good afternoon everyone
so we’ve all heard this saying that time
is money but it’s really not that simple
the idea is that you can convert time
which is really the only thing you have
anyway into money but it doesn’t always
work the other way around
sometimes your time or the time of
someone you love can be robbed away from
you and even though you’d gladly
exchange any amount of money to get that
time back money alone usually can’t fix
these kinds of problems so given that
time is your most precious your most
finite valuable resource what’s the one
thing you could do to protect your time
or to get more of it
well obviously we should be preparing
for the zombie apocalypse
so zombie briefing part 1 you’ve got to
get your survival kit your vehicle your
underground bunker munitions rations
plenty of potable water this is really a
lot to think about okay okay so maybe
I’m a doomsday prepper but maybe there’s
a principle here that also applies you
see I’ve personally found that the very
best uses of my time would indeed well
prepare me for a zombie apocalypse the
very best uses of my time make me
stronger faster and harder to kill even
so there’s some important trade-offs to
be made with the limited time that I
have everything that might make me
stronger may not necessarily make me
faster everything that makes me faster
may or may not make me harder to kill in
scientific terms we could call this a
multi objective optimization problem so
looking back on my life multi objective
optimization problems have really been a
guiding influence and I’ve come to think
a good paradigm for life in general I
was raised up
real West Virginia by my
great-grandparents now we were poor and
there was little chance of that changing
but whereas we were poor in money we
were not poor in values I understand why
that generation is called the greatest
generation they overcame something far
more severe than a zombie apocalypse and
I can assure you that 50 years later as
I was a kid growing up in the 1980s
surrounded by 80 year olds those values
were still on display daily grit
determination resilience perseverance
now I didn’t really think of it this way
at the time but I was absorbing those
values I was making them my own and
those values would lay the foundation of
my own personal quest to become stronger
faster and harder to kill in 1994 when I
was 13 something totally unexpected
happened
my maternal grandfather who I barely
even interacted with up to this point in
my life came for a visit now that was
unusual he lived out of state we had no
interaction but what made the visit even
more unusual was that he bought me a
brand new computer so remember back then
computers were quite expensive and
pretty rare even more so on the road
where I grew up but this computer would
be the most amazing gift it would
fundamentally change the course of my
life the operating system on this
computer was windows 311 that was always
so interesting and I was quickly seduced
by the dark side the DOS terminal so
this is where I would spend countless
hours and I would become completely
obsessed learning how to program and
this actually turned into a real source
of tension in my life as I had to start
thinking more seriously about trading
off my time I could continue doing what
I had been doing status quo or I could
increasingly spend time in this solitary
activity that no one around me seemed to
really care about or understand now the
timing here was pretty good because as I
wandered further and further down this
Aaron thought logic that was programming
I noticed that I started to think about
the world a little bit differently more
systematically more logically the way I
was programming and I was right up
against the biggest optimization problem
of my life at this point figuring out
how I would escape the cultural inertia
around me how would I get into a college
after high school and get this paid for
and what were all the things I would
have to do to try to make that happen to
create some space for myself to to go on
and and be somebody in the world well on
June 30th 1999 just a couple of weeks
after graduating high school I showed up
to Colorado Springs at the Air Force
Academy for what would be the hardest
four years of my life up to that point
and a multi objective optimization
problem unto itself I can assure you but
the values of the greatest generation
combined with this systematic thinking
and this relentless desire to manage my
time it began to pay off handsomely I
would graduate at the top of my major in
computer science and right after being
commissioned as a new second lieutenant
I would go right into grad school and
you’ll never guess what I studied multi
objective optimization problems so I’ll
spare you all of those computer science
lecture notes I’ll spare you all of the
fine print of the master’s thesis and
I’m just going to tell you the most
important thing that I learned in grad
school which was this if you’re really
serious about solving multi objective
optimization problems the very best way
to do it is with software that can learn
from its own experience the very best
way to do it is with artificial
intelligence
you see even with the most powerful
computers we can’t try all of the
possible combinations fast enough to
ever find the breakthroughs that we’re
looking for and nor can we just sort of
wander through the space intelligently
guessing we just can’t intelligently
guess through that space fast enough
just after grad school I was a newlywed
I just recently married my wife and
during this time we were still getting
to know one another and we well she
noticed she had some vision problems
that were kind of coming out of nowhere
we went and we got this checked out and
we we received the most devastating news
she had a she had a cyst growing in her
brain and it had to be surgically
removed as soon as possible now the good
news is that most of the cyst was
removed the bad news is the vision
problems just got worse
the bad news is she contracted
meningitis it was misdiagnosed it was
mistreated with extended and high levels
of steroids which just wrecked every
glandular function in her body just
completely wrecked her into this day we
are working through what I think of in
my own mind is a big multi objective
optimization problem trying to find a
way to restore her health so I asked you
earlier what’s the one thing you could
do to protect your time or to get more
of it more time for you and your family
more time for me and my family more time
for all of us well for me
I’ve increasingly believe that the very
best use of my time is to work on what I
think is maybe the most meaningful and
perhaps even the most complex
multi-objective optimization problem of
them all healthcare healthcare is
insanely complex I mean it literally
encompasses every human being on the
planet decisions mean the difference
between life death or lifetimes of
wellness or chronic suffering this just
cuts right at the core of our humanity
and in health care everything is it’s
scale I’m not aware of any small
problems in health care here’s what I’ve
become completely convinced of you can
look at health care and the pockets
within it as multi objective
optimization
Asian problems and it’s very convenient
to do so because you can start to
quantify it in ways that machine
learning and artificial intelligence can
help us to solve the big problems to get
the leverage that we need to
increasingly make those breakthroughs in
the space now I’ll be honest I have a
pretty conservative view of what I think
is possible with artificial intelligence
but even so I think it’s the most
disruptive thing I’ll see in my life and
as I think about it I think back to
computers and then we networked those
computers together with the internet we
shrink down the form factor which makes
them easier to distribute and get to the
7 billion or so people in the world
social networks and apps grow up and we
begin to collect data about every last
part of our lives now that actually
sounds like the perfect setup for
software that can learn from all that
data software that can improve all
aspects of our lives especially our
health and our healthcare I’d like to
share just a few examples of where AI is
already making a difference in
healthcare and the first is in the
personalized genomic space and I’ll sort
of frame this within the broader
quantified self context sequencing the
genome is a big deal and that’s amazing
we’ve done that work now it’s time to
get busy
understanding the genome understanding
well what are all of those genes and
what all those genes really do and how
do they interact with one another and
express themselves based on whether or
not other genes are or aren’t present I
mean that’s that’s actually beginning to
sound like oh I don’t know
multi-objective optimization problem
perhaps that’s starting to sound like
exactly the kind of situation where we
can use artificial intelligence to get
the leverage we need to make the
breakthroughs happen even more so I’ll
give you an example of this some of you
may already know you could leave here
today you could buy a kit for a little
over a hundred dollars you spit in a
tube you mail it in and a couple of
weeks later you get back this amazing
collection of continually updating
digital reports these reports tell you
everything from your likely tolerance to
caffeine or lactose food allergens which
can make a big difference on your health
and wellness all the way to whether or
not you’re carrying certain recessive
traits whether or not these traits may
be important to know about is you have
children of your own or as you continue
to age I’ve been a quantified self
junkie and I’ve been into this stuff for
years keeping detailed nutrition logs
and workout journals and this stuff is
helping me to take my own health and
wellness to a whole new level and I’m
convinced is you look into it and start
to apply it to your own lives it’s gonna
make you stronger faster and harder to
kill as well now the next thing I’d like
to make sure is on your radar is
machines that can read machines that
process human communication that’s
really nothing new we’ve been doing that
since the beginning the old computer my
grandfather bought me could do that just
fine thank you
but what’s different now is machines
that can actually understand that human
communication to drive faster and better
decisions now this is something very
different this is a very big deal if
you’ve ever studied language it’s a hard
thing to wrap your head around you have
all of these words that mean something
and the context of those words changes
depending on what other words are around
them and you know when they were said
and what information has or hasn’t
superseded them and what you might
already know and I think you already see
where this is going this this is
starting to sound familiar isn’t it this
is starting to look like one of those
multi objective optimization problems
and I can assure you machine learning is
advancing the state of the art here
almost daily this is where I’ve spent
most of my own professional career this
is the wave of the future machines that
are understanding language let me give
you an example in health care of where
this is already making a difference for
a couple of years we’ve been working
with one of the large provider systems
here in town and we’ve built a system
that reads pathology reports we’re using
machines to determine if a patient has
cancer and if so
which kind of cancer they have now to be
clear we’re not interested in automating
away md’s physicians nurses those types
of care providers at all what we are
doing here is we are providing a machine
that can read to intelligently triage
the information that is administrative
it’s the overhead what this ultimately
allows the physician to do is to spend
more time empathizing delivering that
human touch that makes such a difference
when you may have just received the
absolute worst news of your life and
this is a win for everyone it’s it’s a
win for the provider it’s definitely a
win for the patient it’s it’s a win for
the whole health care system it’s making
the whole thing stronger faster and
harder to kill the final example is
stem-cell therapy and this is a
particular interest to me
because of all the chronic conditions
that my wife and I have been through
over the years now much like in the
example I told you about with genomics
understanding the complex interactions
between stem cells and damaged tissues
in the body’s understanding whether or
not an individual is a good candidate
for a particular expensive experiment
experimental stem cell therapy based
upon the chemical signatures in their
body and and what else is going on in
their health that’s that’s sounding
familiar I hope I hope you’re already
thinking I hope you beat me to it that
is a multi objective optimization
problem unto itself and just like in
genomics I mean it’s no surprise that
our bioinformatics community has been
using machine learning for a number of
years but the advancements in the
broader space the increasing speed of
hardware the increasing commoditization
of the open source tools well it’s it’s
benefitting them and they are able to
move even faster to deliver these new
stem cell therapies to all of us who
need them and I’m really happy to say my
wife she was determined to be a good
candidate for one such therapy just last
month we flew out to Arizona so she
could receive that therapy now it’s too
early to know
if this is the miracle we’ve been
looking for in our own big multi
objective optimization problem that
we’re trying to solve but I’m hopeful
I’m optimistic that it’s going to make
her and others like her stronger faster
and harder to kill than ever before so
whether it’s advancing the art of what’s
possible in medicine whether it’s
preparing for the zombie apocalypse
whether it’s just a fresh perspective
for life I hope you see that the
paradigm of a multi objective
optimization problem is a useful one
because of the way it helps you to frame
problems because of the way you can
quantify them and then begin to apply
intelligent machines and artificial
intelligence to get the leverage that
leads to breakthroughs so that you me
and all of us can become stronger faster
and harder to kill thank you
[Applause]
you
[Applause]
so we’ve all heard this saying that time
is money but it’s really not that simple
the idea is that you can convert time
which is really the only thing you have
anyway into money but it doesn’t always
work the other way around
sometimes your time or the time of
someone you love can be robbed away from
you and even though you’d gladly
exchange any amount of money to get that
time back money alone usually can’t fix
these kinds of problems so given that
time is your most precious your most
finite valuable resource what’s the one
thing you could do to protect your time
or to get more of it
well obviously we should be preparing
for the zombie apocalypse
so zombie briefing part 1 you’ve got to
get your survival kit your vehicle your
underground bunker munitions rations
plenty of potable water this is really a
lot to think about okay okay so maybe
I’m a doomsday prepper but maybe there’s
a principle here that also applies you
see I’ve personally found that the very
best uses of my time would indeed well
prepare me for a zombie apocalypse the
very best uses of my time make me
stronger faster and harder to kill even
so there’s some important trade-offs to
be made with the limited time that I
have everything that might make me
stronger may not necessarily make me
faster everything that makes me faster
may or may not make me harder to kill in
scientific terms we could call this a
multi objective optimization problem so
looking back on my life multi objective
optimization problems have really been a
guiding influence and I’ve come to think
a good paradigm for life in general I
was raised up
real West Virginia by my
great-grandparents now we were poor and
there was little chance of that changing
but whereas we were poor in money we
were not poor in values I understand why
that generation is called the greatest
generation they overcame something far
more severe than a zombie apocalypse and
I can assure you that 50 years later as
I was a kid growing up in the 1980s
surrounded by 80 year olds those values
were still on display daily grit
determination resilience perseverance
now I didn’t really think of it this way
at the time but I was absorbing those
values I was making them my own and
those values would lay the foundation of
my own personal quest to become stronger
faster and harder to kill in 1994 when I
was 13 something totally unexpected
happened
my maternal grandfather who I barely
even interacted with up to this point in
my life came for a visit now that was
unusual he lived out of state we had no
interaction but what made the visit even
more unusual was that he bought me a
brand new computer so remember back then
computers were quite expensive and
pretty rare even more so on the road
where I grew up but this computer would
be the most amazing gift it would
fundamentally change the course of my
life the operating system on this
computer was windows 311 that was always
so interesting and I was quickly seduced
by the dark side the DOS terminal so
this is where I would spend countless
hours and I would become completely
obsessed learning how to program and
this actually turned into a real source
of tension in my life as I had to start
thinking more seriously about trading
off my time I could continue doing what
I had been doing status quo or I could
increasingly spend time in this solitary
activity that no one around me seemed to
really care about or understand now the
timing here was pretty good because as I
wandered further and further down this
Aaron thought logic that was programming
I noticed that I started to think about
the world a little bit differently more
systematically more logically the way I
was programming and I was right up
against the biggest optimization problem
of my life at this point figuring out
how I would escape the cultural inertia
around me how would I get into a college
after high school and get this paid for
and what were all the things I would
have to do to try to make that happen to
create some space for myself to to go on
and and be somebody in the world well on
June 30th 1999 just a couple of weeks
after graduating high school I showed up
to Colorado Springs at the Air Force
Academy for what would be the hardest
four years of my life up to that point
and a multi objective optimization
problem unto itself I can assure you but
the values of the greatest generation
combined with this systematic thinking
and this relentless desire to manage my
time it began to pay off handsomely I
would graduate at the top of my major in
computer science and right after being
commissioned as a new second lieutenant
I would go right into grad school and
you’ll never guess what I studied multi
objective optimization problems so I’ll
spare you all of those computer science
lecture notes I’ll spare you all of the
fine print of the master’s thesis and
I’m just going to tell you the most
important thing that I learned in grad
school which was this if you’re really
serious about solving multi objective
optimization problems the very best way
to do it is with software that can learn
from its own experience the very best
way to do it is with artificial
intelligence
you see even with the most powerful
computers we can’t try all of the
possible combinations fast enough to
ever find the breakthroughs that we’re
looking for and nor can we just sort of
wander through the space intelligently
guessing we just can’t intelligently
guess through that space fast enough
just after grad school I was a newlywed
I just recently married my wife and
during this time we were still getting
to know one another and we well she
noticed she had some vision problems
that were kind of coming out of nowhere
we went and we got this checked out and
we we received the most devastating news
she had a she had a cyst growing in her
brain and it had to be surgically
removed as soon as possible now the good
news is that most of the cyst was
removed the bad news is the vision
problems just got worse
the bad news is she contracted
meningitis it was misdiagnosed it was
mistreated with extended and high levels
of steroids which just wrecked every
glandular function in her body just
completely wrecked her into this day we
are working through what I think of in
my own mind is a big multi objective
optimization problem trying to find a
way to restore her health so I asked you
earlier what’s the one thing you could
do to protect your time or to get more
of it more time for you and your family
more time for me and my family more time
for all of us well for me
I’ve increasingly believe that the very
best use of my time is to work on what I
think is maybe the most meaningful and
perhaps even the most complex
multi-objective optimization problem of
them all healthcare healthcare is
insanely complex I mean it literally
encompasses every human being on the
planet decisions mean the difference
between life death or lifetimes of
wellness or chronic suffering this just
cuts right at the core of our humanity
and in health care everything is it’s
scale I’m not aware of any small
problems in health care here’s what I’ve
become completely convinced of you can
look at health care and the pockets
within it as multi objective
optimization
Asian problems and it’s very convenient
to do so because you can start to
quantify it in ways that machine
learning and artificial intelligence can
help us to solve the big problems to get
the leverage that we need to
increasingly make those breakthroughs in
the space now I’ll be honest I have a
pretty conservative view of what I think
is possible with artificial intelligence
but even so I think it’s the most
disruptive thing I’ll see in my life and
as I think about it I think back to
computers and then we networked those
computers together with the internet we
shrink down the form factor which makes
them easier to distribute and get to the
7 billion or so people in the world
social networks and apps grow up and we
begin to collect data about every last
part of our lives now that actually
sounds like the perfect setup for
software that can learn from all that
data software that can improve all
aspects of our lives especially our
health and our healthcare I’d like to
share just a few examples of where AI is
already making a difference in
healthcare and the first is in the
personalized genomic space and I’ll sort
of frame this within the broader
quantified self context sequencing the
genome is a big deal and that’s amazing
we’ve done that work now it’s time to
get busy
understanding the genome understanding
well what are all of those genes and
what all those genes really do and how
do they interact with one another and
express themselves based on whether or
not other genes are or aren’t present I
mean that’s that’s actually beginning to
sound like oh I don’t know
multi-objective optimization problem
perhaps that’s starting to sound like
exactly the kind of situation where we
can use artificial intelligence to get
the leverage we need to make the
breakthroughs happen even more so I’ll
give you an example of this some of you
may already know you could leave here
today you could buy a kit for a little
over a hundred dollars you spit in a
tube you mail it in and a couple of
weeks later you get back this amazing
collection of continually updating
digital reports these reports tell you
everything from your likely tolerance to
caffeine or lactose food allergens which
can make a big difference on your health
and wellness all the way to whether or
not you’re carrying certain recessive
traits whether or not these traits may
be important to know about is you have
children of your own or as you continue
to age I’ve been a quantified self
junkie and I’ve been into this stuff for
years keeping detailed nutrition logs
and workout journals and this stuff is
helping me to take my own health and
wellness to a whole new level and I’m
convinced is you look into it and start
to apply it to your own lives it’s gonna
make you stronger faster and harder to
kill as well now the next thing I’d like
to make sure is on your radar is
machines that can read machines that
process human communication that’s
really nothing new we’ve been doing that
since the beginning the old computer my
grandfather bought me could do that just
fine thank you
but what’s different now is machines
that can actually understand that human
communication to drive faster and better
decisions now this is something very
different this is a very big deal if
you’ve ever studied language it’s a hard
thing to wrap your head around you have
all of these words that mean something
and the context of those words changes
depending on what other words are around
them and you know when they were said
and what information has or hasn’t
superseded them and what you might
already know and I think you already see
where this is going this this is
starting to sound familiar isn’t it this
is starting to look like one of those
multi objective optimization problems
and I can assure you machine learning is
advancing the state of the art here
almost daily this is where I’ve spent
most of my own professional career this
is the wave of the future machines that
are understanding language let me give
you an example in health care of where
this is already making a difference for
a couple of years we’ve been working
with one of the large provider systems
here in town and we’ve built a system
that reads pathology reports we’re using
machines to determine if a patient has
cancer and if so
which kind of cancer they have now to be
clear we’re not interested in automating
away md’s physicians nurses those types
of care providers at all what we are
doing here is we are providing a machine
that can read to intelligently triage
the information that is administrative
it’s the overhead what this ultimately
allows the physician to do is to spend
more time empathizing delivering that
human touch that makes such a difference
when you may have just received the
absolute worst news of your life and
this is a win for everyone it’s it’s a
win for the provider it’s definitely a
win for the patient it’s it’s a win for
the whole health care system it’s making
the whole thing stronger faster and
harder to kill the final example is
stem-cell therapy and this is a
particular interest to me
because of all the chronic conditions
that my wife and I have been through
over the years now much like in the
example I told you about with genomics
understanding the complex interactions
between stem cells and damaged tissues
in the body’s understanding whether or
not an individual is a good candidate
for a particular expensive experiment
experimental stem cell therapy based
upon the chemical signatures in their
body and and what else is going on in
their health that’s that’s sounding
familiar I hope I hope you’re already
thinking I hope you beat me to it that
is a multi objective optimization
problem unto itself and just like in
genomics I mean it’s no surprise that
our bioinformatics community has been
using machine learning for a number of
years but the advancements in the
broader space the increasing speed of
hardware the increasing commoditization
of the open source tools well it’s it’s
benefitting them and they are able to
move even faster to deliver these new
stem cell therapies to all of us who
need them and I’m really happy to say my
wife she was determined to be a good
candidate for one such therapy just last
month we flew out to Arizona so she
could receive that therapy now it’s too
early to know
if this is the miracle we’ve been
looking for in our own big multi
objective optimization problem that
we’re trying to solve but I’m hopeful
I’m optimistic that it’s going to make
her and others like her stronger faster
and harder to kill than ever before so
whether it’s advancing the art of what’s
possible in medicine whether it’s
preparing for the zombie apocalypse
whether it’s just a fresh perspective
for life I hope you see that the
paradigm of a multi objective
optimization problem is a useful one
because of the way it helps you to frame
problems because of the way you can
quantify them and then begin to apply
intelligent machines and artificial
intelligence to get the leverage that
leads to breakthroughs so that you me
and all of us can become stronger faster
and harder to kill thank you
[Applause]
you
[Applause]
Please follow and like us: