
in the 1960s many people thought that
artificial intelligence is just around
the corner even many researchers thought
that in just 10 years time it would be
possible to build computers that are as
good as humans in every cognitive task
so now 50 years later it’s easy to say
that oh that didn’t happen
we still don’t have the AI that was
promised to us in science fiction but I
have been in the field for 25 years
studying and developing artificial
intelligence and today I’m here to tell
you that artificial intelligence is just
around the corner so I’m going to talk a
little bit about why we want to build
this artificial intelligence human-like
computers that can do things that humans
now only only humans can do and then I’m
going to talk about the technology what
is it that makes humans able to do this
stuff where are we now how are we going
to build them and then a few words about
how we can share the benefits of this
new technology with everybody in the
1960s people were dreaming of these
human level computers that would power
Android servants that do all the tasks
that human don’t want to do and so that
humans can spend time with their
families and friends and do all the
things that made life worth living
instead we got the internet and
smartphones they are great tools I use
them everyday myself but they are not
quite what we were expecting are they as
an example a few weeks ago I was
traveling in London I was planning to go
from the railway station
of the airports to the city and I had
booked my tickets with my mobile phone I
was walking to the platform and there
was waiting for the Train but then the
train was delayed okay luckily I was I
wasn’t in a hurry but some people were
worried about their schedules and they
thought frantically typing on their
phones and then they found there is a
person there’s a conductor on the
platform and they were queuing to her
they won’t asked her questions she was
able to reply each one of them
immediately clearly and they were happy
with the service that’s the kind of
service we would like to have that’s the
kind of help that our mobile phones now
can’t give us in unexpected situations
humans are the only ones who can help my
mother for instance is suffering from
using her mobile phone because nowadays
many services can only be reached
through mobile phones she would like to
talk to a human so my goal is to build
technology human-like AI that my a
mother would be happy to work with so
let’s have a look that’s what else we
could do with this technology it’s not
just service but there are many things
that we are now doing with dumb
technology for instance we’re using a
lot of energy and raw materials when we
could instead use intelligence we are
making our environment pretty monotonous
so that our dumb technology can cope
with it we are even making our own
working lives more standardized so that
all the processes can be worked all the
processes work together with the
technology that we have all of this
could change if only we could build
these more human-like machines so we are
not trying to replace humans most of the
time actually what we are trying to do
is we are trying to replace the dumb
technology
that nowadays has automated many of the
things that humans used to do so now
let’s look at this technology can we
build it
what do we have now in the 1960s people
were not only expecting to have this
nice robots to serve them but they were
also afraid of robots rising against
their creators like viruses in this
famous movie 2001 space are the same the
artificial intelligence held 9000 rebels
against humans researchers and industry
was trying to educate people no no no
that’s not how it works that computers
do exactly as they are told no more and
no less that was true then it was true
for a long time
internet smartphones are built based on
that kind of technology we call it
programming people then call it
artificial intelligence nowadays what we
call artificial intelligence is
something different because there are
limitations to what you can do with
programming for instance I’m seeing you
now I’m talking to you
you are listening to me you understand
what I say all of this is very hard to
program and that’s why when you try to
build self-driving cars or you make
virtual assistants that understand
speech or you try to translate language
nowadays we use something which is
called machine learning it means that
you take a lot of data you give verses
instances of images from from traffic
and then you give also the desired
output these are cars these are
pedestrians and so on you take a lot of
data and you build this algorithm you
program the algorithm that learns from
the data that’s how AI works nowadays so
we can say that today computers only do
what they learn from the day
that’s the reality now and this is this
is how it is now and some people think
this is how it will be there is
something in humans that we can do much
better than computers now we are
creative we can respond to new
situations I’ll give an example from
Homo sapiens living his daily life
namely me I’m biking to work basically
every day and this is my way back home
every day I take this spiking Lane and I
turn right a few days ago we had
visitors coming in for dinner and I
needed to go buy groceries from a shop
which is straight ahead
so you’ll guess where I was I find
neuroscientists and cognitive scientists
have figured that they are like two
kinds of two modes of thinking in humans
they call it system 1 and system 2 the
system one is the autopilot it’s the
it’s the part of my brain that’s doing
biking for instance it’s the part of my
brain that took me right and almost back
home it’s very fast it’s intuitive
I don’t need to think about it in fact I
cannot think about it I cannot explain
how I do it
I just do it it’s unconscious fully
automatic and crystallized knowledge
it’s something that I learned from
experience from a lot of data
sounds family right very much like
artificial intelligences these days what
we have now in AI is basically the
system one what we are missing is this
creative system – which is able
handle new situations I was supposed to
use this system to when I’m needed to
make an exception I’ll go buy
ingredients for the great dinner it was
successful in the end
so no worries my brain is able to
imagine things I can I have simulations
going on inside my brain and that’s the
foundation of this system – this is what
we need to build there are some examples
of AI systems that use this kind of
system – for instance alphago built by
google deepmind few years ago they won
the world’s best go player go is a game
which is even more difficult than chess
it’s very famous popular in East Asia
and this program alphago and its
successor alpha zero which a half a year
ago became the world’s best chess player
has a kind of system – so what do you do
is you program in a game engine so it’s
fully programmed thing just programs has
have used that for ages and you can use
this for simulating the game so given a
board position you can start simulating
what what could happen and from that
simulation you create data so the system
system – the simulation creates its own
data and then there are system one which
builds intuition is based on this data
if the simulation itself can be used to
decide how good the situation is and
also suggest new moves that may be maybe
that’s a good move and then these
intuitions feed back to the planning
process an expert player doesn’t need to
think very many moves a beginner will
think think about many many different
moves
but an expert player into intuitively
knows that these are good-looking moves
I’ll think about them only so this is
our system to in alpha zero will help
system one develop intuitions which in
turn makes make that makes the planning
process perfect this is something that
we can do in games but when we are for
instance talking about traffic
self-driving cars we can no longer build
all the rules inside the computer there
is no way to build rules about how
pedestrians work for instance so now we
and others have started developing
systems which can learn the rules of the
world they can learn the rules of the
game how the traffic works house how
communication with people work and so on
and this is the foundation of a new
breed of very high which is going to be
here any time now and which all of you
will be able to use it’s going to be
more human-like it’s going to be
creative it’s going to be able to
communicate in communication the
important thing is transmitting these
mental images if you can’t imagine you
cannot properly communicate if you think
Foresters about that example that I gave
in London that queue of people that was
trying to ask questions from the
conductor what I did I was transmitting
my internal simulations to you you
understood what I was trying to say so
these new computers which are able to
imagine will be also able to understand
us so yes we are building this
technology is happening and it’s a good
thing so always when there is
technological change
potential for not everybody being
winners even though the grow even even
though the pie grows bigger it’s not
clear that everybody will get a bigger
share of the pine so the most important
thing we need to do with this technology
is to make sure that we distribute the
benefits of a ice labor to everybody
there are certain things that will work
just as before our society already has
social security and all of these systems
in place and many things like versus the
pension system I think will work just as
before it’s based on investing in
companies and the companies will
continue making profits so we can share
those profits for everybody but what
about for instance labour now we are
taxing labour and using that to fund our
social security some people have
suggested that if robots will take over
some of the work that humans now do we
should tax the robots well there’s a
problem here is an AI from the movie
here just in case you haven’t seen the
movie here’s the area it’s not a
physical thing it’s a it’s a voice in
the air artificial intelligence is
immaterial computation happening in some
cloud it’s not possible to text that so
I think there’s every reason to be
optimist
there’s this new breed of AI coming and
there is a lot of good stuff we can do
with it so I started with asking can we
build is human-like area yes we can and
that’s a good thing yes we want to do
that because there’s so much good that
you can do do with it so all of you
hopefully in five years time foresters
will be able to use that AI and build
good things in the society but yes but
we have to be mindful this technology is
very powerful and that means that it’s
also potentially dangerous we need to
make sure that we will distribute the
benefits for everybody if we do this
right the future will be bright thank
you [Applause]
artificial intelligence is just around
the corner even many researchers thought
that in just 10 years time it would be
possible to build computers that are as
good as humans in every cognitive task
so now 50 years later it’s easy to say
that oh that didn’t happen
we still don’t have the AI that was
promised to us in science fiction but I
have been in the field for 25 years
studying and developing artificial
intelligence and today I’m here to tell
you that artificial intelligence is just
around the corner so I’m going to talk a
little bit about why we want to build
this artificial intelligence human-like
computers that can do things that humans
now only only humans can do and then I’m
going to talk about the technology what
is it that makes humans able to do this
stuff where are we now how are we going
to build them and then a few words about
how we can share the benefits of this
new technology with everybody in the
1960s people were dreaming of these
human level computers that would power
Android servants that do all the tasks
that human don’t want to do and so that
humans can spend time with their
families and friends and do all the
things that made life worth living
instead we got the internet and
smartphones they are great tools I use
them everyday myself but they are not
quite what we were expecting are they as
an example a few weeks ago I was
traveling in London I was planning to go
from the railway station
of the airports to the city and I had
booked my tickets with my mobile phone I
was walking to the platform and there
was waiting for the Train but then the
train was delayed okay luckily I was I
wasn’t in a hurry but some people were
worried about their schedules and they
thought frantically typing on their
phones and then they found there is a
person there’s a conductor on the
platform and they were queuing to her
they won’t asked her questions she was
able to reply each one of them
immediately clearly and they were happy
with the service that’s the kind of
service we would like to have that’s the
kind of help that our mobile phones now
can’t give us in unexpected situations
humans are the only ones who can help my
mother for instance is suffering from
using her mobile phone because nowadays
many services can only be reached
through mobile phones she would like to
talk to a human so my goal is to build
technology human-like AI that my a
mother would be happy to work with so
let’s have a look that’s what else we
could do with this technology it’s not
just service but there are many things
that we are now doing with dumb
technology for instance we’re using a
lot of energy and raw materials when we
could instead use intelligence we are
making our environment pretty monotonous
so that our dumb technology can cope
with it we are even making our own
working lives more standardized so that
all the processes can be worked all the
processes work together with the
technology that we have all of this
could change if only we could build
these more human-like machines so we are
not trying to replace humans most of the
time actually what we are trying to do
is we are trying to replace the dumb
technology
that nowadays has automated many of the
things that humans used to do so now
let’s look at this technology can we
build it
what do we have now in the 1960s people
were not only expecting to have this
nice robots to serve them but they were
also afraid of robots rising against
their creators like viruses in this
famous movie 2001 space are the same the
artificial intelligence held 9000 rebels
against humans researchers and industry
was trying to educate people no no no
that’s not how it works that computers
do exactly as they are told no more and
no less that was true then it was true
for a long time
internet smartphones are built based on
that kind of technology we call it
programming people then call it
artificial intelligence nowadays what we
call artificial intelligence is
something different because there are
limitations to what you can do with
programming for instance I’m seeing you
now I’m talking to you
you are listening to me you understand
what I say all of this is very hard to
program and that’s why when you try to
build self-driving cars or you make
virtual assistants that understand
speech or you try to translate language
nowadays we use something which is
called machine learning it means that
you take a lot of data you give verses
instances of images from from traffic
and then you give also the desired
output these are cars these are
pedestrians and so on you take a lot of
data and you build this algorithm you
program the algorithm that learns from
the data that’s how AI works nowadays so
we can say that today computers only do
what they learn from the day
that’s the reality now and this is this
is how it is now and some people think
this is how it will be there is
something in humans that we can do much
better than computers now we are
creative we can respond to new
situations I’ll give an example from
Homo sapiens living his daily life
namely me I’m biking to work basically
every day and this is my way back home
every day I take this spiking Lane and I
turn right a few days ago we had
visitors coming in for dinner and I
needed to go buy groceries from a shop
which is straight ahead
so you’ll guess where I was I find
neuroscientists and cognitive scientists
have figured that they are like two
kinds of two modes of thinking in humans
they call it system 1 and system 2 the
system one is the autopilot it’s the
it’s the part of my brain that’s doing
biking for instance it’s the part of my
brain that took me right and almost back
home it’s very fast it’s intuitive
I don’t need to think about it in fact I
cannot think about it I cannot explain
how I do it
I just do it it’s unconscious fully
automatic and crystallized knowledge
it’s something that I learned from
experience from a lot of data
sounds family right very much like
artificial intelligences these days what
we have now in AI is basically the
system one what we are missing is this
creative system – which is able
handle new situations I was supposed to
use this system to when I’m needed to
make an exception I’ll go buy
ingredients for the great dinner it was
successful in the end
so no worries my brain is able to
imagine things I can I have simulations
going on inside my brain and that’s the
foundation of this system – this is what
we need to build there are some examples
of AI systems that use this kind of
system – for instance alphago built by
google deepmind few years ago they won
the world’s best go player go is a game
which is even more difficult than chess
it’s very famous popular in East Asia
and this program alphago and its
successor alpha zero which a half a year
ago became the world’s best chess player
has a kind of system – so what do you do
is you program in a game engine so it’s
fully programmed thing just programs has
have used that for ages and you can use
this for simulating the game so given a
board position you can start simulating
what what could happen and from that
simulation you create data so the system
system – the simulation creates its own
data and then there are system one which
builds intuition is based on this data
if the simulation itself can be used to
decide how good the situation is and
also suggest new moves that may be maybe
that’s a good move and then these
intuitions feed back to the planning
process an expert player doesn’t need to
think very many moves a beginner will
think think about many many different
moves
but an expert player into intuitively
knows that these are good-looking moves
I’ll think about them only so this is
our system to in alpha zero will help
system one develop intuitions which in
turn makes make that makes the planning
process perfect this is something that
we can do in games but when we are for
instance talking about traffic
self-driving cars we can no longer build
all the rules inside the computer there
is no way to build rules about how
pedestrians work for instance so now we
and others have started developing
systems which can learn the rules of the
world they can learn the rules of the
game how the traffic works house how
communication with people work and so on
and this is the foundation of a new
breed of very high which is going to be
here any time now and which all of you
will be able to use it’s going to be
more human-like it’s going to be
creative it’s going to be able to
communicate in communication the
important thing is transmitting these
mental images if you can’t imagine you
cannot properly communicate if you think
Foresters about that example that I gave
in London that queue of people that was
trying to ask questions from the
conductor what I did I was transmitting
my internal simulations to you you
understood what I was trying to say so
these new computers which are able to
imagine will be also able to understand
us so yes we are building this
technology is happening and it’s a good
thing so always when there is
technological change
potential for not everybody being
winners even though the grow even even
though the pie grows bigger it’s not
clear that everybody will get a bigger
share of the pine so the most important
thing we need to do with this technology
is to make sure that we distribute the
benefits of a ice labor to everybody
there are certain things that will work
just as before our society already has
social security and all of these systems
in place and many things like versus the
pension system I think will work just as
before it’s based on investing in
companies and the companies will
continue making profits so we can share
those profits for everybody but what
about for instance labour now we are
taxing labour and using that to fund our
social security some people have
suggested that if robots will take over
some of the work that humans now do we
should tax the robots well there’s a
problem here is an AI from the movie
here just in case you haven’t seen the
movie here’s the area it’s not a
physical thing it’s a it’s a voice in
the air artificial intelligence is
immaterial computation happening in some
cloud it’s not possible to text that so
I think there’s every reason to be
optimist
there’s this new breed of AI coming and
there is a lot of good stuff we can do
with it so I started with asking can we
build is human-like area yes we can and
that’s a good thing yes we want to do
that because there’s so much good that
you can do do with it so all of you
hopefully in five years time foresters
will be able to use that AI and build
good things in the society but yes but
we have to be mindful this technology is
very powerful and that means that it’s
also potentially dangerous we need to
make sure that we will distribute the
benefits for everybody if we do this
right the future will be bright thank
you [Applause]
Please follow and like us: