Press "Enter" to skip to content

Human Values and Power in a World of Artificial Intelligence | Olaf Groth | TEDxHultAshridge


ladies and gentlemen we stand at a
watershed moment in humanity’s vast
digital uncertain future
a potent technology artificial
intelligence has been developing with
breathtaking speed thanks in large part
to advances in neural network
technologies that are roughly modeled
after the human brain AI is able to
ascertain patterns in massive
unstructured data sets make
recommendations even make decisions with
ever increasing precision because it
learns by itself without the
interference of messy complicated
emotional human beings fascinating and
scary at the same time isn’t it this is
why I think we need to ask ourselves
this fundamental critical question for
humankind
how will we shape and govern the power
of AI so that it serves human values and
human power this man here is I’m sure
you’ll agree knows a fair bit about
power for better or for worse and
apparently he knows about the
combination of artificial intelligence
and power this is hot off the press is
quoted in a meeting just last week with
young AI leaders as saying those who
lead in the development of AI will end
up ruling the world I hate to tell you
in this particular instance he may be on
to something because AI is about power
it’s about the power of the technology
and how it relates to and interferes
with human power and the power of
societal institutions that is why we
need to get in ahead of this and make
sense of it before the machine that we
created the technology run circles
around us faster than we can count the
laps let me back up and talk about a
person that is closer to me than mr.
Putin luckily this is my daughter Hannah
she’s 12
let me ask you hands up if you have
children thank you I’m glad we’re not
suffering from any extinction problems
in this room children are great they
make you think about the future your
children my children will live in a
world that is purvey
safely permeated by artificial
intelligence this is not science fiction
this is not some unbelievable future
this is starting right now Hannah is
starting to play a rich computer games
providing information about herself her
cognitive patterns her behavior social
competitive potentially even her
socioeconomic status and that of her
family as we’re starting to pay for
games in a few years she’ll be driving a
car or she’ll be driven in an autonomous
car and she would be providing
information about herself to automakers
and insurance companies about her risk
profile about her Geographic movement
patterns thanks to GPS and AI coming
together about her social and emotional
needs as she communes in the car with
other people and as she draws content
games music from the web into the car
the Machine the a I will then mesh all
of this and draw conclusions on her
quote unquote pattern of life her
pattern of life it will then make
predictions about what Hanna might want
to do or need to do next
tremendously useful of course and also
problematic
let me project forward to let’s say I
don’t know 2030 Hannah goes to the
doctor there’s a medical AI assisting
the doctor the AI will be able to look
at disparity de stream such as her
nutritional intake and food preferences
her family history and DNA the
geographic area she lives in and what
the pollution levels are they’re even
stress levels in her relationships and
at work and it will be able to make a
prediction about Hannah’s likelihood of
incurring breast cancer at some point in
the medium-term future now right about
now you might say but that’s fantastic
she now has choices that you and I never
had and that may be so unless of course
the Machine says the AI says Hannah
you’re gonna have to follow my
recommendations for drastic changes in
your lifestyle if you want to stay in
short now I don’t know about you but I
think this should be Hannah’s choice it
should be driven by Hannah’s value set I
may have my preferences for Hanna’s
well-being but this is her as a person
calling these shots privately I might
add so we need to make sure that she
retains the right to self-determination
on these critical questions let me talk
a little bit more about values because
values are at the core of how we as
human beings are empowered in society
you might think we transact commercially
or in social media and give a fairly
holistic picture of who we are but
that’s not true it’s a very partial
picture you have dormant values you have
emergent values your values change as
you stumble through life serendipitously
meeting people meeting challenges
feeling failure and and success pain and
joy you evolve and so does your value
system you make hair line thin
trade-offs between values every day now
this complicated picture needs to be
included in the governance of AI maybe
even your personal assistant AI now not
to make things too complicated on top of
this society also has an interest in
this and has its own hierarchy of values
how will that hierarchy expressed in
political processes and community
conflict with your value hierarchy as
you jointly govern the AI let me give
you an example it’s called the Somalia
problem what if all of a sudden your
personal AI will tell you I’m sorry boss
but I’m gonna have to go to Somalia
because our government has summoned us
to go help with a humanitarian crisis
and then you say to the AI let’s call
him George George you can’t leave I need
you here in say California to help me
manage our water problems I’m sorry boss
but starving children more important
than water in California but George
we’ve got sunshine and marijuana here
and and it says dude I’m a software
program and off it goes to Somalia right
so so who determines what values govern
a
that’s a critical question let me talk a
little bit more about values
who here has how to run-in with the law
I’m sorry you can’t answer this cuz
we’re on camera but let me ask you then
who here has seen the movie Minority
Report right great movie
science fiction but it isn’t it’s
actually here now
it’s called predictive policing the NYPD
is working with big data companies to
get predictions on where in the city
crime might arise and then they can
preemptively dispatch officers to those
locations now right about now you might
feel a lot safer as a tourist going to
New York but the communities in question
feel stigmatized and they’re asking the
NYPD how did your machine arrive at
these conclusions and the machine can
tell them that similarly some judges in
America are getting assistance from
machine learning algorithms AI on
determining bail payments for alleged
offenders and yet the machines can tell
the judges how they arrived at these
conclusions now I think that’s a problem
in a society that’s governed by the
transparent rule of law and because of
that the US government represented by
DARPA the Defense Advanced Research
Projects Agency that’s a mouthful has
launched a program called explainable AI
so they’re using one AI to explain
another AI right now we have remember
this we have layers upon layers upon
layers of values and layers upon layers
upon layers of AI who is going to govern
all of this transparently right it’s a
big issue for us to solve I want to go
from policing to psychology 45 million
Americans have some kind of
psychological trouble or another and yes
that’s gotten worse since the election
I’m sorry to say 80 percent of them
don’t think they can get adequate
accessible inexpensive care income the
psychotherapy bots eliza le woo bot and
sim sensei some of these are developed
by medical experts and the military in
order to provide relief to soldiers who
are suffering from trauma depression
increased stress levels right great
service to them but we need to be
conscious that this kind of technology
is now permeating
other areas of life who here has
interviewed for a job let’s say over the
last five years all right some of you a
lot of you in fact that tells me you
know how anxiety evoking that is this
kind of technology here facial
recognition voice recognition body
language recognition algorithms are
comparing you analyzing you in your
interview performance against that of
other interviewees then they’re using
that to compare all the interviewees to
other people on the inside of the
company who have already been performing
well on these jobs now they want to
predict how well you can do in these
jobs understandable but uncanny isn’t it
because you thought you were in control
of your own appearance so we need to
talk about this I want to go to the
political realm now and think big China
is spending billions of dollars on AI to
assign a trust rating to each one of its
citizens a publicly available rating why
are they doing this they want to
facilitate peace and harmony for the
glorious harmony society to increase
peaceful transactions in the economy
understandable given China’s history but
also scary why who’s in charge of these
ratings what kind of recourse does a
citizen have to correct
misunderstandings or mistakes and by the
way aren’t we always told we should
learn from mistakes so then are we gonna
get a bonus to make up for some
shortcoming on a mistake how is all this
going to be determined this has a direct
impact on the power to determine the
livelihood of a billion five Chinese
right a big thing I want to highlight
that there is a lot of benefit to a lot
of the things that I just told you I’m a
Technol I want to show you some other
things that are unequivocally positive
this is what I might call the social
Factory in Germany these are experiments
where the machine is learning basic
parameters physical psychological about
the human operator here on the left then
integrating with an operator more
effectively for more collaborative
higher productivity that is great for
economic growth and for
labor right you might have heard about
Japan’s aging society 25 percent of
Japan’s population is 65 years or older
they’re facing a looming crisis of a
million caregivers a shortage of a
million caregivers by 2025
therefore Japan has invested in robots
and an AI to facilitate longer healthier
less lonely lives for its elderly right
I think that’s a phenomenal idea I have
an aging parent on another continent I
would love to call up the AI let’s calm
George right and ask hey George how is
how’s mom doing right and get an
objective opinion on how she’s doing day
or night
instead of getting her on the phone
saying oh honey don’t worry about me I’m
fine
right you want that authenticity now so
therefore I think that’s a great thing
climate change nai possibly the biggest
impact positively speaking for all of
mankind humankind we have AI working
with the UN IPCC see the body that won
the Nobel Prize for its climate change
research and activism determining which
of its climate change models are the
most effective that will impact all of
Earth it doesn’t get any bigger than
this does it now you might ask where is
all of this going to go in the future
and there are four emerging perspectives
I leave you with the first one is the
upper left essentially saying that a lot
of people who are inventing AIO
innovating on it are introverted people
right that frankly don’t care very much
about human irrationality and big Uwe T
ambivalence political decisions it’s
messy put them all on a couch let the
Machine run the world right the second
version is a much bleaker version of
this which is that the machine the AI
will eventually eradicate us why because
either it has figured out that we are a
non optimizable function in this world
or because some evil door is going to
hijack it and kill all of us okay this
is mr. Hawkin or mr. musk version of the
future the third perspective
is that of the singularity the machine
the AI and the human essentially
becoming one right this is reycarts walk
reycarts Wells version and I suppose
that means we don’t have to be afraid of
AI because we’re gonna become one with
it okay
the fourth perspective is that we need
to keep the human being understood by
the Machine observed by the machine in
the center of the machines attention
always in control and empowered the best
the best representation I think of this
perspective is Stuart Russell and the
Center for human compatible AI at UC
Berkeley this ladies and gentlemen I
think is our ticket that’s our path we
need to go from the upper left to the
lower right bypassing the upper right
here in this picture altogether how do
we get there we need to do something
that we’re not usually very good at
which is learned from history go back
800 years King John returning from
France to England faced with 20-odd
angry barons saying we no longer want
your in transparent rule they sat down
together and eventually the Magna Carta
came out of it the Magna Carta was a
quasi contractual document that was the
precursor to the Renaissance the
Enlightenment and eventually
constitutional democracy I think we need
a new Magna Carta a digital Magna Carta
for the era of AI and that document
needs to put values based human
empowerment at the center let me say
this again because it’s important values
based human empowerment we have to get
there through a multi-stakeholder global
Congress this needs to be a forum for
what all of us and our children can be
represented not just the experts and
policy makers right if we can get this
done then I think I will sleep much
better and I would submit to you you
will sleep much better too knowing that
our children will live the kinds of
lives that are worth living thank you
very much
you
Please follow and like us: