AlfdhA0GD74

[Music]
in fact this here's Tyrion joining us
our first guests on the a hi cafe series
I actually created a video with him and
one of my other mentors Tim Duncan
some time ago about artificial
intelligence even Watson that was really
really interesting but luckily he's back
in Toronto and rabl to make another
video about something really fundamental
with artificial intelligence which is
the fact that pattern detection that our
visual intelligence is not artificial
intelligence oh yes absolutely yes my
name is Jerry Dammers here with our
ecosystem and 10 million I've been
working for a little while together we
were having fun actually it was a
pattern detection and in fact and I
without pattern detection it's just that
I got exactly and we're gonna change
things down exactly what we're going to
talk about we're going to have one of
those conversations that doesn't involve
keyboard it's kind of human and we're
talking about neural networks patterning
detection and what it means and what not
and how useful is it but but
fundamentally you can understand that it
is the core isn't it exactly well now
what that pattern detection there's no
point to a guy on first place that's
right and it even goes deeper than that
and and this is what's gonna be fun
about this this new series cafe right
yes is that we can talk about it and we
Caffe belong in this case week we're
gonna turn this guy and and allow the
analog device to get all right like the
other analog device which is our brain
and so fundamental to pattern detection
is our ability to as humans or any any
cognitive organism for that matter to
detect patterns in patterns create
memories right exactly and then when the
pattern changes what happens well you
say what's happening exactly and that's
the biggest thing so I've got a pattern
that changes what's happening well isn't
that what people are trying to really
find out isn't that what makes us aware
it's a VAR environment exactly
so isn't it fun to think that actually
today we have technology was AI that is
based on that fundamental belief or that
fundamental principle I should say then
we have to believe in it right but it's
a fundamental principle yeah that in
fact a pattern when we absorb when we
move through life or when a system moves
through data right we observe pattern
and the beauty was there's no network
nai is that as we move through the data
or as we move through life we remember
right and we observe that there is that
there is a reoccurring pattern and
what's interesting about it is that if
it's recurring it's soothing and it
doesn't matter who you are and then you
but what happens when the pattern always
occurs it's always the same time in
order it's boring exactly what you
expected it's what you know exactly
right so so we can we can go talk about
this about
are you of it Howard I pleasant
technology if you want you know bring
the analogy around an or network around
the wonderful work that you do or some
of the project to work out the cognitive
projects but besides that project alone
you know yeah how do you say okay great
I'm doing who's that
how do you bring it in to a point where
people understand and how can you
express that's a no-no really what we're
doing here is pattern detection exactly
all right so one of the projects that of
course we're actually collaborating on
is called the cognitive story may have
heard of it before it's basically a
project where I first of all I mean
generally we're trying to use AI to
enhance people's lives but our first
goal is to actually help out a
quadriplegic girl so as I mentioned
she's spotted a jerk unfortunately
suffers from Rhett syndrome meaning she
cannot communicate whatsoever if she's
you know thirsty or hungry or wants go
to the washroom
her mom has to make good informed guess
as to what Boo is trying to lie so there
you go
so why does she can she make a guess
because she's able to recognize those
paths there you go exactly that's it and
so what works like to do that is see how
can we unlock more data to gather
pirates fun
well booze mom is looking at while she
was trying to understand what who's
talking about if she said okay how is
boom moving what's the time of the day
when it's the context so it's a
situation like she's looking at all
these different variables which might
indicate towards what we was trying to
mention and these are the different
types of patterns that she's working at
recognizing but there are lots of more
patterns that she is a human cannot
recognize because us humans are
sometimes bias towards symbol repeatable
patterns correct we can't look through
vast amounts of
flex noise and noise data and say hey
this is a pattern correct but artificial
neural networks okay and so that's why
we're actually trying to use
electroencephalogram handsets and deep
learning algorithms to actually
recognize patterns in lose EEG brain
waves when she's trying to answer a
question and see if we can convert those
brain waves to actual intent or what she
wants to do and that's a perfect example
of how we're actually detecting quite
literally patterns in her brain waves
even though it's so noisy these neural
networks have relatively easy time
finding patterns in such a noisy and
fast amounts of data because they're
able to see repeatable yet changing
patterns throughout all the data it's
exactly right and what's fascinating but
when you bring in in that context they
say okay so the technology is reading
the signal that mean us as human we're
not capable of reading exactly so you're
complimenting and what I love about what
you said is that you're complimenting
what we call me if you won't be the
person who the intimate interpreter is
what we call bird on research right so
the mom is the intimate interpreter you
say today if it's daytime depending on
the weather
bunch of vectors exact hole that she's
observed she's able to correlate them
and says oh observing your child it's
based on the patterns of knowing how she
behaves if she Krantz is she looks a
certain way she knows that what's
beautiful but but the project is that
now we're saying patterns we don't see
the computer can see inside and then it
brings them in and it adds a new
dimension right and then when the
pattern is different but a pattern
changes let's say in conjunction was her
wanting to eat her being you know
nothing comfortable or wanting to watch
your teeth and somebody's in the way
like it happened
yeah I was there he said she there was a
reaction that in conjunction with the
data gives us the ability to take that
cognitive component that we all take for
granted and put it in assistant that can
assist her that's the augmentation right
there right we're limited by the amount
of data that we can take in and how we
can take in that data we humans are only
good taking in data with you know the
senses that we have listed numbers you
would never understand I mean you know
we'll try our best and visualize in your
head you would understand it but it's so
much easier for me to show you a graph
and you will immediately understand what
we're doing with something as complex as
EEG brain waves we cannot do that we
cannot have you just look at a graph of
the EEG brain waves and say okay do see
a pattern in these is the fact that
they're so complex they are literally
dependent on each other all the
individual data points to find out and
that's exactly correct and this is the
correlation that gives it sense it's
alright because I can look at a pattern
I kid I'm gonna be able to recognize it
but what I see a change if I cannot
correlate it with something else it's
meaningless and so on so you take this
in the context of business and market
fluctuation and whatnot and when we use
all these wonderful techniques to do
pattern detection we're looking at the
single streams of data
factors right and we're saying one of
the patterns so the system can learn the
pattern right and in its creat what it
creates in aggregate it creates a
collective I would say set of neural
pathways right all right and patterns
and then we're looking at this and say
ok with something changes changes what
happens next and that's what people in
business usually look for into business
intelligence and competitors for
instance a pattern like hey look
everybody and Twitter of my competitors
competitor
yeah well that's meaningful but if I
didn't understand the actual pattern how
would I have been awarded to what I
don't know exactly right and that's
where it's beautiful that's that's the
trigger
exactly so now we can do it this more
thing than just the limitation of your
physicality and my physicality in the
same room exactly right the burden would
show up over here it would be this
person to the pattern with very thick
even if you were standing over there
exactly so the system's unified that
improves decision dramatically in fact
artificial intelligence because the fact
that is running on technology has some
advantages that we humans can never have
mm-hmm good or drunk okay
this is basically a huge paddock
elimination problem if we do if we want
to do it actually but the thing is
driving me sometimes distracting person
why because the thing is our birth
family focus on so many things at once
yes our brains can't focus on 50 things
at once we can look you know in front of
us we can rearview mirror we can see
other nears we can see what speed we're
going to look at the GPS so many things
to look at so many things to cordoned
okay a pretty good job but this is our
machine is coming machines
cars so much better than us because they
can take a look at all the data points
make meaningful pattern recognition
alpha yes and then from there take
everything that it's seeing in
conjunction and then make a decision
because their full attention
exactly they have full attention
everywhere because they are capable of
making full attention every single data
point in your honor to go to your
example and somebody drives a car they
are used to seeing traffic around them
they were used to seeing people so it's
actually it becomes subconscious in a
way so their attention is only related
to the immediate decision after mating
and if something interferes exactly
right and systems are much more open to
the very largest right this is
interesting but you see this this is
where I think that the competition on
the pattern detection is very meaningful
because when people people talk about AI
who talk about natural language
processing oh no that's bad well it's
arbitrary
I mean think of it this way the world
around us yeah the world the natural
world shows patterns Laurens
demonstrated it was the strange
attractor where you know in a closed
environment that is highly dynamic we're
able to observe patterns exactly oh it's
no different this is one of the things i
postulate at all there was no difference
between the closed environment of nature
and the closed environment of human
cognitive right so the information or
the elements in that chaotic environment
of these feel patters yes right and
their computers now because we capture
so much of our firm ation is what is our
information when we
it's an intent exactly and we have an
intent in order to those because we
observe something and yeah and
communicated or we have an intent to
achieve something it's better that's why
we put it in the digital format
especially with unstructured data oh
yeah what more wonderful thing to do
then capture the intents of human
expression and analyze the pattern how
concepts evolve in the company in
collective memories these actually I
mean how cool is that and this is the
work we love to do I'm going to take all
the information you have I'm going to
slice it and correlate it and show the
patterns of your content show the story
you collect a story when you said about
the car examples it's exactly the same
because we're saying I'm not aware of
everything I'm not dismissing anything
and the thing is what I'm saying
okay we were trained by this one drivers
this is now my driving style what am i
able to do is it able to take say a
hundred drivers of its loading or maybe
200 300 drivers yes and each one before
I see how hundreds of hours of driving
data we're not telling machine ok
because the roads curving and these
would be the reason right this is how
you're gonna turn the steering wheel no
we're not able to do we're teaching it
we're giving all the data from these
drivers and saying learn how to drive
and it's implicit or lies all those
patterns it's not learning one specific
driving style look for example if you're
a human you're taught how to drive you
will learn a specific driving style with
AI we can look at all these data points
and find patterns among all of them and
say okay this is the average way to
drive the proper way to drive at least
this is how most units which are thank
you if you have any right and this is
why pattern detection is fundamental
when we talk about
because I mean one thing people look at
our special intelligence with respect to
human interface right now if you can't
fool me you know to think that a
computer can reason and I can say a lot
of this computer can reason therefore
it's artificial intelligence this is
where people are fiction their reality
is it's much wrong simple so what I'm
talking about this to me we demystified
for people
what a I really is because we're telling
this is fundamental this is a more
organic property of any company that
organism and we can do the same thing
with technology when you take it from
there I mean doesn't that simplify a
conversation in fact you could just say
artificial intelligence it's basically
just a huge pattern recognition problem
in general we're trying to give an
abstract and huge amounts of data and
see how can we find a specific patterns
that allow us to achieve this end result
there's also things like okay let's take
a look at the entire history here and
see when this similar type of pattern
has happened what usually happens after
this no longer a linear regression right
sorry it's a lot more interesting a lot
more organic exactly because you say
hello that the stock market is a human
invention so human extraction exactly
like for example with artificial
intelligence we're trying to emulate
things like empathy in humans computers
we cannot emulate things like
I should understand things like your
personality we can understand it's a
sentiment with which you're talking for
example let's just say that you're you
know tweeting about something that's
just saying that your entire company is
receiving millions or hundreds of
thousands of tweets we're filters here
whenever you say okay well this specific
keyword what sentiment attached this
keyword okay what types of catters allow
us to say okay the sentiment would this
is positive or negative or people are
angry or fearful and you're right and
this is complex any of you you've seen
this before you've seen people who do
like sentiment analysis on Twitter all
the time say hey what do they do they
simply use terms yeah then they said
this is positive this is negative this
is not this is so deterministic it has
no flexibility and you know people will
use certain language if in a positive
context when we for most of the people
that language is unacceptable in seen as
negative Zack how do you learn you know
so you have to inject disability in the
system so to learn and to give you
insight and then from there it's not
just that one time learning it is you
learn once you have base knowledge now
you're working on top of that building
more and more and more knowledge as you
go as you want inference or training
more and more as you drop okay you're
about to get into an accident learn from
that you learn that okay I shouldn't do
this one I should do this are you
telling me that companies will invest in
a
have to embrace the idea that it is good
to observe mistakes absolutely go you
guys got that this is true right you
have it's a learning system it is such a
beauty I mean think about it the
projects are on a college in health care
you know helping Diagnostics I mean it's
all pattern detection and validation by
professional if somebody asked me this
what is it we throw all of this data
into into the bucket into Watson and we
wanted to answer a completely arbitrary
question without proper training right
so that interaction to just is it it's
at home does it take for a human being
to learn to be a doctor
why would you expect a bunch of trial
and error okay when you put that into a
system you expect the system to do a
better job than a human being but number
two you gotta look and say well what is
the value of having trained assistants
spend perhaps billions of dollars
training your system exactly morning now
I can bring that knowledge into areas in
the world where thousands of
professionals can benefit from it you
tell me all valuable about this right
now
Oh cancer specialists sort of diagnosis
or you know treatment of cancer okay
they're all getting their own experience
they're all gaining their own individual
knowledge with every single diagnosis
and treatment I know where you're going
what happens we can have one single
cloud that has all the experience all
the knowledge so every time saying
someone's diagnosed in Canada
someone in Dubai can benefit from that
knowledge yes an entirely different
doctor entirely different Hospital
entirely different method but still
they're benefitting from the fact that
someone in Canada was treated by are you
telling me here that we can build
collective knowledge and benefit from it
quite literally yes there you go and
that's beautiful because as humans right
we can you talk about
right our organism our system is unified
in order to be stronger I mean build
city to be stronger we build Network and
you know it's beautiful
just like you all network just like that
all right it's funny how every time two
entities can communicate the sum of the
value is higher can you imagine I mean
the whole reality as well is that call
connected devices the fact that they're
connected give them their own governance
that only a cognitive awareness that we
don't need to comprehend as a difficult
like to simulate other cells in your
body little 10 is 10 what you're
thinking but they're participating in
the equation exactly every more official
cognitive system because it's it's but
it's negative so now we're unifying this
is what I always just said this
knowledge is converging we're gonna find
in technologies enabler exactly if you
take a look at this why are you in so
great it's because we can communicate
with each other in such a natural way
and we can actually transfer knowledge
of one person to other really quickly
but it's a common trail right that's why
the user so great
preserving knowledge for literally years
and you know gaining experience but this
meeting is not some limited mmm all
right as I mentioned going back to the
example can undo if you can't talk to
learn something in camera but it teaches
a few other doctors limit
those doctors go everything to each but
with AI that's not limited at all what
if we could take the fact that humans
can communicate with each other and
share their knowledge to the next level
what if we can make it set behind this
need to share knowledge gets just one
individual AI that does cancer treatment
diagnosis right Enix for it is the sum
of all the knowledge all the knowledge
every single honesty I mean everybody's
got to realize this is when we talk
about pattern detection looking L arche
I said defective the system can remember
oh the fact that the system can
aggregate defected the loop feedback
that the learning mechanism improves I
mean and I want to be metaphysical my
buddy but we are capturing the the in in
a very sustainable and long-term way the
essence of human cognitive and capacity
exactly for the service of humans oh
yeah you know and this is what's
fantastic so back to the beginning here
right yes parent detection fundamental
we do it organically technology enables
us to actually go act on it and
convergent when the pattern changes when
something is different if we can act on
it
[Laughter]
[Laughter]
over many years they've gathered
experience these are the stocks made to
buy sell they've learned how to
recognize patterns yes but the thing is
the only problem is it's limited to that
individual
they cannot completely distill all of
that knowledge that they have and give
it to another person true exactly so
what if we would have used AI to do
automatics in fact in the u.s. I believe
over 70% of stock trading done on us
bought in 2014 was all done by boss yeah
yeah so what if one central AI that
eventually learns all these patterns
it doesn't need to share any knowledge
or otherwise it's just one AI that
continuously learns and that's the AI
want to keep in a box in a basement
right and I throw some voltage at it and
because that's the distinction
unfortunately you know are we still
operate around a model of failures and
success to solve and necessity is a
necessity right I mean otherwise we will
be doing what we're told today but the
point is what you say you're absolutely
right if this was knowledge what I think
that think the expert in your case yeah
and then you bring the insight and you
reveal the patterns of data you reveal
the anomalies and you turn that into
something that they can use to make
better decision to give them a
competitive
now the same way that we talk about
normalizing patterns and not can you
imagine if everybody had the same tool
to market who no longer be a market oh
[Music]
yeah in the seats of purpose but the
example is absolutely right because
eventually this is what can happen
because of so this is fun so we have we
have a lot of example we could be
talking about the time he was talk
marketer talking about healthcare
talking about driving cars talking about
collective knowledge of looking at the
movement of memories in the organization
you know a thank you a great example
when you talk about as convergence its
two entities mergers and acquisition one
company has all this data all this
knowledge and can abstract it to make it
useful and find it inaccessible and then
you have another company who's got like
they just acquired it and something over
here happens that is unusual and guess
what it matches with something over here
that happened ten years ago you can drop
just like your memories yeah I can draw
from that knowledge check when I see
this it reminds me when my mom used to
make eggs five-minute eggs like this I
mean that's my was done right so my
correlation was this is most probably
different than yours so so the fact is
that uncorrelated that same thing was
Enterprise easy they acquire this
knowledge is collective and they can
bring it out and say well this is
meaningful for me right now I do a
virtual an acquisition gosh I didn't
know that you did stuff stem-cell
research 20 years ago the government
rules change
that's an emerging pattern exactly to
change well can we match that change as
a memory has a trigger to see what we
did in the past
and elevate all the entities associated
with it that's what we do every day
one can organs
it was actually the data the store in
your archives which to me you know what
it is is a graveyard of knowledge
graveyard it's silly look it was the
reality of things you know no it changes
haven't changed Diebold was technology
c'est moi maybe there's something I can
really do with say exactly its values in
there first of all cancer yes there is
humans cannot keep physically impossible
for us to read through all of that and
not just read through it but understand
all of it and apply it to make you
exactly yeah but technology it can only
store all that information but with AI
not only do we store that info we make
sense of it and use it can take all the
important stuff that we're going through
all the important stuff that we create
and then understand it and make use of
atoms interact with is very natural
exactly and that's bird can go ahead and
ask a question and it'll get done in
semesters based on the analysis it's
brilliant it is really and also another
great example which relates to this it's
actually
put a link to that in the description
about I believe with mining or password
and so what they were thinking doing is
collaborating with IBM and actually
taking most of their knowledge bases
most the knowledge sense you know
feeding it to Watson and they're placing
a good fraction of their workforce just
just like that okay mm-hmm but what they
found out was that related to the
equation the movie workers they added
war because they realize exactly what
they realize is that we don't need to
mess trick that with IRA people because
not all of our workers knowledge
knowledge an easy way hey I will take
all that knowledge and distill it down
make it so easy for us humans to access
them in fact especially the oil industry
I think you're talking about the example
and the oil rig and the deep-sea
exploration is that the one beliefs they
might be right but the beautiful part
about this and the oil industry is is
one company that used a lot of
contractors want you know depending on
the market and depending on other
resource allocations and they often lose
them and there's been a huge knowledge
attrition in this industry especially in
Louisiana Texas
you know it's a different move different
the different movement in the in the in
the price and the resources availability
and all that so these guys come in and
out you know and feel as it was a
knowledge denotes that they take off
notice so so knowledge knowledge
acquisition exploitation usage you know
in a way to remedy against now if you
Tricia is great but but here's the thing
that's fun
yes you know you used email how many
people say oh my gosh
place jobs we heard this but I'm fully
up to that in the description as well
but you know artificial intelligence is
one of the reasons that it's so hard to
develop with is because before I even
existed everyone expected exactly okay
[Laughter]
it doesn't mean that we will lose it
people we augment people we make them
more knowledgeable so they can make
better decisions that and it's true I
mean automation in the has taken a lot
of jobs with people I don't you know
what I think really is fine what it's
the people today and the fear about so
so before came a decision that affected
people who weren't into factories now
the making decision around the holiday
on their job somehow they feel that oh
my gosh so it's it's I think it's it's
it's it's humorous in some way but
fundamental back to what we're saying
yes and I think we're about to wrap up
this wonderful conversation this skeptic
I think we're about there
and I think we have about about just
this particular time so I think it's
important just for people to really
understand it is wonderful when an
organic model like pattern detection
like cognitive science is actually the
foundation of a technological
representation in abstraction and that
makes it intimate between us and the
technologies is the technology can start
to do things we expect ourselves to be
able to Jennifer what a wonderful
connection that represents oh yeah so
think of a I'm just but we were talking
about just stuffing a I'm without
pattern detection is not a I and embrace
it because what I'm truly believe that
it is a true transposition of something
that is innate in every one of us and it
only makes sense so that was the first
episode this was fun
thank you very much for joining in today
I really hope all of you enjoyed this
this new series we're doing some more of
these very soon as well officer Ian
thank you very much everyone for joining
it today hope you enjoy if you have any
questions suggestions or feedback please
leave it down in the comment section
below you can email it to me attend
original dark comm or you can tweet to
me I drew Manning apart from that Terry
how would you contact you Oh beacon
contact me directly at Darwin ecosystem
calm or cheery and Darwin ecosystem that
come we're happy to continue the
conversation all right thank you ten
wonderful fun conversation no computer
no keyboard no keyboard to me
keywords a lot of time moving all right
come on