cogito I am the okay I’ll be the greater
Japan I am the CEO and co-founder for
cog mystics so this is my third startup
company and I introduced the five you
know for what artificial intelligences
where it is added and I’ll also walk you
three different case studies so what
we’ve been working on practical
applications of AI so I saw this you
know I kind of might have heard about it
absolutel in fact we have been around
for a while so there was an analogy that
really captured me so you need a growth
Chinese bamboos they put the seed in
five years nothing happens you don’t see
anything and six weeks that really
shoots up and it’s very quickly 90 feet
tall so artificial intelligence also the
same way so if you’re in the prime time
now where you seeing uber cars driving
around us by themselves obviously on
powered by AI
but this also started a long time ago it
started in the 1950s with Alan Turing he
uses anybody wants to be equal to
computer science
but me a geek curing was fantastic from
that perspective there’s a conference at
Dartmouth that started at all but all
that stuff is great that was a lot of
research done and now is when we really
reaping the benefits now is whether
bamboo is shooting out to be 90 feet
tall so I’m going to talk about
contemporary practical applications of
artificial intelligence we’re in an age
where computers have learned to really
mimic human beings pretty well they can
see they can look they can watch videos
for us see what they are seeing is part
of the videos they can understand now
you can talk to computers I’m sure all
of you have done that with Siri
right or other versions we’ll see I wish
CD was a little bit smarter but that’s
okay they Apple will eventually make its
powder and make it better and so they’re
working on that stuff but now computers
really have the ability to understand
what you’re saying
interpret it making decisions based on
it and the thirds are fast practice
computers can act we obviously have
robots doing work you have cars driving
themselves so that’s kind of where we
are going quickly talking a little bit
more about sensing right now we’re at a
point with computers in detect images if
you put an image in app somebody elseís
a cat or a dog or a person skiing
downhill or for some scuba diving
the computers as well as human beings
are in their ability to detect that same
thing with word error rate as I’m
speaking computers can detect what I’m
saying
translators voice file into a text file
and the accuracy is just like anybody
else out here listening and sort of
comprehending that so the word error
rates have really really gone down
another thing I know this is a big slide
but this was one of my co-founders and
I’ll talk about him is dr. Eric neighbor
wreck at the school of computer science
he was very instrumental in IBM building
Watson that won the Jeopardy challenge
that’s some of the framework Watson
users to do deep QA don’t worry I’m not
going to go through each one of those
blocks but the whole idea is you can ask
the computer a question right so you
know what are the three made I’m going
to talk about India let’s see how many
people have okay so what are the three
major rivers in India so how we all can
our brains start working in that
direction and we sort of think about
that or what are the main rivers in
Pittsburgh right we quickly change it
the sentences are very similar but we
synthesize based on certain key words
and then same thing they we sort of what
what they did was they were able to
teach the computer to be able to
understand these questions in real times
in 3040 seconds even less and come up
with an answer
so that’s surf through computers now
have the ability to do that as well and
of course living in Pittsburgh we are
pioneers we see these cars around us so
this is the first city where ruber
decided to test their self-driving cars
so now obviously think of it you know
driving is a pretty mundane task we
people have been good for a while but
now computers have the ability they can
stop signs those a hard thing but
traffic lights Lane changing other
things they’ve been able to sort of
master that effect obviously safety is a
big concern there but computers can do
that now and my favorite is computers
have sort of also learned how to play
games recently I don’t know if you to
many of you know this now but there was
a poker challenge here in the casino and
the computers a queue the computer
actually won the challenge they did
better than any human being did as part
of that challenge and of course this all
started in 60s and 70s where we strained
two bureaus to play checkers then
obviously chess deep blue and then
Watson winning the Jeopardy challenge
and now poker so overarching theme so if
you put all these pieces together right
well now you have computers that can see
things sense things think and think like
human beings analyze a lot of data you
can solve a lot of really
really cool applications so that’s what
we’re doing at Cognos sticks that’s what
fascinates me – the field of artificial
intelligence so I’m going to be talking
about three different applications first
one’s a little technical it’s cognitive
clustering second one is more fun in the
retail space for cognitive offers and
the third one is interactive agents
that’s my favorite one where you will
soon be talking to machines to come up
with sandwiches or coffees or other
things and they’ll be serving as and
when you need understanding you knowing
your preferences so those are the three
journeys I’m going to be talking about a
little bit of background I think I’ve
covered most of it most of the research
that our company is based out of comes
from Carnegie Mellon school of computer
science so this is my I don’t know if
this is my third startup in fact Eric
Nyberg who’s one of my other co-founders
was also involved early on with Lycos I
had a company called online choice
another company called Intel eons which
we sold our technology to Google in fact
both Eric and I back in the early 90s
used to work for a company another CMU
spin-off called Carnegie group and it
that in fact also was doing some AI work
as well I was very happy doing some Lisp
programming and C++ programming but then
obviously the technologies have changed
since then and working on Department of
Defense projects so one big things are
of which which are a lot of what we do
really really well is in the area of
deep QA that’s what I have talked about
sort of here
now you’ve trained computers to be able
to understand questions interpret the
questions find the right answers ranked
the right answers of five different
options how do I rank those options and
then present the right answer to you so
before we go into the three journeys a
lot of people ask me this sort of you
know can a I help my business what is
the key thing
what is it that makes a given problem a
good problem to solve for computers or
artificial intelligence so first thing
is computers actually are not good as
unless human beings or what we call
subject matter experts train them so you
really do need people to train the
computers so the science fiction thing
that people talk about computers taking
on the world and doing things on their
own that’s really far way out there I
don’t even think it’s going to happen
because at the end of the day these are
program is programs that are running
whether even its neural networks or
other things that are self learning
people have to be behind it to train
that another key component for AI to do
and do well is lots of data right at the
end of the day if you don’t have good
clean data it’s hard for computer
scientists to train the models to see if
the data looks like it’s like teaching
somebody math you got to show should
tell them what’s two plus five what’s
five plus eight and eventually they get
the concepts they’re getting addition
they get multiplication same thing you
show them a lot of data and say hey if a
form looks like this it’s probably a
high-risk form if a form looks like this
it’s a low-res swap so and so and so
forth so for insurance purposes and
claims processing you give it a lot of
data computer eventually gets it and
then it is able to sort of make
decisions on its own more than most
important things that we face when we
talk to companies and what one thing we
want to tell them is very important for
the executive team and the people to buy
in we’re data science or cognitive
computing can really have an impact
because if that’s not there the
project’s not going to be successful
from that perspective so the three these
are the three ones I’m going to talk
about cognitive clustering cognitive
offers and cognitive information agents
so one of the biggest challenge which
Eric was very excited about came
actually from a company really close to
here called Tye versa
how many people here know about the deep
dark web
this was fantastic sir so as big as the
web is the deep broad web is like 8/10
50 times bigger than that so Towers has
a company here and what they do is they
collect a lot of data on a daily basis
about 22 million files that they
download from peer-to-peer networks
BitTorrent all sorts of networks they
have figured out to collect all this
large amount of data
now imagine if I’m a company and I get
22 million files a day this could be
Hillary Clinton’s emails this could be
people’s tax records this could be
travel itineraries lot of it is audio
files video files it is very hard to
sift through that data and make sense of
it so what we were able to do with this
with that massive amount of data right a
clustering algorithm basically that
looks at a file looks at various regular
expressions
what is part of that file and
automatically organizes it hey here are
all lots of tax forms so we see 1040 and
tax returns and Social Security numbers
I’m going to begin to cluster them
together or personal financial
information
so we’ve now been come to a point where
the algorithm automatically takes these
clusters all these files together
organizes the data shows it works in
conjunction with human beings and they
can either decompose these clusters
further or they can combine certain
clusters together if they need to so
this way tie versa which is right here a
local company can make sense of this
massive amount of data and really use
artificial intelligence to figure out
what we are downloading on a daily basis
and what we can do with that information
I’m going to move on and before we move
on I’m going to actually play a little
video for you guys
[Music]
[Music]
and
[Music]
thank you so this is an AI powered app
that we built for mr. tire Munroe
Bufferin brakes so what is really cool
about this app in fact in the next
iteration it gets information from the
car so there’s an onboard diagnostic
computer obd device you put plug the
device into the car it reads the sensor
so if you if one of your cylinders is
misfiring I know you ain’t talked about
cars so that’s perfect you guys are all
refreshed with that stuff now so so as
soon as one of the spark plugs is
misfiring it actually generates an error
code we have the obd device embedded in
the car that reads these error codes and
then says based on this you would want
to sort of you know schedule a service
appointment now we are taking mileage
information we are taking service
records all different past transactions
then using that data to send you timely
reminders hey this would be a good time
to get new brakes or negative tires it’s
all being powered by artificial
intelligence a quick sort of you know
big thing about the system is again like
I was talking about earlier is looking
at getting all sorts of data so
transaction data clickstream data if
you’re looking going online if you power
the call center if you have send emails
any third party data obd data so what
the algorithms do what artificial
intelligence really enables us to do is
synthesize all this information boil it
down to what is the right content
promotion or offer that would make sense
to you at this time presented to you in
the right format to basically drive more
revenue and more customer loyalty form
at all so the results have been pretty
good on an average I don’t know if you a
lot of you new deserve but people are
not too loyal
where they get their oil change they get
a coupon from Jiffy Lube or other places
so they used to get a customer once
every two years so you see a customer
the next time you customer see the same
car it could be two years later now
we’re seeing the same customer coming
every six months as and when they need
this information so it’s been phenomenal
from their perspective I’m actually
going to skip the Smart Force one but
this is a so this is the last one I
wanted to talk about this is an example
the project we were working on is
actually confidential they didn’t want
to use the information but imagine this
what you will be able to do and what AI
will allow you to do in the future is
the following you get in front of a
kiosk or a computer or your smartphone
it recognizes your face right it knows
I’m Sanjay it knows it’s beautiful day
in Pittsburgh’s it’s 70 degree outside
it knows my preferences what I’ve
ordered in the past whether it’s from
sheets or subway or Starbucks for that
matter right and it comes up with the
right recommendation for me also what is
really cool about these application is
versus having to go through menu than
options I can actually talk to the
computer it understands me it will offer
me hey Sanjay do you want your
cappuccino says no I’m in the mood for
an America americano today or I’m in the
mood for a turkey sub and I make those
selections it understands me it gives me
options and then I’m a very happy
customer as a result of this so this is
really you know taking all the pieces
all the things that I can do putting it
all together and then providing
delightful services to customers for
various organizations so that’s another
one that we are pretty proud of how are
we doing on time okay so if you surf you
look at it what does it take to make
applications like these how does the how
do
these applications work so what is the
magic behind it this is where we talk
about so like I talked about early on
what is very critical in artificial
intelligence is good data that’s where
we ingest data like I talked about the
preferences like what do people like to
drink what do they like to have so you
ingest a lot of these data you create
objects based on those data then this is
where the humans are very important the
subject matter experts they begin there
and tag the data they annotate the data
enrich the data make it cleaner make it
better made make it index and then we
build models to match the data like
regression is a good example of that
simple but a good example of that and
based on those models you can begin to
make predictions so once you have
received what we call Hydra level of
precision and recall from a machine
learning perspective that’s when you let
these models out there and let them
interact with people so people interact
with them and so what the models are
doing at the background are looking at
the right options retrieving these
objects ranking the object and saying at
this point
John would be most likely who would most
likely take this offer and run with that
offer so that drags the object and sort
of you know we basically are able to
present the right offer at the right
time to the right customer or make right
decisions based on what the data and the
models are telling us so this is a sort
of either as a pretty basic but really
good framework that is behind most
artificial intelligence systems okay so
in conclusion artificial intelligence is
really here to stay
it is going to have a huge impact on us
we actually deal with it on a day-to-day
basis Google uses it a lot too so
was parallel to serve various ads to us
Amazon uses it a lot to do predictive
analytics and collaborative filtering
people which have who have read these
books are also interested in this or
have bought this would also be
interested in this we’re going to have a
lot more applications like the one I
talked about where there’s going to be
smart kiosks topical BOTS where if you
go to your healthcare provider and
you’re waiting out there you’d be able
to get much more information based on a
horse’s sort of random searches on the
web which people do a lot which
sometimes can be dangerous much more
topical BOTS which will interact with
you know you know your history and deal
with it so that surface where we are
going and you know what we like to do at
cognos Texas to enable a lot of these
applications and really make the world a
little better place by democratizing yes
so thank you
you
Japan I am the CEO and co-founder for
cog mystics so this is my third startup
company and I introduced the five you
know for what artificial intelligences
where it is added and I’ll also walk you
three different case studies so what
we’ve been working on practical
applications of AI so I saw this you
know I kind of might have heard about it
absolutel in fact we have been around
for a while so there was an analogy that
really captured me so you need a growth
Chinese bamboos they put the seed in
five years nothing happens you don’t see
anything and six weeks that really
shoots up and it’s very quickly 90 feet
tall so artificial intelligence also the
same way so if you’re in the prime time
now where you seeing uber cars driving
around us by themselves obviously on
powered by AI
but this also started a long time ago it
started in the 1950s with Alan Turing he
uses anybody wants to be equal to
computer science
but me a geek curing was fantastic from
that perspective there’s a conference at
Dartmouth that started at all but all
that stuff is great that was a lot of
research done and now is when we really
reaping the benefits now is whether
bamboo is shooting out to be 90 feet
tall so I’m going to talk about
contemporary practical applications of
artificial intelligence we’re in an age
where computers have learned to really
mimic human beings pretty well they can
see they can look they can watch videos
for us see what they are seeing is part
of the videos they can understand now
you can talk to computers I’m sure all
of you have done that with Siri
right or other versions we’ll see I wish
CD was a little bit smarter but that’s
okay they Apple will eventually make its
powder and make it better and so they’re
working on that stuff but now computers
really have the ability to understand
what you’re saying
interpret it making decisions based on
it and the thirds are fast practice
computers can act we obviously have
robots doing work you have cars driving
themselves so that’s kind of where we
are going quickly talking a little bit
more about sensing right now we’re at a
point with computers in detect images if
you put an image in app somebody elseís
a cat or a dog or a person skiing
downhill or for some scuba diving
the computers as well as human beings
are in their ability to detect that same
thing with word error rate as I’m
speaking computers can detect what I’m
saying
translators voice file into a text file
and the accuracy is just like anybody
else out here listening and sort of
comprehending that so the word error
rates have really really gone down
another thing I know this is a big slide
but this was one of my co-founders and
I’ll talk about him is dr. Eric neighbor
wreck at the school of computer science
he was very instrumental in IBM building
Watson that won the Jeopardy challenge
that’s some of the framework Watson
users to do deep QA don’t worry I’m not
going to go through each one of those
blocks but the whole idea is you can ask
the computer a question right so you
know what are the three made I’m going
to talk about India let’s see how many
people have okay so what are the three
major rivers in India so how we all can
our brains start working in that
direction and we sort of think about
that or what are the main rivers in
Pittsburgh right we quickly change it
the sentences are very similar but we
synthesize based on certain key words
and then same thing they we sort of what
what they did was they were able to
teach the computer to be able to
understand these questions in real times
in 3040 seconds even less and come up
with an answer
so that’s surf through computers now
have the ability to do that as well and
of course living in Pittsburgh we are
pioneers we see these cars around us so
this is the first city where ruber
decided to test their self-driving cars
so now obviously think of it you know
driving is a pretty mundane task we
people have been good for a while but
now computers have the ability they can
stop signs those a hard thing but
traffic lights Lane changing other
things they’ve been able to sort of
master that effect obviously safety is a
big concern there but computers can do
that now and my favorite is computers
have sort of also learned how to play
games recently I don’t know if you to
many of you know this now but there was
a poker challenge here in the casino and
the computers a queue the computer
actually won the challenge they did
better than any human being did as part
of that challenge and of course this all
started in 60s and 70s where we strained
two bureaus to play checkers then
obviously chess deep blue and then
Watson winning the Jeopardy challenge
and now poker so overarching theme so if
you put all these pieces together right
well now you have computers that can see
things sense things think and think like
human beings analyze a lot of data you
can solve a lot of really
really cool applications so that’s what
we’re doing at Cognos sticks that’s what
fascinates me – the field of artificial
intelligence so I’m going to be talking
about three different applications first
one’s a little technical it’s cognitive
clustering second one is more fun in the
retail space for cognitive offers and
the third one is interactive agents
that’s my favorite one where you will
soon be talking to machines to come up
with sandwiches or coffees or other
things and they’ll be serving as and
when you need understanding you knowing
your preferences so those are the three
journeys I’m going to be talking about a
little bit of background I think I’ve
covered most of it most of the research
that our company is based out of comes
from Carnegie Mellon school of computer
science so this is my I don’t know if
this is my third startup in fact Eric
Nyberg who’s one of my other co-founders
was also involved early on with Lycos I
had a company called online choice
another company called Intel eons which
we sold our technology to Google in fact
both Eric and I back in the early 90s
used to work for a company another CMU
spin-off called Carnegie group and it
that in fact also was doing some AI work
as well I was very happy doing some Lisp
programming and C++ programming but then
obviously the technologies have changed
since then and working on Department of
Defense projects so one big things are
of which which are a lot of what we do
really really well is in the area of
deep QA that’s what I have talked about
sort of here
now you’ve trained computers to be able
to understand questions interpret the
questions find the right answers ranked
the right answers of five different
options how do I rank those options and
then present the right answer to you so
before we go into the three journeys a
lot of people ask me this sort of you
know can a I help my business what is
the key thing
what is it that makes a given problem a
good problem to solve for computers or
artificial intelligence so first thing
is computers actually are not good as
unless human beings or what we call
subject matter experts train them so you
really do need people to train the
computers so the science fiction thing
that people talk about computers taking
on the world and doing things on their
own that’s really far way out there I
don’t even think it’s going to happen
because at the end of the day these are
program is programs that are running
whether even its neural networks or
other things that are self learning
people have to be behind it to train
that another key component for AI to do
and do well is lots of data right at the
end of the day if you don’t have good
clean data it’s hard for computer
scientists to train the models to see if
the data looks like it’s like teaching
somebody math you got to show should
tell them what’s two plus five what’s
five plus eight and eventually they get
the concepts they’re getting addition
they get multiplication same thing you
show them a lot of data and say hey if a
form looks like this it’s probably a
high-risk form if a form looks like this
it’s a low-res swap so and so and so
forth so for insurance purposes and
claims processing you give it a lot of
data computer eventually gets it and
then it is able to sort of make
decisions on its own more than most
important things that we face when we
talk to companies and what one thing we
want to tell them is very important for
the executive team and the people to buy
in we’re data science or cognitive
computing can really have an impact
because if that’s not there the
project’s not going to be successful
from that perspective so the three these
are the three ones I’m going to talk
about cognitive clustering cognitive
offers and cognitive information agents
so one of the biggest challenge which
Eric was very excited about came
actually from a company really close to
here called Tye versa
how many people here know about the deep
dark web
this was fantastic sir so as big as the
web is the deep broad web is like 8/10
50 times bigger than that so Towers has
a company here and what they do is they
collect a lot of data on a daily basis
about 22 million files that they
download from peer-to-peer networks
BitTorrent all sorts of networks they
have figured out to collect all this
large amount of data
now imagine if I’m a company and I get
22 million files a day this could be
Hillary Clinton’s emails this could be
people’s tax records this could be
travel itineraries lot of it is audio
files video files it is very hard to
sift through that data and make sense of
it so what we were able to do with this
with that massive amount of data right a
clustering algorithm basically that
looks at a file looks at various regular
expressions
what is part of that file and
automatically organizes it hey here are
all lots of tax forms so we see 1040 and
tax returns and Social Security numbers
I’m going to begin to cluster them
together or personal financial
information
so we’ve now been come to a point where
the algorithm automatically takes these
clusters all these files together
organizes the data shows it works in
conjunction with human beings and they
can either decompose these clusters
further or they can combine certain
clusters together if they need to so
this way tie versa which is right here a
local company can make sense of this
massive amount of data and really use
artificial intelligence to figure out
what we are downloading on a daily basis
and what we can do with that information
I’m going to move on and before we move
on I’m going to actually play a little
video for you guys
[Music]
[Music]
and
[Music]
thank you so this is an AI powered app
that we built for mr. tire Munroe
Bufferin brakes so what is really cool
about this app in fact in the next
iteration it gets information from the
car so there’s an onboard diagnostic
computer obd device you put plug the
device into the car it reads the sensor
so if you if one of your cylinders is
misfiring I know you ain’t talked about
cars so that’s perfect you guys are all
refreshed with that stuff now so so as
soon as one of the spark plugs is
misfiring it actually generates an error
code we have the obd device embedded in
the car that reads these error codes and
then says based on this you would want
to sort of you know schedule a service
appointment now we are taking mileage
information we are taking service
records all different past transactions
then using that data to send you timely
reminders hey this would be a good time
to get new brakes or negative tires it’s
all being powered by artificial
intelligence a quick sort of you know
big thing about the system is again like
I was talking about earlier is looking
at getting all sorts of data so
transaction data clickstream data if
you’re looking going online if you power
the call center if you have send emails
any third party data obd data so what
the algorithms do what artificial
intelligence really enables us to do is
synthesize all this information boil it
down to what is the right content
promotion or offer that would make sense
to you at this time presented to you in
the right format to basically drive more
revenue and more customer loyalty form
at all so the results have been pretty
good on an average I don’t know if you a
lot of you new deserve but people are
not too loyal
where they get their oil change they get
a coupon from Jiffy Lube or other places
so they used to get a customer once
every two years so you see a customer
the next time you customer see the same
car it could be two years later now
we’re seeing the same customer coming
every six months as and when they need
this information so it’s been phenomenal
from their perspective I’m actually
going to skip the Smart Force one but
this is a so this is the last one I
wanted to talk about this is an example
the project we were working on is
actually confidential they didn’t want
to use the information but imagine this
what you will be able to do and what AI
will allow you to do in the future is
the following you get in front of a
kiosk or a computer or your smartphone
it recognizes your face right it knows
I’m Sanjay it knows it’s beautiful day
in Pittsburgh’s it’s 70 degree outside
it knows my preferences what I’ve
ordered in the past whether it’s from
sheets or subway or Starbucks for that
matter right and it comes up with the
right recommendation for me also what is
really cool about these application is
versus having to go through menu than
options I can actually talk to the
computer it understands me it will offer
me hey Sanjay do you want your
cappuccino says no I’m in the mood for
an America americano today or I’m in the
mood for a turkey sub and I make those
selections it understands me it gives me
options and then I’m a very happy
customer as a result of this so this is
really you know taking all the pieces
all the things that I can do putting it
all together and then providing
delightful services to customers for
various organizations so that’s another
one that we are pretty proud of how are
we doing on time okay so if you surf you
look at it what does it take to make
applications like these how does the how
do
these applications work so what is the
magic behind it this is where we talk
about so like I talked about early on
what is very critical in artificial
intelligence is good data that’s where
we ingest data like I talked about the
preferences like what do people like to
drink what do they like to have so you
ingest a lot of these data you create
objects based on those data then this is
where the humans are very important the
subject matter experts they begin there
and tag the data they annotate the data
enrich the data make it cleaner make it
better made make it index and then we
build models to match the data like
regression is a good example of that
simple but a good example of that and
based on those models you can begin to
make predictions so once you have
received what we call Hydra level of
precision and recall from a machine
learning perspective that’s when you let
these models out there and let them
interact with people so people interact
with them and so what the models are
doing at the background are looking at
the right options retrieving these
objects ranking the object and saying at
this point
John would be most likely who would most
likely take this offer and run with that
offer so that drags the object and sort
of you know we basically are able to
present the right offer at the right
time to the right customer or make right
decisions based on what the data and the
models are telling us so this is a sort
of either as a pretty basic but really
good framework that is behind most
artificial intelligence systems okay so
in conclusion artificial intelligence is
really here to stay
it is going to have a huge impact on us
we actually deal with it on a day-to-day
basis Google uses it a lot too so
was parallel to serve various ads to us
Amazon uses it a lot to do predictive
analytics and collaborative filtering
people which have who have read these
books are also interested in this or
have bought this would also be
interested in this we’re going to have a
lot more applications like the one I
talked about where there’s going to be
smart kiosks topical BOTS where if you
go to your healthcare provider and
you’re waiting out there you’d be able
to get much more information based on a
horse’s sort of random searches on the
web which people do a lot which
sometimes can be dangerous much more
topical BOTS which will interact with
you know you know your history and deal
with it so that surface where we are
going and you know what we like to do at
cognos Texas to enable a lot of these
applications and really make the world a
little better place by democratizing yes
so thank you
you
Please follow and like us: