LaBeouf’s throwing a riot a whole bunch
of robots going crazy well it’s one of
my favorite movies uh not just because
Will Smith my favorite actor because it
was the first movie I ever watched that
showed me a real like somewhat real
depiction of an artificial intelligence
like this idea that humans could create
something born of Technology that seems
human but is very distinct and it’s not
just me
Media cinema they’re all enamoured by
the idea of an artificial intelligence
the idea that like an AI could somehow
rise and surpass humans up guys know
this guy
Ultron from from the Avengers I think I
think this guy epitomizes the way that
we look at artificial intelligence the
idea that like we’re going to create
some kind of artificial being with
personality very similar to a human and
yet and yet he’s going to be so
different uh but because of depictions
like Ultron people also think that AI is
something that doesn’t exist yet and
that’s the question I going to ask you
know the duck does AI exist yet well
though you’re not going to onion on your
drive home you’re not going to drive by
Ultron I can bet that each and every
single one of you today has interactive
than artificial intelligence and you
didn’t even know about it
see artificial intelligence doesn’t just
exist today but it’s modern incarnation
has existed for over a decade and it’s
in everything and it exists every single
day in your life and if you don’t
believe me I’m about to prove it to you
today did you google anything
maybe when you’re on your way to the P
AC or you know after you go you’re going
to find a place to go to dinner well if
you have then there’s a good chance you
interacted with an AI and AI called
rankbrain
our rankbrain is google’s artificial
intelligence that basically goes to a
whole bunch of user profiles right
it’ll earns what makes people similar
and what makes them different so that
way when you put in an order for
something it’ll it’ll tell or when you
put in a search it will tell you what
you want not just on what you typed in
but also because it knows who you are
and its intent is gained from there did
you read the news by chance well if so
there’s a good chance that you’re the
reticle
that you read was not written by a
person rather is written by a robot the
associated press releases about 3,000
quarterly reports and not one of them is
read or is written by a human they’re
all written by an AI called wordsmith
that takes raw data goes through a
formula and then puts out a whole bunch
of paragraphs about it same thing with
Fox Sports for baseball what they do is
they take raw raw data gives short
postgame updates in the form of in the
form of tiny little paragraphs in
sentences and if you ever do fantasy
football you’ll know that Yahoo Sports
you probably under you who at Yahoo was
time to write me this well it’s not
you’ve been lied to is the robot who
wrote it all this time so then you’re
probably wondering okay so if it’s not
this you know Ultron asked artificial
intelligence that the what is artificial
intelligence so in order to get to what
artificial intelligence is what AI is
you have to know intelligences itself so
psychologists Computer Sciences they
tend to think of it as pattern
recognition the smarter you are the
better you’ll be able to recognize
patterns for example something that we
would consider dumb like an ant will
keep bumping into a mirror right it’ll
it doesn’t know where it’s going and
will keep bumping and it’ll keep bombing
until it dies it’s not capable of
realizing that if I go into this but if
I go into this mirror like I’m not going
to be able to move forward ah they don’t
take a slightly smarter animal like a
dolphin a dolphin if you put a mirror in
like the ocean might swim towards the
thing it’s not a dolphin it’ll bump
against it and realize oh I recognize
that this object is not another dolphin
and I’ll go swim away but human were the
most intelligent of all creatures we
don’t even need to bump into a mirror to
know that okay it’s mirror we’re not
going to bump into it
so then artificial intelligence
basically is a computer program that’s
capable of recognizing patterns what it
does is it runs the code it gets a
result and then uses that result to
affect its own code it rewrites its own
algorithms it reaches itself so with
this we have a whole capabilities a
whole bunch of possibilities about AI
not just limited to fantasy sports a
high-spade treating high-spade a speed
treating form firms like Citadel and
aadya use artificial intelligence its
core to their trading platform what they
do is they have an AI look at how the
market has been doing over the past few
months
and then we’ll move find a prediction
for the next for the next few months and
we’ll act
exclusively on that there’s very little
human input in Citadel they would rather
hire a PhD in computer algorithm puter
algorithms than they would someone with
an MBA Tesla your new favorite car
company also is a heavy user of machine
learning a derivative of artificial
intelligence
each Tesla car is lined up with
chock-full of sensors that help
determine not just the location of your
car the location of other cars and where
they were before then the machine
learning algorithm then determines not
just where the car is but where it will
be in the next few seconds which allows
it to determine the location of cars in
the future and then it pretty much
predict the crash before it even happens
so basically artificial intelligence
it’s utility to us is as humans is it
autumn eyes as things why would I Drive
a car when I’m going to have a car drive
itself for me why would I hire brokers
when I can have a computer make
thousands of trades in one second so
ultimately automation has major benefits
like it’s easier it’s cheaper it’s
faster both organization you lose
something very crucial you lose control
and as we give artificial intelligence
bigger and bigger tasks we sacrifice
things that are getting more and more
important in fact one day an artificial
intelligence will have to choose between
life and death because as we keep on
giving it more and more important tasks
we then put our own safety our lives at
risk let me give you an example so one
day I believe we’re going to have a
fully automated transportation system
all cars are going to be part of some
kind of car internet and they’re going
to be able to interact with each other
same way user profile’ and you won’t
have to do any driving so let’s take you
to that year let’s in 2035 so I’m going
to juice you two car one car one has one
mother she’s a soccer mom she’s taking
care of her youngest child who’s a
toddler they’re both in the backseat
they’re just on their way to they’re on
their way to school and in the other car
we have one c-suite executive at a very
major pharmaceutical company and what’s
happening is these two cars are go in
opposite directions and all of a sudden
due to some unknown error there is some
kind of crazy malfunction and now these
cars are headed on on its are going in
on a head-on collision
so now you’re probably asking yourself
all right what can the computer system
do to fix it so the computer system
realize this is too late and then runs a
million calculations and realizes that
there’s only there’s there’s only one
way you can only save two of the three
lives so that means either the toddler
the executive or the mother dies so in
this case we have a real dilemma we have
no idea well the computer has no idea
how to calculate or how to justify the
taking of one’s life over the over the
lives of others well that’s not entirely
true
computer scientists are trying to find a
way or trying to find multiple ways in
which that we can derive a solution this
problem uh so that’s that’s basically it
how can you create a formula can you
program morality into a computer so
there are two there are two measures I’m
going to talk about the first one is
called well I’d of it the fiscal measure
the idea that we’re going to weigh each
person’s life based on the net impact
they have on society so for example
that’s the suite executive if we say
that he is to die then what is the net
impact on the rest of the world on the
rest of the economy
well when any normal company X
experiences some kind of shift and
management especially executive level
their stocks are gonna plummet uh so
imagine if their c-suite executive their
CEO dies companies would investors would
back out companies would lose millions
of dollars in infrastructure and as a
result the company may have to lay off
hundreds if not thousands of people
compare that to say the death of a
toddler toddler doesn’t employ anyone
doesn’t contribute to the job force
their death as unfortunate as it would
be would be much less impactful than the
death of the executive so if we were to
implement this method what would happen
is thousands of lives or not thousands
laps thousands of dollars millions of
dollars possibly would be saved and we
wouldn’t even know about it uh the model
does ignore certain things though for
example it ignores potential the idea
that a child may be worth later later on
in the future however we can’t recognize
that immediately but there’s also
another fatal flaw with this the idea
that well if you’re influential it’s
more likely it’s because
you’re wealthy and what happens is when
you input it when you put this method in
wealthy people are going to be kind of
getting their privilege extended before
they just got to drive nicer cars
probably more safer cars but now with an
AI guiding the system using this
protocol they’re pretty much immune to
car accidents and what’s even worse is
poor people are now facing an even
greater amount of car accidents so then
there’s the other matter the social
matter the societal measuring are many
people in philosophy dictate that we our
morals are determined by society so if
we’re going to create a moral machine we
should have society determinate so the
people should decide who lives and who
dies so you might be wondering how can
you do this you can’t crowdsource that
much information oh it’s just small and
such an infant has infinitesimal time so
what we do is it’s an artificial
intelligence we teach it uh what we do
is we collect a whole bunch of
information be it a census we kind of
get some kind of site together and we
learn from humans morals we learn from
how they behave and we learn at the way
that they treat each other and then from
that we take the mean of all their
answers we collapse that and we figure
out okay this is the this is the outcome
that would leave the least amount of
people unhappy and in that case most
people would say according according to
this one according to this one program
by MIT called the moral machine what
they did is they ran hundreds of
simulations I like this so you can
actually participate this in right now
if you want you have to choose between
cool lives and who dies and they found
out that children are the least likely
to die in executives and criminals are
more likely to die so what happens is we
have a quote unquote socially optimal
scenario but is it really socially
optimal the issue with this is inherent
within society itself that we as a
society are flawed the fact that we have
our own biases and not only that how is
this going to scale the views of society
within one state be it Texas is very
different than the views of one state be
it New York and then when you scale it
to a national level or an international
level you’re going to have major
disputes and who is going to determine
what kind of AI it is you know is it
going to be from New York is going to be
from Texas whose morals are
instilling within the robot itself so
ultimately what I want to get across
here is the fact that no matter which
method we use no matter which computer
algorithm we’re smart enough to generate
it will fail we will always be
disappointed with our results not based
on not because we’re not good enough
technologists but because the idea of
what we’re doing itself is flawed in
essence we cannot outsource humanity we
can’t we can’t let go of the part which
makes us human when we when we when we
oughta Mize characteristics we’re doing
things because they’re too easy for us
but for things that are too difficult
like making decisions like these we can
never allow ourselves the opportunity to
let this go now I’m not trying to say
that we should halt production of any
artificial intelligence engines AI to me
still is most fascinating field there is
in computer science and I want us to
continue but as we continue in this
field I want us to remember that we’re
the creators we’re the builders were the
makers and we’re the ones we’re going to
be maintaining them fueling them
controlling them
we can’t allow artificial intelligence
to control us or determine our live
paths thank you [Applause]