study of artificial intelligence or AI
for short so you can imagine that this
question is AI evil is very much on my
mind to get started let me show you a
few items and get your gut reaction evil
or not land mines land mines are clearly
evil they kill and maim innocence they
stick around for decades we’ve signed an
international treaty to ban them how
about technologies such as washing
machines antibiotics and textiles they
save our time eradicate diseases provide
affordable clothing good stuff right
well historically many new technologies
were greeted with some resistance and
suspicion the 19th century Luddites
fought the introduction of looms the
beginning of the textile industry they
threw shoes into the looms to disrupt
their operation ultimately a misguided
gesture so how about AI is AI good or
evil should we attempt to ban AI as we
did with land mines how many of you have
some very real concerns about AI please
raise your hands high no surprise a eyes
gotten a bad rap in Hollywood films such
as The Terminator in the press from
prominent scientists such as the
physicist Stephen Hawking Hawking says
that AI could be the worst thing that’s
ever happened to the human race
worse than climate change worse than
nuclear weapons the fear is real I get
it
some of the most virulent statements
against the AI have come from the
entrepreneur Elon Musk he says that with
a I were summoning the demon that AI
could be a threat to the very existence
of the human race frankly this demon
talk hurts but let’s put emotions aside
and and spend our time trying trying to
discuss AI rationally is AI an
existential threat to the human race
Newsweek magazine sees AI as a
clear and present danger with headlines
like AI is coming in it’s poised to wipe
us out and Newsweek isn’t an outlier we
see these headlines over and over again
frankly they sell they sell newspapers
but still the question is is AI poised
to wipe us out the renowned roboticist
rod Brooks says if you’re worried about
ai technology still has a long long way
to go but my colleague and Goering likes
to say if you’re worried about AI
turning evil that’s like worrying about
overpopulation on Mars it’s just way
beyond the pale but nevertheless what
happens in the future what happens when
the robot finally figures out how to
open that door well to answer that
question we need to take a step back and
ask what is AI exactly in Hollywood
films it’s typically depicted as a
creepy being I can’t do that Dave but we
need to separate Hollywood from the real
world and science fiction from real
science to get deeper into this let’s
look at one of the most impressive AI
systems ever built in March 2016 the
computer program alphago defeated the
human world champion in the ancient
board game of go like more than a
hundred million people I was following
the match riveted but unlike some I
wasn’t surprised by the outcome when it
comes to puzzles two board games AI is
unstoppable
it’s able to look at billions of
permutations at superhuman speeds but
the real world is more complex is more
nuanced and the fact of the matter is
that alphago is a very limited system to
illustrate this I wrote down some
questions that I would ask alphago if we
were introduced in some AI cocktail
party I would say alphago can you play
poker the answer is No alphago can only
play go it doesn’t even extend to other
games
alphago can you cross the street
again the answer is No alphago cannot do
what a human child can do it doesn’t
even know what a Street is alphago can
you tell me about the game turns out
again the answer is No alphago doesn’t
understand the game
it can’t explain itself it doesn’t even
know it won there’s a key distinction
here that we need to make to understand
alphago and AI systems in more in more
general terms and that’s the distinction
between intelligence on the one hand and
autonomy on the other hand this is the
key so please stick with me here I’m
going to illustrate this using a
two-by-two table in the top left we have
high autonomy with minimal intelligence
that’s a bunch of teenagers drinking on
a Saturday night and on the bottom right
we have high intelligence with minimal
autonomy and that’s AI systems like
alphago nothing more than a fancy
calculator with humans we have free will
we make choices we’re intelligent
intelligence and autonomy go hand in
hand but machines are different you can
be super intelligent in some narrow
sense like playing go but have no
autonomy whatsoever as another example
look at computer viruses they’re
autonomous they’re dangerous but they’re
not intelligent so now we can go back to
answer this question what is AI the real
kind not the Hollywood version AI is a
tool for people to use a pencil a fancy
pencil sure but still something that we
as people draw with my six-year-old
mikey is more autonomous than any AI
system he can make his choices he can
cross the street he can explain himself
he can understand English when he wants
to
that’s that’s the difference let’s look
at weapon systems look military weapon
systems are scary enough
imagine a weapon system that can operate
with no human-in-the-loop that can fly
around halfway around the world to hunt
somebody down and kill them that’s the
stuff of nightmares right but notice the
nightmare is the autonomy autonomous
weapons are a nightmare and AI
researchers are unequivocally opposed to
these systems we’ve written a letter to
President Obama saying so no autonomous
weapons okay how about jobs that’s a
concern AI and jobs how many jobs have
been taken away dude automation how many
more will be taken away due to the
advance of AI Hal Varian Google’s chief
economist says that the old jobs will go
away and new jobs will come but how
quickly will that happen and what will
we do in the meantime this is a valid
concern I stay up at night worrying
about this my point is though that the
doomsday headlines the Terminator
scenario those are distractions from the
real concerns like AI and jobs the
concerns that we ought to be thinking
about you might say okay then why play
with fire as it were why not just
declare a moratorium on AI and be done
with it and first of all much as we
would like to stop technology sometimes
technological change is hard to stop to
even slow down if we slow down our
progress other nations will overtake us
if you think about the past what if we
slowed down our progress with
technologies like electricity like
antibiotics and think we want to live in
a world where infant mortality is way up
and life expectancy is way down we don’t
want to go back to the Stone Age but the
most important reason to investigate AI
is that it has enormous potential
benefits potential benefits that can
save lives Paul Allen founded the
nonprofit Allen Institute for artificial
intelligence in Seattle with a clear
mission AI for the
and good my colleagues AI my colleagues
and I at the Institute are not mad
scientist and in love with our creations
like dr. Frankenstein
we’re sincere well-meaning folks trying
to unlock the potential benefits of AI
and let me give you two quick examples
here first from the realm of science
scientists researchers doctors were all
inundated with documents and data and
starving for insight and wisdom we need
help to cut through the clutter to home
in on the information that we need last
year we launched a scientific search
engine called semantic scholar it’s free
it’s available over the internet and
it’s been used by millions of people to
do better science over time we’re going
to see more and more sophisticated
science search engines and science
scientific helpers imagine a situation
with a clue the cure for an intractable
cancer is hidden somewhere amongst
thousands and thousands of tedious
clinical studies and experiments an AI
search engine like Symantec scholar
could help scientists connect the dots
and find a cure that would otherwise be
overlooked hey I can help scientists
find the answers to Sciences thorniest
problems my Microsoft colleague Eric
Horowitz likes to say that it’s the
absence of AI technologies that is
already killing people due to medical
errors preventable medical errors
preventable medical errors are the third
leading cause of death in American
hospitals we can change that AI systems
can warn exhausted and overworked
doctors about mistakes that can point
out to them new treatments they can make
a difference and save lives when I look
around this room I see that AI will save
your life and that’s true for a hundred
maybe 200 of you more over time and even
more when we think about our families my
final example is from the realm of
driving human drivers worried me
particularly now that myself
ten-year-old is driving regularly he’s
texting and driving I know this because
I get the text at least he’s not
drinking and driving although I worried
that some of his classmates are we have
over 30,000 deaths in our highways each
year in AI bay safety systems at car in
cars could reduce those substantially a
few weeks ago a friend of mine was
jogging in Seattle and he was hit by her
car the driver was 96 years old we want
our grandparents our parents and
ultimately us to be as independent as
possible to keep driving
but at the same time we need to keep us
and other pedestrians safe well AI base
safety systems and cars can do exactly
that and remember just as with automatic
transmission or anti-lock brakes
mechanical driving decisions will be
automated but the human is still in
charge the human still determines the
destination where the car will go it’s
not like a thousand cars are going to
band together and try to take over the
White House so is my point that AI is
the best thing since sliced bread
no the potential benefits are enormous
but I do acknowledge that there are some
real risks my point is this AI is
neither good nor evil it’s a tool it’s a
technology for us to use and to
paraphrase Spider Man with great
technology comes great responsibility so
I invite each and every one of you to
join me in advocating for the
responsible use of AI in advocating for
AI for the common good in using AI to
help us tackle problems that we as humans can’t tackle alone thank you