three rules of hegel
to think by itself and if you take the
first sentence as an input like we did
in the example the sentence artificial
intelligence is intelligence exhibited
by machines the computer will transfer
this sentence to AI is intelligence and
stupidity because it is a counter height
of intelligence which is recreated
intelligence in a created model and this
is more or less the same we were
thinking about in the example before and
the way the computer gets there is by
using machine learning to analyze the
syntax of the sentence so how the words
are connected and something which is
called the Semantic Web so there is a
known connection of words something
might be a sooner noon for another word
or something might be a type of another
word and you can use these information
to feed them into a computer model and
because it’s a computer model you can
now feed in any other kind of sentence
and many sentence make a lot of sense
and the computers constructing
interesting thoughts well of course many
sentence don’t make sense right now but
it’s a prototype so what do you want but
we still have the last rule the
reflection essence is the movement from
nothing to nothing and thereby back to
itself and this is so important here
because the whole point is my computer
program has no idea what artificial
intelligences I’m not claiming this this
would be silly and it has no idea what a
counter fight of recreated intelligence
is so it starts from nothing and moves
back to nothing but the important thing
is that the way how you get from this
first nothing to the second nothing
there is a lot of sense in this way and
there is a deeper connection it’s like
you are expanding your concept and this
is what thinking is about thinking is
about expanding your concept so of
course the computer still doesn’t know
what a is but he knows how to connect
the first part AI with a broader concept
and because the result is speech and the
input was speech as well the computer
can now start to think about its own
outputs this might be very very
important because then different outputs
will it lead to totally different ways
and maybe computers will start
developing a consciousness on their own
based on different starting points well
let me add one final word because there
is a lot of fear about artificial
intelligence going around Ellen mask is
arguing about super intelligence and
even Stephen Hawking has said super
intelligence is very very dangerous but
if he could really
make computers think like humans do and
not in a very very strange way which I
described in this from A to B style but
really thinking then I think there is no
reason to believe why some entity that
is smarter than we are but things on the
same level should act like an asshole
and war and terror is never very clever
so I think if you could reach real
artificial intelligence we’re going to
see great times thank [Applause]