resources where every single human being
on earth has access to basic needs like
enough food clean water and shelter and
that there’s no need to make trade-offs
in allocating resources no need to worry
about running out of resources in
welfare or migration policies and no
inequality based on wealth
how would you redistribute prosperity
and reorganize social hierarchies with
this cornucopia imagine a world without
jobs in which citizens are paid to
pursue leisurely activities like reading
writing self-discovery building
relationships and other hobbies how
would you redefine self-worth without
work imagine a world without death in
which cancer has been cured terminal
diseases have been reversed and genetic
disorders are treated at the prenatal
level how would you decide who gets to
live forever and how would you balance
immortality with population control all
of these scenarios have two things in
common the first is that deciding how
society operates under each of them is
ultimately a political question resource
scarcity division of labor and fear of
death have always defined our discourse
in political philosophy because they
determine how individuals interact and
how they conceptualize justice and
freedom they also determine how
nation-states interact in a global
economy in times of peace and in
preserving national security in times of
war philosophers from Aristotle to
Hobbes to Montaigne have taken scarcity
to be an inevitable part of human
existence and aimed to tackle the
competition that arises from scarcity by
drawing on ideas of natural altruism
or by deferring to a Leviathan to
prevent all-out war Plato and Rousseau
believed that
division of labor directly leads to
inequality and injustice Adam Smith saw
division of labor as an engine for
economic growth in a technical necessity
but also leading to corruption and
degeneracy Karl Marx saw division of
labor as alienation from one’s true
nature and the results of increase in
class hierarchies as we see philosophers
have always assumed that scarcity
division of labor and death are
permanent parts of human life and that
this need to work and this desire to
always have more than your neighbor even
when he doesn’t have enough is just part
of human nature
justice conceptualized from scarcity is
all about property fairness and avoiding
death self-worth conceptualized from
division of labor is all about how much
you produce and how valuable your work
is to other people but what would the
world look like if these things didn’t
exist what would justice look like if we
weren’t worried about running out of
food or water would there still be
inequality if everyone had the same
access to resources how do we determine
our self-worth if it’s not based on the
value of our work this brings me it’s
the second thing that these scenarios
have in common which is that they will
be made possible by the rise of
artificial intelligence AI has been
described as the last invention we’ll
ever need to make driving the 4th
Industrial Revolution AI has two forms
artificial narrow intelligence and
artificial soup general intelligence AI
is already all around us in the form of
artificial narrow intelligence these
forums are really good at accomplishing
one task and well and some of them are
at or have surpassed human level
abilities these include computers that
play chess or go self-driving cars that
new Gmail feature that generates short
responses for you the ability of Netflix
to suggest yet another 90s rom-com for
me
self med in medical imaging diagnostics
and that facial recognition feature in
iPhones artificial general intelligence
can be roughly described as an AI that
can accomplish any intellectual task
that a human being can another term
super intelligence is an AGI that is far
better at performing any and every task
that a human being can this includes of
course manufacturing and computation but
also art writing teaching producing
music and even perhaps philosophy and so
in this one machine is the next
Shakespeare the next Rembrandt the next
Bach the next Plato but also the best
surgeon investment banker and architect
in the world so whether it’s high level
artificial narrow intelligence or
artificial superintelligence in the
future we will have machines that are
capable of innovating in radically new
ways and finding solutions to the
world’s biggest problems
hey I could eliminate scarcity either by
maximizing the efficiency by which we
allocate the resources we already have
or by expanding our resources by
genetically modifying crops to yield
bigger harvests or by directly growing
meat and vegetables from stem cells hey
I could systematically screen every
cancer genome much more quickly than a
human then a team of humans can and
prescribe personalized medicine for each
unique cancer patient hey I put screen
and changed the mutations for terminal
diseases while the baby is still in the
womb it could reverse manmade global
warming by directly reducing atmospheric
co2 levels ultimately these are all
empirical optimizable questions that can
be tackled by algorithms so if all the
scenarios that I listed earlier seem
like science fiction or Society hundreds
or thousands of years from now I’m here
to tell you it’s not artificial narrow
intelligence is already here and from a
survey of the world’s leading AI experts
artificial super intelligence
Agence is possible or even likely to
happen by 2060 but even if the opposite
were true if all scientists agreed
that’s part of that superintelligence is
impossible or that AI can’t do
everything that humans can like falling
in love or having free will
policymakers shouldn’t ignore the
political implications of artificial
narrow intelligence super intelligence
is one thing but there are many
political issues to tackle before we get
there under estimating our future
technological progress in this area is
poor planning this is the failure to
recognize that technology is and always
will be a political issue technological
progress is not deterministic and the
application of technology may not be
convergent nor moral determining how or
even if we should pursue these
advancements is ultimately determined by
our political nature so what is our
political nature
well politics has been defined as the
study of who gets what when and how and
these advancements will ultimately
challenge our existing assumptions
regarding the who the what and the how
whether a is developed by governments or
companies democracies or authoritarian
regimes is extremely important
first off whether AI progress occurs
through companies or governments will
ultimately determine how that AI
technology is applied governments have
the incentives to ensure the safety and
prosperity of its citizens maintain
regime stability and pursue global peace
whereas companies like Apple Facebook
Google and Microsoft are driven by
profits and subject to market
competition but there are advantages and
disadvantages to each the advantages of
a government taking control are that it
would have full control over funding
different projects and determining the
research agenda it would be incentive I
and have the power to pursue certain
ends like decreasing economic inequality
or even just stalling the progress of
self-driving cars to let infrastructure
catch up and these ends cannot always be
reached by market forces even with the
same technology available the
disadvantages of a government taking
control are authoritarian or perhaps
even totalitarian tendencies that giving
a government control over a technology
that will determine our jobs our
well-being in our lives may be granting
it too much power but if companies were
to be the sole actors in AI progress and
they begin amassing political influence
or infringing on certain rights and
freedoms how much should a government
intervene this question becomes
especially important when we consider
that the two nation-state frontrunners
in AI are the United States and China
two countries with arguably very
different forms of governance that stem
from different political theories and
have different paths of innovation
whereas AI research was heavily
dominated by the u.s. in their late 20th
and early 21st centuries China has
quickly established itself as an AI
heavyweight between 2010 and 2014 China
saw a 186 percent increase in AI patent
applications whereas the US only saw an
increase of 26 percent Chinese startups
have won International speech and facial
recognition competitions in China’s
largest Internet companies Baidu
Alibaba and 10 cent have enormous
amounts of data to work with from
e-commerce social media and search
engines with over 750 million Internet
users and so given that these two
countries are pretty much neck-and-neck
when it comes to AI capabilities what
are some of the most important political
issues to consider the some of them
include job retraining balancing
economic growth from automation with
massive job loss and
or growing inequality reallocation of
resources regulating police and military
power once all weapons become automated
cyber security privacy and freedom of
speech now I’m not gonna have the time
to go into all of these in detail but
let’s start with jobs perhaps the most
immediate policy issue in the
foreseeable future
ai will be able to complete most of our
jobs better than we can possibly
rendering the very concept of jobs
obsolete how will each governments
handle the transition from lives or
revolve around work to lies or revolve
around something else will they continue
with the assumption that jobs are
absolutely necessary to happiness and
fulfilment and an intrinsic desire of
humans will they pay citizens to
complete mundane tasks or with a
universal basic income or do away with
jobs altogether without human labour
what are we striving towards as
individuals once AI begins expanding our
resources or our ability to allocate
resources which government will enforce
new welfare policies how will each
re-evaluates
existing economic inequalities and our
current focus on retributive justice
will they converge upon a new definition
of justice when there’s no more
competition for resources and no more
need for wealth disparities another
important policy issue is surveillance
and the use of AI and facial recognition
the Chinese government has already begun
collaborating with companies that
develop this technology to enhance its
social credit system which is currently
in the works but may theoretically
assign each citizen a score that
determines loan statuses insurance
premiums international travel abilities
and even access to education for their
children cameras all over China matched
the faces of citizens with their social
media information medical data spending
habits and criminal records to catch
crimes as they happen but also to
increase trustworthiness by policing
behavior
what the us do the same or would this be
considered an infringement on privacy
now that we have the technology to
regulate immoral behavior and catch
criminals what do we do with it finally
how will each government’s approach the
ethical implications of AI is morality
universal or as a relative as an example
let’s take self-driving cars how will
each government approach the question of
whether a self-driving car should ever
sacrifice the life of a singular driver
for the lives of a bus full of children
or in this case the lives of three
pedestrians for five in the car and who
is liable if it makes the wrong decision
will these two nations reach the same
conclusion to this real-life trolley
problem my answer is not necessarily
but I hesitate to make a judgement on
who would make the morally better
decision as another example let’s take a
gorilla’s that estimate risk assessment
for crimes in 2016 in the United States
a program wrongly labeled black
defendants as twice as likely to
reoffended white defendants this was not
due to the algorithm itself but the
biases that already exist in the US
justice system and this is worrying
implications so whatever prejudices our
nation’s have will show through in the
data that AI relies on in this case in
this case that was arrest records
postcodes social affiliations and income
and because each nation and its data
have different prejudices they will also
create different AI programs and so that
leaves me with three questions one which
nation has fewer or at least less
prejudice biases and would create more
just AI programs to which nation is
working to overcome these prejudices in
a meaningful way so that they aren’t
programmed into machines and three are
there certain applications of AI that no
nation should be allowed to pursue
finally if and when we reach
superintelligence its values must be
perfectly aligned with ours or else that
optimizes for the wrong goal we may not
be able to predict the actions of a
super-intelligent machine and this
presents an existential risk benign
example of this is open a eyes program
that plays coast runners which is
basically this old-school video game
that’s where the goal is to finish a
boat race as quickly as possible but the
but the game rewards its players with
points for hitting certain targets laid
along the route and so when the program
played this game it found in island to
circle around over and over ultimately
going in the wrong direction but having
a higher score obviously amidst the
whole point of the game which was to
cross the finish line but it did exactly
what it was told to do which was
maximize the score so because we are
unable to bridge the actions of machines
the way we can with humans if we fail to
specify exactly what we want we face a
king – problem getting what you asked
for may not always be what you want but
instead of turning everything at touches
into gold super intelligent AI as a
perfectly rational actor may take any
small task like manufacturing paperclips
or optimizing a financial algorithm to
the extreme and its actions may be
irreversible it may decide that the most
efficient way to do these tasks is to
turn all of the carbon in ourselves into
paperclips or to manipulate the stock
market and destroy the global economy or
to colonize a universe and turn it all
into a huge supercomputer to run
simulations and I know this sounds a
little extreme and far-fetched but the
inability to predict the actions of a
more intelligent being is exactly why we
have control over the animal kingdom we
determine whether other animals live or
die
not because we are physically superior
but because we are more intelligent and
so a super intelligent AI
must have its values perfectly aligned
with ours which includes not killing any
human being not colonizing the universe
and maintaining happiness and prosperity
and freedom in human life but it may
also include values that are somewhat
less universal even the first one not
killing any humans may not be universal
if we decide that some lives are worth
more than others we may value the lives
of the moral over those of the wicked
the lives of those who live now over
those of the future we may decide to
achieve perfect equality or to maintain
current socio-economic hierarchies for
stability to maintain current democratic
and authoritarian regimes or to do away
with governments altogether and pursue
anarchy now that there’s no more
competition for resources and because
superintelligence is so difficult to
control and potentially impossible to
compete against the first nation or
government the first government or
company to reach superintelligence may
determine the set of values that the
rest of society and future generations
will operate under a nation-state may
decide to maintain current global power
dynamics or to ensure global hegemony
for itself and so my question is how
will each government concretely and
perhaps permanently define its values
and which nation’s values align with
those of all of humanity now to take a
step back for a second why do I ask
these questions in the first place is it
useful to think about what society will
look like a hundred years from now is it
possible to control the who the what and
the how of politics and technology I
think it is I think each governments
must begin preparing for the future and
which will be dominated by AI and for
these different scenarios if we
eliminate resource scarcity jobs and
death and if we reach superintelligence
and they should begin soon
the magnitude of change the likelihood
of rapid AI progress and the potential
permanency of global power conferred
upon a nation by AI Russian President
Vladimir Putin recently declared AI is
the future not only for Russia but for
all humankind
it comes with colossal opportunities but
also threats that are difficult to
predict whoever becomes the leader in
the sphere will become the ruler of the
world and so the United States and China
must anticipate how the other will
employ this technology and how they will
differ they must represent the public
interest and think critically about the
applications of this technology in fact
this is a global public interest and
thus global governance is needed as a
neutral non partisan species wide third
party perspective to ensure global
prosperity and peace AI is the most
important political issue of our
generation because it will revolutionize
how we approach some of the most
fundamental questions in political
philosophy our history of assumptions
regarding human nature civil society and
existence will be turned on its head
how should society be structured when
there’s no more competition for
resources how do we define our
self-worth in the absence of jobs and
money should we pursue immortality what
are the values and morals that we should
program into superintelligence who in
the first place should answer these
questions should it be companies like
Apple and Facebook or governments like
those of the US and China or a global
governance body these are the most
important political questions of our
generation that we have never had to
consider before and the questions that
we must begin assessing now thank you
[Applause]
[Music] you