The Extension of Human Intelligence
[Music]
[Applause]
hey siri
the weather today sunny with a high of
okay can you tell me a joke
sure why did the students eat their
homework
because their teacher said it was a
piece of cake
let me try asking something more
abstract hey siri
what’s the meaning of life
all evidence shows it’s chocolate
just now i had a little conversation
with siri and i’m sure many of you have
spoken to some sort of digital assistant
whether it be siri alexa bixby you name
it
it’s so convenient that sometimes we
feel that they’re even smarter than a
real person
until we give them some creative
questions the technology behind siri and
her relatives
is artificial intelligence or a.i
today from spotify which uses ai to
give us better music recommendations to
verizon
which uses ai to improve internet
services
to instagram using ai to tackle online
bullying
artificial intelligence is all around us
in fact ai is now called the fourth
industrial revolution
following steam electricity and the
internet
but for too many people artificial
intelligence seems scary
but don’t worry today i’ll talk about my
research on ai
its core mechanism and also address what
ai holds
for the future the concept of artificial
intelligence was first seen through
depictions of robots
in films such as metropolis in the 1920s
in 1956 the term artificial intelligence
was officially coined
at the dartmouth summer research
conference by a group of scientists from
dartmouth
harvard and ibm this was a monumental
moment to nail the flag to the mast
the field of ai was launched by the bold
vision
that computers could be made to perform
intelligent tasks
but when we talk about ai we often think
of those cool robots
but actually a.i and robotics are
separate fields of study
robotics deals with machines that
interact with the physical world via
sensors
on the other hand ai tackles learning
and logical
logical perception in short while ai
acts as the brain robots act as the body
but the evolution in both areas
continues
and the line between a i and robotics is
blurring
because when combined you can get an
artificially intelligent robot
and in 2018 spacex actually flew an
artificially intelligent robot named
simon
to the international space station to
better understand how ai works
let’s look at a simple example of how
this how the computer learns
this is a short python program called
the capital cities of the world
it works when you type a country’s name
into the search box and it’ll pull up
the country’s capital
now if a response wasn’t previously
reported then it will prompt you to put
in the right answer
now some of you may be saying oh i’ve
known country capital since i was in
elementary
i don’t need ai to do that for me
however this is only the simplest
learning process here’s another example
the price of a diamond can be determined
based on seven different features
for us humans considering two or three
features can be easy
but adding on more features may hurt our
brains
for a i considering five or even ten
more features to compute
is only a matter of milliseconds from
here you can see that ai can compute at
an incredibly complex
level to understand how ai works under
the hood
we can break ai down into two main
subsets
narrow ai and machine learning narrow ai
is good at executing predetermined
actions siri
is a great application of narrow ai upon
receiving your request
siri records the sound waves from your
voice and translates them into a code
she then breaks down the code to
identify particular patterns
and keywords and when you hear her
response
it’s not a person behind the scenes but
rather just a search result
from her tremendous database and when
siri doesn’t have an answer
she simply shows you a web search that
matches your keywords
and finally she converts the results to
an audible response
but for more complicated questions we
don’t have answers and there’s no
database to search
therefore there are no predetermined
actions so we as humans learn through
experience
by trial and error and make educated
guesses via neurons
in our brain guess what a.i learns the
same way
scientists are replicating this process
using artificial neural networks
and this is the core of ai called
machine learning
and this is super cool right who
wouldn’t want to make a good educated
guess when there’s no way of telling
what’s right i certainly do
so here at school i co-founded a
forecasting club
forecasting is the versatile skill of
predicting a future event analytically
we started with fun events such as the
us open and we’ll continue with the
upcoming elections
we’re using a basic software from
microsoft and just playing around with
the data
but i have to admit with so many moving
parts it’s difficult considering all the
factors
so moving forward we’ll invite local
university professors
and ai experts to help and teach us
along the way
now here’s how machine learning actually
works there are four major components
the first is the mail carrier who
collects data for ai to learn
then we have the construction worker who
builds the model
now not fashion model but a set of
functions that calculates the outputs
this is the simulation of the human
brain that’s why these models are also
called
neural networks then the police officer
the loss function comes in to monitor
the performance
calculate the errors and redistribute
them back into the model to be
recalculated
and finally the manager who makes the
calculation process
faster via a fancy jargon called the
optimizer
and this process repeats for several
rounds until the errors
are minimized curious on how to better
use ai
this past summer i participated in the
national center of atmospheric research
artificial intelligence summer school
working with scientists from stanford
university and
noaa i learned that essentially machine
learning is fitting
inputs and outputs into an algorithm
otherwise known as
curve fitting and now let’s look at an
example
from the summer school of rain
prediction to see how this whole machine
learning process plays out
we used temperature pressure and
moisture data as inputs
to predict the chance of rain we used
past rain events to train ai and obtain
a curve
we then use the curve and sets of
functions to calculate the output
and in the meantime we used physical
functions such as mass and energy
because the calculations still need to
follow physical laws such as mass
conservation
and this process repeats for several
rounds until the curve
is fit and as you can see this machine
learning process requires ai
experts in the field today to
collaborate with other disciplines such
as
environmental science medical science
engineering financial business
etc to create models that advance each
field
and teamwork is key for future
development
while the benefits of ai are tremendous
new technology can cause concerns
like every other innovation in human
history the risks
exist consider the invention of cars and
airplanes
we all understand the risks of traveling
by these means of transportation
but they have allowed traveling to
become so much easier
and even impacted areas outside of basic
transportation
such as logistics and emergency rescues
for the
reasons ai technology is built on the
foundations of
improving efficiency and scientists are
like the wright brothers
being brave and cautious to improve this
technology
another concern is human employment it
may be intuitive to think that the
growing development of ai
will eventually take over many existing
jobs
however the world economic forum states
that automation will generate 60 million
more new jobs than existing jobs
and i.t companies gartner and the
mckinsey global institute
echoed that thought now if you watch a
lot of science fiction movies like me
you might be thinking what if the robots
from the matrix and the terminator team
up with ai to dominate
humanity well oxford university released
a study claiming that a.i will
outperform humans and many activities in
the next 10 years
such as translating languages by
boosting efficiency
ai will allow us to see a decrease in
manual labor
and an increase in new jobs by
developing new products
and services our relationship with ai
should not be like
that seen in the terminator but rather
like that we
see in star wars with r2d2
in fact mit professor eric brynjolfsson
says that we can use the power of ai
to work on the elimination of global
poverty disease reduction
and the provision of better education he
tells us
not to ask what will happen but rather
what will we choose to do with this
technology thank you