OPEN TRANSCRIPT

00:00
thanks for attending the disruption now
00:02
virtual summit
00:03
2020
00:06
this session is brought to you by fuse
00:09
by cardinal health
00:11
your panelists are will hayes
00:14
and doug mccullough this panel will be
00:17
moderated by lauren burke
00:21
and now the host of the disruption now
00:24
virtual summit rob richardson
00:27
welcome to the disruption now virtual
00:29
summit we are glad to have you here as
00:31
we talk about the human connection to
00:32
technology
00:33
hope everybody's doing well we have
00:35
people from all across the country so we
00:36
got my man
00:38
will hayes who's uh joining us from the
00:40
bay from
00:41
lucidworks how you doing man doing
00:42
really good really happy to be here
00:44
thanks for having me
00:45
uh thanks for being on doug mccullough
00:47
our man who's we've done a lot of events
00:48
together now
00:49
i think we're kind of becoming uh we're
00:51
kind of becoming pros at this thing how
00:52
you doing doug
00:55
doc could be here well there we go you
00:57
good to see everybody
00:59
yeah doug is the cio the city of dublin
01:03
uh one of the leaders of black tech 614
01:06
we appreciate all his partnership and
01:08
everything he does and last but
01:10
certainly not least
01:11
and the moderator for this event lauren
01:13
burke with
01:14
women in analytics how you doing good
01:16
how are you
01:18
hey living the dream living the i guess
01:21
the virtual dream because we can't do
01:22
this in person
01:23
uh but we're gonna make the most out of
01:25
this moment and uh
01:26
you know this time has allowed us to
01:28
connect with a lot with a lot more
01:29
people sometimes so
01:31
we're making the most of it and we're
01:33
learning as much as we can so
01:34
lauren i'm going to turn it over to you
01:36
and you get to take us an mcd event
01:39
all right thank you rob all right so
01:42
since this is the human connection to
01:44
technology
01:45
well this is something you're very close
01:46
to and you work close with can you
01:48
you start off by telling us the
01:49
difference between artificial
01:51
intelligence
01:52
and augmented intelligence yeah
01:54
absolutely i think it's a really good
01:55
concept for for people to understand
01:57
as we think about where we have you know
01:59
sort of good use of technology of ai of
02:03
machine learning to either enrich our
02:04
lives enrich our experiences or
02:06
where things can go a little sideways
02:08
and i think a perfect example is if you
02:10
think about what machine learning
02:11
actually is is it's a machine and one of
02:13
my favorite examples was a tweet that a
02:15
woman put out one time to amazon saying
02:17
hey amazon
02:18
um you know i i bought a toilet seat
02:20
because i replied i needed to replace a
02:21
toilet seat
02:22
um i'm not like an addict or anything
02:23
why do you keep advertising and pushing
02:25
me more promotions and telling me when
02:27
there's new toilet seats and
02:28
that's an example of where just
02:29
artificial intelligence just looks at
02:31
numbers and makes decisions so
02:32
our belief and with my company lucid
02:34
works what we try to encourage is how do
02:35
you intersect
02:36
humans in the middle to make those final
02:39
decisions to use that human intellect
02:40
that human instinct that intuition
02:42
that a machine's just never going to
02:44
have think about artificial intelligence
02:46
as ways to provide
02:47
more efficient choices to humans but i
02:50
do not think that artificial
02:51
intelligence should be making choices in
02:52
our daily lives that's where this idea
02:54
of augmented intelligence comes in where
02:55
we become much more efficient
02:57
but a human is always in the loop to
02:58
ensure that we're providing that
03:00
human experience on top of things first
03:03
and foremost
03:04
yeah absolutely that's incredibly
03:06
important especially when you're working
03:08
with technologies that touch large
03:10
amounts of people
03:11
doug so as the cio of the city of dublin
03:14
how are you
03:15
integrating humans into the process of
03:17
developing your technologies that affect
03:19
the citizens
03:22
well this is interesting you know
03:26
small communities not develop ground
03:29
technologies you know we tend to specify
03:32
them we purchase them from others and we
03:34
can figure them
03:36
so you know someone like will is the
03:38
person who's developing the technology
03:40
but i do want to introduce one concept i
03:42
sit on a work group with the mid-ohio
03:45
regional planning commission that is
03:46
developing
03:47
a survey and tool kits for local
03:50
governments
03:51
and people in my situation to try to
03:53
enhance what we can do
03:55
as we begin to delve into things like
03:57
machine learning
03:58
and artificial intelligence and quoting
04:01
the
04:02
human element or verifying and keeping
04:04
the human element
04:05
into our technology development and so
04:09
uh as will kind of has alluded to
04:12
our goal is to this is the unexciting
04:16
and unsexy part of artificial
04:18
intelligence which is
04:19
the slogging forward i thought it was
04:22
all sexy
04:23
[Music]
04:25
no well it's all sexy to me but i mean
04:29
to the audience but there's a lot of
04:30
spreadsheets involved
04:32
well i mean the thing is is that you've
04:35
got to
04:35
interject and intervene
04:40
the concepts of equity and and equality
04:44
into every part of the technical uh
04:47
rollout
04:47
and that is a slog it's just it's just a
04:50
pickaxe kind of hammer and chisel
04:52
working away like fred flintstone on on
04:55
these things and
04:56
we're just one small community which is
04:58
why we try to act as a region
05:00
in central ohio to answer these
05:02
questions for
05:03
uh many communities and not just one
05:06
because if one community does a
05:07
fantastic
05:08
job and nobody knows about it then it's
05:10
not going to help these things have to
05:12
scale
05:14
yeah and i think there's some there's
05:16
like when people
05:17
depending on who you talk to even some
05:19
people are scared to death when they
05:20
think about
05:21
artificial intelligence because i really
05:24
like the way will said it because people
05:25
see artificial intel some do
05:27
see artificial intelligence and saying
05:29
like okay are we replacing
05:31
humans the ultimate goal um i mean
05:35
i would say no i mean this this uh i'm
05:37
trying to remember the book i think it
05:38
was the second machine age
05:40
and it talked about the fact that okay
05:42
we do we now have
05:44
we now have a computer that that can out
05:46
that can beat
05:48
regular a regular one person a chess
05:50
master if it's just one chess master but
05:53
if you have a team that's actually just
05:55
knows how to use the computer
05:57
they were able to be a super computer
06:00
and chess masters because they knew how
06:01
to use the
06:03
they knew how to use computers in the
06:04
right way while also interjecting
06:06
human potential and human thinking so i
06:09
do think that's how we have to think
06:11
about it and that's how the process has
06:12
to be
06:13
and it shouldn't just scare us because i
06:14
tell look technology is coming
06:16
one way or another we got to figure out
06:18
how we use it
06:20
and make sure it has the most impact for
06:22
all involved absolutely
06:23
absolutely if i could tag in on that
06:25
because i'm really impressed with the
06:27
way that will
06:28
is talking about on minted intelligence
06:30
i think
06:31
of this in that way like computers can
06:35
calculate
06:36
in a way that's far and away more than
06:37
any human is going to be it's just raw
06:39
brute strength
06:40
but making decisions and having judgment
06:43
uh
06:43
and nuance is not that great was
06:46
a strength for them so when we think
06:48
about
06:50
i'm going to make decisions but when i
06:52
need to do
06:53
massive raw computation i'm going to
06:55
call on a tool
06:56
and that's what we need to think of
06:57
artificial intelligence as as just a
06:59
tool
07:00
and then we can do a lot of it and it's
07:02
a little bit less scary i think
07:03
and this is my point about presenting
07:05
choices not making decisions right
07:07
because a lot of
07:08
you know and i come from a data world
07:10
i'm a little biased but you know a lot
07:11
of what we're trying to accomplish with
07:13
with artificial intelligence is how do
07:15
we make sense of just massive amounts of
07:17
information and data and present it in a
07:18
way that's
07:19
you know readily available for rob to
07:21
take an action with
07:22
and so when you start thinking about
07:24
sort of all of the different
07:25
permutations and the dimensions that you
07:27
have to kind of start working through in
07:29
order to make those decisions obviously
07:30
this is where the machine algorithm
07:32
approach
07:32
makes the most sense but at the end of
07:34
the day it's about how do i distill that
07:36
massive amount of information down to
07:37
something that's consumable
07:39
so then a human being can actually
07:40
interject that choice because so the
07:42
point that you made doug i mean
07:43
the intuition the intellect the instinct
07:45
the emotion that we
07:46
we absorb in part of that interactive
07:48
experience is what makes us human
07:51
and and we're never going to achieve
07:52
that at least today with with technology
07:55
yeah absolutely it's incredibly result
07:58
driven
07:59
and a lot of times you really should be
08:01
considering who
08:02
is going to be using that at the end who
08:04
is your audience that you really want to
08:06
be
08:07
able to make that decision off of so
08:09
something we
08:10
hear a lot of in like algorithms around
08:12
machine learning and artificial
08:14
intelligence
08:15
is that the expression like garbage in
08:17
means garbage out
08:18
so when you're dealing with poor limited
08:20
or biased data sets
08:22
so like in some cases those bad data
08:24
sets can
08:25
result in a faulty product so like
08:28
for due to limited training data one
08:30
example is voice technologies often
08:32
misinterpret or have trouble
08:34
understanding like particular accents or
08:36
women's voices
08:38
and another that i know you've talked
08:39
about before will is chat bots and
08:41
recommendation systems so without
08:43
human intervention in those those can
08:45
have some disastrous results
08:46
and this goes back to how these things
08:48
are trained i mean rob and i had a fun
08:49
conversation about this on the podcast
08:51
where
08:51
you know a lot of these these disasters
08:53
that have occurred like cortana for
08:55
instance was because they were training
08:57
on conversations occurring on twitter
08:59
and so
09:00
you know you got to ask yourself like
09:01
have you ever been on the internet have
09:03
you ever gone through the internet
09:04
comments that is not necessarily a place
09:05
where you want to train why is that
09:06
important
09:07
because sentiment is how we determine
09:10
how
09:10
we should emotionally react to something
09:13
and when you start training a machine to
09:15
learn sentiment from conversations that
09:17
are happening in cyberspace with no
09:18
controls with complete anonymity with
09:20
complete toxicity
09:22
we start to associate sentiment around
09:23
gender sentiment around race
09:25
sentiment around socioeconomic status
09:27
something around whether you've been
09:29
involved in the justice system
09:30
and that starts to then impact the way
09:33
decisions are being made by these
09:34
systems now it's one thing if i'm trying
09:36
to promote a pair of tennis shoes it's
09:37
another thing if i'm trying to
09:38
determine a sentencing guideline and so
09:40
that's why when you think about garbage
09:42
in garbage out garbage in garbage out
09:43
could mean that
09:44
you know somebody promotes to me the the
09:46
next pair of high heels which has no
09:48
real functional value to me okay that's
09:50
garbage and garbage out you made a bad
09:51
decision i'm not going to transact
09:53
but it could also mean making a decision
09:55
about an individual based on the way
09:56
they look
09:57
based on the way they talk based on
09:59
their gender and having that
10:01
built inherently into the system making
10:04
these decisions this is not like a human
10:05
being having a bad day this is a machine
10:07
that has been trained
10:08
to understand bias and to apply bias in
10:10
his decision making
10:12
that is a critical mistake that we have
10:13
to avoid going forward and we've seen
10:15
plenty of examples
10:16
of folks who have just trained on these
10:18
data sets and unleash things to the
10:19
public and have very bad results
10:22
yep and i would say just to add to that
10:24
before you move on to doug i know you're
10:25
going to ask question there
10:27
it also leads to economic results like
10:29
it's of course it's we don't like it
10:30
being diverse people here we don't like
10:32
it when there are
10:34
negative implications for women when
10:36
there are negative
10:37
implications for black and brown people
10:39
it affects us in a personal way
10:40
but you talked about this on the podcast
10:42
too it goes beyond that like you know
10:44
when microsoft released the racist bot
10:46
that hurts your reputation and and
10:48
reputation travels real fast right now
10:50
on the internet
10:51
it's bad business right absolutely so
10:53
yeah garbage in
10:54
means you know not only having the right
10:56
data sets but i'm sure people didn't go
10:58
in and say oh
10:59
i want to create garbage in they just
11:01
didn't have the right people there to
11:03
say
11:03
like you said if you and i were there or
11:05
doug was there like wait wait wait wait
11:07
wait
11:07
you're going to get your information on
11:09
twitter like are you out of your mind
11:11
now now we're getting into the economic
11:13
advantage of having an inclusive team
11:15
because again that was the part that we
11:16
just yeah again like how people just sat
11:18
in a room and no one
11:19
thought this would be a bad idea clearly
11:21
was lacking representation
11:23
simple representation by the way if i
11:25
could
11:26
put a perspective in there before you
11:28
ask your next question lauren no
11:30
please one of my interpretations is that
11:32
uh you know microsoft did not release a
11:34
racist bot
11:35
microsoft released a reflective bot onto
11:38
a
11:39
racist data set right and
11:42
it's even better you know what i mean i
11:44
mean it reflects
11:46
what we do it is you know the technology
11:49
worked
11:50
perfectly right as exactly as designed
11:53
i understand why they did it and how
11:55
this happened as well and i'm not giving
11:57
them a pass or anybody a pass for doing
11:59
this
12:00
but you know the question implies is
12:02
also
12:03
the we're trained as data people that
12:06
the larger your data set
12:08
the more accurate and valuable your
12:10
results are going to be so what you
12:11
really want is the biggest data set you
12:13
can have
12:14
and if you're trying to train for
12:16
conversationality
12:17
twitter is a pretty good choice people
12:20
are having conversations on it
12:21
and the data is massive and so i think
12:25
in addition to the people conversation
12:27
we're probably about to have
12:28
uh we need to think about if you're just
12:30
talking about purely the data before you
12:32
got to the team
12:33
or anything like that uh we actually
12:36
need to be proactive and intentional
12:38
to create data sets that are not
12:41
racist because our actual behavior
12:46
is racist so our bots are going to be
12:49
right so if you want non-racist bots
12:52
we're
12:52
going to have to actually proactively
12:55
and intentionally
12:56
insert data so so mass
13:00
of data is actually less valuable than
13:03
than the accuracy and the target design
13:06
that we're going i think what i would
13:09
what i would i wouldn't counter but what
13:10
i would argue there though
13:11
not to not give i'm not trying to attack
13:13
and i'm also not giving a pass but you
13:15
know there's got to be an intentionality
13:17
behind it right because to your point
13:19
yes the the the data sets that you were
13:21
going to examine as part of these
13:23
exercises
13:24
you have to put thought behind those
13:25
things and i go back to just simply
13:27
representation
13:28
if you're if you're releasing a chat bot
13:29
with the intention of putting it out
13:31
there to the world to serve all of the
13:32
people
13:33
then you should probably be thinking
13:35
harder about the how much representation
13:37
is within those data sets so i do think
13:39
there's some accountability there i hear
13:40
what you're saying and yes i mean as a
13:42
as a technologist as a scientist you
13:43
know i i often
13:45
the same thing right you give me a
13:46
massive corpus of conversations
13:48
happening out there
13:49
in cyberspace and i can do a lot of
13:51
really cool and interesting things with
13:52
them
13:52
but again is my intention to do academic
13:55
research or is my intention to provide a
13:57
service
13:58
back out to the public and so there's
14:00
just again i think there needs to be a
14:01
discipline and an accountability
14:03
to how you think about that data set how
14:06
are you calling it how are you ensuring
14:07
that it's a representative data set
14:09
if that is what your intention is to go
14:11
do um so but i don't yeah but i
14:13
absolutely agree with you doug i mean
14:14
this is a byproduct of
14:16
of something else not necessarily it's
14:18
not the ball
14:19
it's not the bot's fault let's just say
14:20
that you know it's like how there's no
14:22
bad dogs right
14:23
right but the self-awareness part while
14:25
we're on this i think this is just a
14:26
good conversation here
14:27
and really important for us to learn
14:29
from because there's a there's two
14:30
lessons one
14:32
uh right thing to do but also
14:34
economically like your
14:35
reputational damage happens quickly and
14:38
it happens
14:39
fast and so understanding that even if
14:41
you don't believe in this which we i'm
14:43
sure
14:44
most of us are preaching to the choir we
14:45
believe in being inclusive we believe in
14:48
diversity you don't have to believe any
14:49
of that like you said we talked about
14:50
this you
14:51
it is good for your business because
14:53
here's what's going to happen right
14:54
you don't approach this right uh it will
14:57
affect you
14:58
and so there has to you and you have so
14:59
and you have to know i think we have to
15:01
accept
15:02
kind of the doug's point that you are
15:04
biased
15:05
and in order and and your default is
15:08
going to be
15:08
to do biased things unless you have
15:12
a process in place to check
15:15
your own bias and people have to
15:18
want to start from there the
15:21
acknowledgement that i am biased i'm
15:22
including myself in this we are all
15:24
biased and unless you and until you have
15:27
a process
15:29
that holds you accountable to not
15:30
letting your bias
15:32
affect and infect the data this will
15:34
happen
15:35
again and again and again that's kind of
15:37
my point
15:38
yeah absolutely i completely agree
15:41
because
15:42
as much as we we are talking about like
15:44
the responsibility aspect of it like
15:46
companies and individuals
15:48
no one's setting out to create a racist
15:50
or sexist algorithm
15:51
but at the same time while we know
15:53
that's not their intention
15:55
they also have to be intentional
15:56
themselves to sort of mitigate
15:58
that bias and that implicit like
16:02
bias that is causing harm for the users
16:04
of this because
16:06
if you don't have people asking those
16:08
questions along the way
16:10
it does come back and it is your
16:11
responsibility that this happened
16:13
and this has this poor result um and
16:16
the negative publicity i think is
16:18
helping to
16:20
get more policies and more processes in
16:22
place so people are getting ahead of
16:24
that
16:24
and ensuring that these sort of things
16:27
aren't happening over and over again and
16:29
aren't
16:29
affecting people in a negative way um so
16:32
something that we're like i think we're
16:33
getting at is this like disconnect
16:35
between the data
16:36
and the actual users so we know and it's
16:40
widely known that ai can perpetuate
16:42
forms of discrimination in ways that
16:44
disproportionately
16:45
affect the impacted populations whether
16:48
that's by race
16:49
sex gender religion anything like that
16:52
and it can cause ongoing economic and
16:54
societal problems for those individuals
16:56
so when have you seen the failure of
16:59
algorithms
17:00
that have gone wrong and why do you
17:02
think it's important that we
17:03
look into that further i'll just tag
17:06
something in there on that it's more
17:08
question than answer but
17:09
uh bias is human and and we shouldn't
17:12
demonize it
17:13
uh i i i'm gonna speak to one person
17:16
in a different way than i speak to
17:17
another person because of visual cues
17:20
that that's just how the human computer
17:22
works and uh we need to recognize this
17:26
the other thing is as a technologist
17:28
this
17:29
is not how we develop technology and
17:32
computer science as a discipline
17:34
did not really develop over the last 30
17:36
to 50 years
17:38
uh mechanisms for this human aspect
17:41
because this is this is new uh we we
17:44
have
17:44
a merging of technical capability and
17:49
scale
17:50
with human uh things that
17:53
we honestly need the disciplines that
17:55
we've been poo pooing for a long time
17:57
why are you going into art history why
17:59
are you going into ethics
18:00
when computer science is the field
18:05
to go i i i mention that also because
18:08
algorithms predate uh computers right so
18:11
computers scale that algorithm and
18:13
automate it but the algorithm itself
18:15
i would say something like redlining or
18:18
gerrymandering
18:19
or uh you know some of those
18:22
economic decisions that have been made
18:26
with pencil and paper for a long time
18:28
so i just want to make sure i throw that
18:29
in there that's a really
18:32
interesting point actually absolutely
18:35
yep and i think we kind of already
18:37
touched it obviously we made a couple of
18:38
examples about how the algorithms got it
18:40
wrong we went through the bots and
18:42
and you know we we want to make sure we
18:44
actually get it right
18:45
as for the two reasons we already really
18:47
discussed so this is why i think when
18:49
companies hear this and
18:50
you know why it's in their economic
18:52
interest to have diverse perspectives
18:54
you know one of my clients they have a
18:57
company that they just do
18:59
they help they help companies avoid
19:01
making stupid mistakes when it comes to
19:02
advertising you know that
19:04
that should be easy like there was one
19:05
example of uh
19:07
i can't remember the company but it was
19:08
a company with selling beer
19:10
and and so they they went and they had
19:13
the
19:13
it was a bunch of people of color in
19:15
this bar and they pushed past
19:17
all the people of color just to a white
19:19
person and everybody else
19:20
and they didn't see they didn't see a
19:22
problem with that so you know
19:24
it obviously if you can't see and that
19:25
seems super obvious to me like that's it
19:27
but like if people can't see that how
19:31
you can imagine how much of a challenge
19:32
they have when when they're looking at
19:34
analyzing data when they think is going
19:37
to be
19:37
objective they think is as as doug said
19:40
as long as we have more
19:41
uh we're going to be more accurate but
19:43
that's actually not true
19:45
and if you have people that have
19:46
different perspectives because it's not
19:48
and it's not just
19:49
technical aspects actually a lot of the
19:52
proof has been shown that people can
19:53
make
19:53
contributions to science that are like
19:56
tangibly there that actually don't have
19:58
the level of expertise that's why it's
19:59
important to have
20:00
multiple point of views we're not just
20:01
talking race but that's important too
20:03
but also perspective
20:05
absolutely and you know it's interesting
20:06
i mean i think we talked about this a
20:07
little bit before when it comes to ai
20:08
what i'm excited about
20:10
is that it the i don't want to call it
20:12
the bar but the requirement for
20:14
contribution comes down dramatically
20:16
right this is not no longer just
20:18
requiring your your data scientists it's
20:20
no longer just requiring your data
20:22
engineers and your phds and your
20:23
computer scientists i mean we need real
20:25
people
20:26
involved in training these algorithms
20:28
and analyzing these data sets and
20:30
ensuring that they make sense and so i
20:31
think there is a real opportunity
20:33
also back to what we said before i mean
20:35
you know i think we could have a whole
20:36
panel we could talk about
20:37
inclusion and diversity and equity and
20:39
why these things are important and i'm
20:41
sort of just tired of it
20:42
frankly but what what is actually
20:44
relevant here is back to the point that
20:45
we're talking about this is about
20:46
economics and having economic advantages
20:49
you may not
20:50
think that you know seeing the people
20:53
past the bar
20:53
is that big of a deal you may not even
20:55
know to even ask those questions to
20:56
lauren's point these people don't get
20:58
together
20:58
with the intention of let's go create a
20:59
really offensive advertising for our
21:01
beer like you know that's not the
21:03
that's not the objective but again we
21:05
can see right there where that lack of
21:07
inclusion
21:08
led to a mistake which has a negative
21:10
impact
21:11
so if there's anything else to take away
21:12
it's to look at your teams that are
21:14
making decisions on behalf of the
21:15
customers
21:16
that you serve because i guarantee you
21:18
almost everybody listening has an
21:20
inclusive set of customers or at least
21:21
an intention to serve an inclusive set
21:23
of customers
21:24
so if you are not representing those
21:26
customers in that room
21:27
you have a disadvantage and you need to
21:29
think about your strategies whether it's
21:30
around how you recruit how you build
21:32
teams how you promote your leadership
21:34
specifically for the goal of not missing
21:36
out on economic opportunity
21:38
if for no other reason moves you yeah i
21:40
agree
21:42
i just want to put a shout out at that
21:43
leadership level as well because
21:46
uh we have inclusion and a lot of
21:48
different things and
21:49
and just i don't know this commercial
21:50
that you're talking about but i can see
21:52
how
21:52
different uh concepts can can collide
21:56
so at one level you might have had
21:58
someone saying we need more people of
22:00
color
22:00
in our advertising yes yes and they put
22:04
them in there
22:05
and then another person in a different
22:06
part of the organization was designing
22:08
the storyboard for this other thing
22:10
but none of those people would be able
22:12
to kill this ad before it came out
22:14
that's going to take a vp or a senior vp
22:17
or something like that
22:18
and so you can have a level of inclusion
22:21
at the
22:22
at the employment level but we need
22:25
those bored people
22:26
you need those people that you you know
22:28
the good phrase is who thought this was
22:30
a good idea
22:31
now it passed a whole lot of people who
22:34
didn't have the ability to say
22:36
uh this doesn't look good and and so i i
22:39
just want to
22:40
re-emphasize will what you're saying
22:41
about especially at the leadership
22:43
level you need ceos you need board
22:46
members you need
22:47
senior level people who can say uh we're
22:50
gonna have a
22:51
serious conversation about this ad or
22:53
this algorithm
22:55
or this decision um because i've got the
22:57
authority to do it i'm safe
22:59
i'm not worried about it no
23:03
you guys get rid of me or anything like
23:05
that i'm here to do that and and so
23:08
that's important well the term i use is
23:09
empowered are there folks not only
23:12
at the table are they empowered right
23:14
are they empowered to have a voice
23:15
to to make a decision to to to veto
23:18
something and yeah so absolutely agree
23:20
yeah and i would say on that point
23:21
before i move on because there's a
23:23
couple things i'd like to talk about
23:24
um you know will you talked about the
23:27
technology and ai and actually lowering
23:30
the bar in terms of entry that that is
23:31
true
23:33
there's actually there was a stanford
23:36
actually did an online they opened up
23:38
they did this a while ago they opened up
23:40
artificial intelligence classes online
23:42
and
23:43
and 400 and i think the best student at
23:46
stanford
23:47
came at 400 out of that class so that
23:49
tells you okay
23:50
stanford is supposed to you know only
23:52
let in the best and brightest
23:54
there's a lot of people that certainly
23:55
have the potential that never took ai
23:57
before
23:58
and are able to understand it so i think
24:00
this is there's an opportunity for
24:02
companies if they're willing to be more
24:04
inclusive to have a lot of different
24:05
perspectives here and they have to look
24:07
at
24:07
they have to look at just not the
24:09
traditional way but
24:11
how the technology we have now is really
24:13
democratizing the process and i would
24:15
say that for individuals too
24:16
for individuals listen you know you
24:17
don't you don't have to you know
24:20
well no you didn't you didn't graduate
24:21
from college you don't have to go to
24:22
college necessarily
24:24
to have a skill set in this you just
24:26
have to be passionate and be open and
24:28
be open to learning and there are many
24:30
tools and opportunities to actually do
24:31
so
24:32
absolutely yeah the learning mindset is
24:34
is so key
24:36
yeah and absolutely like nowadays there
24:39
are so many
24:39
boot camps certificate programs even
24:42
particular companies like google
24:43
they are now having training programs
24:45
where you don't need a college degree
24:47
but they will train you in what you need
24:48
to
24:49
know to be successful in that
24:51
environment and i think
24:52
like ai and data science you see a lot
24:55
of diversity of backgrounds and a lot of
24:57
diversity of people which i think is why
24:59
it's growing so
24:59
fast and you are seeing a lot of
25:03
like new things coming out and um it's
25:05
definitely something that
25:06
like in the data science teams i've
25:08
worked on i can tell the like diversity
25:10
of perspective
25:11
and the diversity in the decision-making
25:13
process like doug has mentioned
25:15
is something that is incredibly
25:17
effective and
25:18
that's something i'd really like to
25:19
touch more on so
25:22
not only is it useful to have like
25:23
diverse data sets and diverse
25:25
perspectives but diversity in that
25:27
decision-making process
25:28
especially at high levels and that's
25:30
something that like companies venture
25:32
capital firms investors
25:34
they actually put themselves at it just
25:36
like distinct disadvantage by not
25:38
incorporating those diverse groups of
25:40
people in that process
25:41
so will i know that something you've
25:42
mentioned it's absolutely economical
25:45
to include people and it shouldn't just
25:47
be social responsibility it should be
25:49
in your mind that this is a positive to
25:52
your business and a positive to your
25:54
customers for you to be
25:56
more inclusive in that whole process and
25:58
a negative not to be which is also
26:00
i think important to state right right
26:02
it affects both both sides yeah
26:04
absolutely uh
26:07
if if i could comment on that as well um
26:10
i think we are
26:12
kind of predisposed to avoid friction
26:15
and when you have more diversity and
26:19
more creativity
26:20
and more uh opposing viewpoints
26:23
it really does feel like friction and i
26:26
think that especially in fast-moving
26:28
businesses and high-growth companies
26:30
we're inclined to say uh that's bad
26:34
like you guys have been arguing about
26:37
this for an hour
26:38
and that's bad yeah and as opposed to
26:40
the more creative companies that are in
26:42
organizations that are kind of like
26:44
this is awesome we're getting to
26:46
insights that we wouldn't
26:47
if we all agree and so i think that's
26:50
really critical because
26:51
especially in technical innovation it
26:54
truly requires
26:55
friction and if we've been in
26:59
environments that were never diverse
27:00
before
27:01
we're going to experience some
27:02
discomfort and
27:04
that's just a saying that all of us who
27:06
are trying to achieve diversity and
27:07
inclusion
27:09
have to come to terms with and augmented
27:12
uh uh artificial intelligence and
27:14
machine learning are no different from
27:15
any other type
27:16
and i want to touch i i use the word a
27:18
lot with even within my own company
27:20
about just the
27:21
the discomfort being uncomfortable and i
27:23
think it's something that we have to get
27:25
much better at as a society
27:26
and you know keep that safe space
27:28
mentality as well and i often tell
27:30
people they tend to misinterpret when we
27:32
talk about safety is comfort safety is
27:34
not comfort
27:35
in my mind safety is the ability for you
27:37
to show up as your whole self contribute
27:38
your whole self and we're not
27:40
putting you in a situation where you're
27:41
holding back because again that is a
27:43
disadvantage to me
27:44
as a business leader if i have resources
27:46
within the company that cannot show up
27:48
and fully engage
27:49
but i do not need you to be comfortable
27:51
in fact i want you to be uncomfortable
27:53
at times because that's when we're
27:54
growing that's when we're challenging
27:56
our assumptions
27:57
i call it breaking the bottle we got to
27:58
get out of this trap that we've sort of
28:00
set for ourselves and so
28:02
i think for organizations to start to
28:04
not only just when we're talking about
28:05
diversity and inclusion or ai
28:07
just bring that to your culture
28:10
celebrate discomfort get people just
28:12
used to it
28:13
that way we walk into situations that
28:15
are uncomfortable with the objective
28:16
that we're going to go find a
28:17
commonality and go find solutions
28:19
i think as we start to kind of avoid
28:20
these types of things because they feel
28:22
like friction they feel like conflict
28:23
we're just we're just stuck in the water
28:25
we're not going anywhere you've ever
28:26
been
28:27
white water rafting right when you're
28:29
pushing you're moving right so
28:31
it's definitely going to be hard at
28:32
times but that's when we make progress
28:34
i like word that comes to mind when
28:35
you're talking about that is really
28:37
relevant to this because it is that like
28:39
disruption
28:39
of your current environment and your
28:42
process and
28:43
i think part of that is knowing that it
28:46
doesn't
28:46
have to continue going the way it has
28:49
been and it's okay
28:50
to revisit a process revisit a system
28:53
and change what needs to be fixed
28:55
instead of just going with the flow when
28:57
you know that there's something that
28:58
needs to be
28:59
affected and done to it like effect
29:01
change in that way
29:03
yeah i mean there are several companies
29:04
that look at like all the most creative
29:06
companies as doug say they have
29:07
different versions of this right
29:09
i think uh jeff bezos talks about going
29:11
the day one where you're always in the
29:13
mode of changing changing building
29:15
and not going to the default of okay
29:17
accepting the process because this is
29:19
the process
29:19
of what we've always had and then you
29:21
get into the kodak situation and if i'm
29:23
sure most folks know where
29:24
kodak actually created the technology
29:26
that is instagram
29:28
and they said well it might disrupt our
29:29
entire business and you know they were
29:31
right and they no longer have a business
29:32
i mean that's how that works if you're
29:33
not willing to challenge yourself
29:36
and have what malcolm gladwell called a
29:38
constructive rivalry because you can you
29:40
have something where you are actually
29:41
you
29:41
encourage the environment to challenge
29:44
the system
29:45
in a constructive way but challenge the
29:46
system and actually empower people as
29:48
well says that that should be the
29:49
environment
29:50
that you are working in and i'll make a
29:52
guarantee to you if that's not the
29:54
environment that you're in
29:55
it's not going to if it's a startup it's
29:57
not going to last but if it's a company
29:58
it's probably not going to last either
30:00
absolutely yeah i um so i'd like to
30:03
bring it back to like will you
30:04
specifically
30:05
and what you're doing with search engine
30:07
so an example i have that i'd like to
30:09
bring up of how something went very
30:11
wrong
30:12
back in 2009 sophia umya noble she
30:15
searched for the terms
30:16
white girls on a search engine and came
30:19
up with a lot of nice stock images
30:20
and then she searched for the terms
30:22
black girls latina girls and asian girls
30:24
and each time the first page was filled
30:26
with pornography
30:27
so she wrote a book about this called
30:29
algorithms of oppression and she poses
30:31
the idea that search and
30:32
engine results are not nearly as neutral
30:35
as they may seem
30:36
so i want to ask you what people
30:38
commonly get wrong when they think about
30:39
how
30:40
search engines work and what are you
30:41
doing differently
30:43
yeah i mean well yeah okay let me answer
30:45
those separately so i mean
30:47
in the case of google and i'm fully
30:49
aware of the example and you know do the
30:51
same thing with with males as well look
30:53
up white teens look up black teens
30:55
you're going to see
30:56
you know kids playing in the park and
30:57
you're going to see you know
30:59
people being incarcerated literally this
31:02
one is
31:02
tricky when you look at it from the
31:04
perspective of let's just say
31:06
google and you know their algorithms
31:08
because
31:09
a lot of what informs ranking within
31:12
google
31:12
is again the internet and the linking
31:15
across the internet the way people
31:18
reference
31:19
women of color versus white women the
31:22
what they link to what they talk about
31:24
the volume
31:25
right and so what you're seeing there is
31:27
again a pretty
31:29
dark reflection of society and
31:33
the question i think is more so along
31:35
the lines of
31:36
what is the responsibility of somebody
31:38
like a google
31:39
to curate that sort of content and that
31:42
sort of behavior and you can imagine
31:44
that that opens up a whole can of worms
31:47
right because now it's interesting we
31:49
talk about sort of augmented
31:50
intelligence to prevent us from just
31:51
doing something stupid
31:53
but now when you're talking about sort
31:54
of shaping information
31:56
you're now actually having a human again
31:58
inject their bias into
32:00
what what is okay
32:03
what's not okay can i celebrate the
32:05
female body can i not i mean there's all
32:07
these questions that sort of come out of
32:08
this so
32:08
i just want to call out that that's a
32:10
very difficult question that i think
32:12
even as a society we're not necessarily
32:14
sure how to answer i don't know that i
32:16
can blame
32:17
google for using linking as part of
32:19
their algorithm it clearly points to
32:22
where the internet has decided that
32:24
these things are associated
32:25
right but we we probably have a lot of
32:27
discussion to determine
32:28
at which point is a provider responsible
32:31
for curating that and as they start to
32:33
curate does that even change the dna of
32:35
who they are
32:37
so meaning like am i no longer just this
32:39
search engine am i actually like a
32:41
content provider
32:42
a content stream am i you know am i uh
32:44
an editorial you know that type of thing
32:46
so it's something to think about
32:47
when it comes to us so so my company we
32:49
build um what we call sort of ai powered
32:52
search engines
32:53
we work with a number of top retail
32:55
brands we work with
32:56
large companies that are just trying to
32:58
use artificial intelligence and search
33:00
technology to help people
33:01
find information within the company if
33:03
you ever worked for a big business or
33:04
worked for a
33:05
a government like like doug you know
33:07
there's just massive amounts of
33:08
information and so we really help people
33:10
use the same approaches that google or
33:12
amazon uses to their experience
33:14
to understand their corporate
33:15
information because of that kind of to
33:17
what doug
33:17
said we have one advantage in that we
33:20
are typically looking at a very specific
33:22
corpus if i'm in retail
33:23
obviously everything i'm doing is in and
33:25
around that catalog and how i'm helping
33:27
people
33:27
engage that catalog more discover that
33:29
catalog more if i'm working within a
33:30
bank
33:31
it's all around financial services and
33:33
the data that they have
33:34
within the bank that proprietary data so
33:36
we don't get
33:37
exposed as much to some of these issues
33:40
that others might have
33:42
but we do often need to think about
33:44
again like the way we're making
33:45
assumptions about
33:46
ranking there was an interesting
33:48
discussion that went into my company
33:49
about we were we were using color
33:51
as a higher weighted um field basically
33:54
a piece of metadata
33:56
compared to another field and it got
33:57
into an interesting conversation that
33:59
is that a gender bias to assume that a
34:01
recommendation engine that color matters
34:04
is that just something that we just
34:05
assume because we happen to be looking
34:07
at
34:07
you know a women apparel store but the
34:09
moment we go somewhere else
34:10
does that go out the window so you see
34:12
that little thing this wasn't
34:13
necessarily about inclusion or exclusion
34:14
but it was one
34:15
sort of interpretive point that could be
34:18
seen as being biased that could take us
34:20
down a variety of paths which may or may
34:22
not be advantageous to our user
34:24
so there is some complexity that comes
34:25
into those those scenarios but i do have
34:27
the advantage of
34:28
we're not out there trying to consume
34:30
the internet and build recommendation
34:31
models and things where
34:32
again as doug was saying we're taking
34:34
trillions of records because that's how
34:36
we get a good inclusive data set
34:38
and we're allowing the the biases and
34:40
the sentiments within those records
34:42
to make decisions we don't really get
34:43
into that as much
34:45
yeah what a great problem though
34:47
speaking of friction
34:49
and controversy and oh that's a longer
34:51
conversation
34:52
because we're going to come up with
34:53
better products where that's concerned
34:55
and i agree with everything you've said
34:57
will thinking about you just sort of
35:00
remove google from that
35:01
and what do all of us have as a
35:03
responsibility to building
35:05
effective search and what are our actual
35:08
objectives
35:09
the original algorithm of google has to
35:12
do with
35:13
rankings and has to do with what is the
35:15
top because that's what the internet
35:17
says you know that based on everything
35:19
that's out there
35:20
and it's probably what you're looking
35:22
for and that's kind of good enough
35:24
but search now is
35:27
is more detailed and has
35:30
actual objectives and so forget about
35:33
google for a second
35:34
when you're searching for employees and
35:38
you're
35:39
you're using your network and you're
35:41
looking for candidates you're really
35:42
acting as a search engine
35:43
and when you are making decisions about
35:45
parole
35:46
for offenders you're you're kind of
35:49
acting
35:50
as an algorithm it's a much smaller data
35:53
set
35:53
and i do believe and this is
35:55
controversial will so i'll be
35:57
interested in what you think about this
35:59
i do believe we need to
36:01
insert some intentionality into
36:04
the objectives that we're trying to
36:06
achieve we have a universality
36:09
and a illusion of no
36:12
bias that says i'm not going to touch
36:15
the data
36:16
because that would bias it well data's
36:18
already biased and the results are
36:20
published
36:21
so now if we want to accomplish a
36:23
certain hiring
36:24
or fairness and sentencing or uh
36:27
fairness in giving loans out
36:29
we're going to have to insert objectives
36:32
and then we're going to need to
36:33
look at the data set that comes back and
36:35
iterate through it
36:36
and say well did this come back with the
36:38
kind of inclusion that we were looking
36:40
for if not we need to iterate
36:43
and that is controversial and and it
36:45
should be and and
36:46
i'm thankful that we're at this moment
36:48
right now where we're having an honest
36:49
conversation
36:51
about race and sex and sexism and and
36:54
and you know abuse and and these things
36:57
so that we can begin to say
36:59
wait a second this goes against values
37:01
that we used to have
37:03
but today's values we want to use our
37:06
technology to accomplish our societal
37:09
goals
37:09
and that makes this a legitimate thing
37:12
to move after
37:13
absolutely yeah i agree completely in
37:16
that we should
37:18
figure out ways to use data in a way
37:21
that's intentional i know
37:22
um china has some technology that they
37:26
are doing more
37:27
in terms of inserting in the process of
37:30
how the criminal justice system works
37:32
and and then
37:32
they're looking to see okay is this
37:35
prosecutor or judge
37:36
out of line in terms of how they're
37:38
sentencing with these cases and if they
37:40
are is there are there some corrections
37:41
that need to be made
37:42
uh i do think there are and there should
37:44
be opportunities
37:45
uh to use data i i think the opposite is
37:49
being done right now but i do think
37:50
we need to be more intentional the one
37:52
that comes really comes to mind is
37:54
facial recognition and how that's being
37:56
used
37:56
and there's a lot of challenges there
37:58
because they're you know it
38:00
facial recognition we know is very good
38:02
at pale mail
38:04
and it's but nothing else and so but but
38:06
still
38:08
uh you know people on the security side
38:10
want to use it to identify
38:11
people even though it's horrible at
38:13
identifying
38:14
anybody but white men for this sport but
38:16
they still want to use the technology
38:18
though it's not ready to be used in a
38:20
way unless we're ready to have the
38:21
conversation about intentionality so
38:23
i agree completely but i don't i'm sure
38:24
there's some disagreements on that but i
38:26
do
38:28
yeah i think it's a big thing right now
38:30
it's
38:31
really coming to light and people are
38:32
really invested
38:34
in making sure that companies and
38:36
individuals are using this data and
38:38
using these algorithms in an ethical way
38:40
one that comes to mind when we're
38:42
talking about like the ethics of how
38:44
you're using your data is facebook and
38:46
the election data
38:47
so whether you know it or not the things
38:50
that you're seeing based on
38:51
the data you consume and the data you
38:53
put out
38:54
can change your opinion and it can help
38:57
you form even more biases and it can
38:59
convince you of certain things and so i
39:02
think
39:03
like it's really good that it's coming
39:04
to light right now and people are
39:06
very invested in ensuring that our
39:08
algorithms and our data is being used in
39:10
the right way
39:12
and one person that i'd like to bring up
39:14
is joy la bolewini who founded the
39:16
algorithm
39:17
algorithmic justice league and she made
39:19
her mission with this
39:21
is to bring light to like the harm that
39:22
these discriminatory algorithms and ai
39:24
technologies can have on these affected
39:26
populations
39:28
and so i wanted to like talk about if
39:30
there's any other companies and
39:31
initiatives that
39:32
you know about that are taking like
39:34
these matters into their own hands and
39:36
like helping
39:36
address and mitigate this bias in data
39:39
algorithms and ai
39:40
yeah and if i can expand that question
39:42
too if you don't mind too i think to
39:43
to say what do you see as the future
39:48
of our relationship with data and how
39:51
how can we shape that i know well we've
39:53
talked about that but i think doug it's
39:54
a question that you've talked about a
39:56
little bit but i think
39:57
really expanding on that is it's a
40:00
really key concept so
40:01
i have one just point of view on this
40:03
that i think is going to come here
40:04
sooner rather than later i'm not aware
40:06
of as many organizations back to
40:08
lauren's question but i'd love to learn
40:10
and get to know them
40:11
um but but one of my my my sort of
40:13
theories for the future is is that we
40:15
are going to see a world where
40:17
we become far more transparent with in
40:19
terms of our privacy exchange
40:21
so if you think about it today we all
40:22
know this none of these things are free
40:23
i think everybody's kind of fully aware
40:25
of that we're using all these services
40:26
we're paying for them we're paying for
40:28
them with personal information
40:29
it's a commodity now i'm not here to
40:32
debate whether we should be doing that
40:33
or collecting personal information it's
40:34
it's it's a free country human being
40:36
should be able to
40:37
engage any way that they want but the
40:39
point is that today there's not a lot of
40:40
transparency
40:42
both in the sense that the companies
40:44
don't really know what they're going to
40:45
do they just know they want to go
40:46
collect it all
40:47
and i don't really understand the value
40:49
of what i'm exchanging with you
40:51
so i see a world in the future where
40:53
that exchange becomes far more
40:55
transparent almost like
40:56
any other market exchange that we see so
40:58
let's take facebook for an example with
41:00
the facial recognition and the facial
41:01
scanning
41:02
facebook might decide to say hey look
41:03
we're going to scan your pictures
41:04
because we're building a big
41:06
you know corpus or training set of faces
41:08
that we're going to go use for whatever
41:09
the heck we want to go use it for i
41:11
should be able to decide you know what
41:12
i'm not okay with that
41:14
opt out what is the exchange 50 bucks a
41:16
month
41:17
okay well i can make those decisions i
41:20
can start to make those trades with that
41:21
commodity far more transparently
41:23
so i don't love regulation when it comes
41:25
to technology i think it can get in the
41:26
way of innovation
41:28
but i do think regulating that the way
41:30
that my
41:31
information gets used and providing me
41:33
insight into that so i can make a
41:35
decision
41:36
as a consumer is something that i would
41:38
definitely like to see
41:39
some regulatory involvement in because
41:42
if anything else is just allowing me to
41:43
participate fairly
41:45
in this value exchange that we've been
41:47
having now for a decade and it's new and
41:49
it's moving quickly
41:50
and this is also why i think a lot of
41:51
mistakes occur because companies are
41:53
sitting here
41:53
worried that if i don't move
41:55
aggressively if i don't move quickly
41:56
back to rob's point i'm going to miss
41:58
this wave and i'm going to become
41:59
irrelevant
42:00
so i'm empathetic to some of those
42:02
challenges i'm also just again as a
42:04
technologist
42:05
i'm fascinated by what's possible with
42:07
what's out there
42:08
but definitely again opportunities now
42:11
to kind of think about
42:12
new ways of engaging in this digital
42:14
world and creating a fair world where
42:16
you don't have to be
42:17
a digital native you don't have to have
42:19
a deep understanding of privacy
42:21
and technology to use these services in
42:23
an informed way
42:26
yeah well we'll in in addition to being
42:28
a consumer you are also a citizen
42:30
and that's the thing people have
42:32
different roles
42:34
in society and we've seen the private
42:36
sector
42:37
has figured out how to really work hard
42:40
on
42:41
on that consumer part but government and
42:43
public sector
42:44
we we're way far behind in dealing with
42:46
that system but that's coming
42:48
and i think when you look at china you
42:50
you can see
42:51
that you know the relationship between
42:54
government and citizen and government
42:55
and individual is pretty important and
42:57
we need to get behind it so i want to
42:58
put a shout out here
42:59
to say we need some more doug's uh we
43:02
need to have conversations in government
43:04
uh we need to sit down and talk through
43:06
this uh it's
43:07
it's important let's have the fight
43:09
right um
43:11
so in dublin at the risk of saying even
43:13
more controversial stuff
43:14
i think about this a lot and we are
43:17
experimenting with a digital identity
43:20
product uh really system and platform
43:23
that is based on blockchain
43:25
so not going to go down the road and
43:26
talking all about blockchain and
43:28
everything
43:28
except for to say that we have the
43:31
opportunity today to develop
43:33
systems based on things that governments
43:36
and big corporations
43:37
cannot influence as much as we're
43:40
equalized
43:41
individuals are going to have some
43:43
influence where that's concerned
43:44
absolutely there's an opportunity for
43:46
transparency
43:47
and potentially in the world that i
43:50
envision the future world
43:52
we could have a situation where all of
43:54
the data that's being gathered
43:56
and observations that quite frankly this
43:59
is happening around us you don't have
44:01
control
44:02
over your data necessarily but we as a
44:05
government or
44:05
public sector should be in a position to
44:08
empower you as an individual
44:10
to see these things as they're gathered
44:12
and as you say will
44:14
opt out or exchange value sorry
44:17
and i don't know that you can have a
44:20
one-on-one relationship
44:22
with every company that has a value
44:24
stake in your data
44:26
but you might have a one-on-one
44:27
relationship with the community in which
44:29
you live
44:30
which might as a service have a way of
44:34
you
44:35
controlling or at least being able to
44:37
observe and have some transparency about
44:39
your data
44:40
certainly with regard to government
44:41
services so i do see a
44:43
a wider future at some point in which a
44:46
government
44:47
where you know taxes is not going to be
44:49
the way we necessarily fund ourselves
44:52
may have a value-added service that is
44:54
trusted
44:55
by people that i'm going to use this
44:57
service to interact
44:59
with a thousand different companies and
45:01
i'm going to know
45:02
when i'm being observed when when my
45:04
credit is being used
45:06
for a particular purpose and and it's
45:08
just going to
45:09
be different than what we think about
45:11
today and one last point i'll make about
45:14
this
45:14
i think this can be used as a tool
45:16
against bias
45:17
and against discrimination against
45:19
racism because it has the potential at
45:21
least
45:21
to equalize people's ability to do
45:24
something like
45:25
it might my access to a phone or a small
45:28
digital tool
45:29
is democratized it's not
45:32
i don't need a cray super computer in
45:34
order to
45:36
do this i should need a cloud-based
45:38
service
45:39
that i can contract with for pennies a
45:41
month right so that's the future i
45:43
yeah i love it oh that's a great answer
45:45
those are great answers that
45:47
i didn't i didn't think about it from
45:48
the government point of view but that's
45:49
a really important point
45:50
that you can kind of be the intermediary
45:53
in between
45:54
you know and to make sure that that
45:56
happens because it's not gonna it's real
45:57
i see it difficult to try to figure out
45:59
well how do you do that with
46:00
every single person that's trying to get
46:01
data that's fascinating and i think it's
46:03
really innovative that's a
46:04
really good thought process i never
46:05
thought about that way yeah i really
46:07
appreciate the
46:08
idea of transparency and that you're
46:11
bringing that
46:12
up so often and including that in like
46:14
your thought process for this
46:16
one of the things that like really
46:17
struck me when i was first learning
46:18
about data ethics was this idea of like
46:21
the unequal exchange of information
46:23
so back before like back in the day
46:25
before internet before computers before
46:26
we all had cell phones
46:28
your neighbor down the street you know
46:29
as much about them as they know about
46:31
you
46:32
in the case with these large companies
46:33
that have your information it's an
46:35
unequal exchange you're giving them your
46:37
information
46:38
and you aren't getting information back
46:40
in return so i like this idea of the
46:42
transparency in this more equal exchange
46:44
in terms of that and especially if
46:46
government can
46:47
help provide that for the specific
46:49
communities i think that's a great place
46:51
that we're moving toward
46:53
yep i agree absolutely um another
46:56
go ahead go ahead i was gonna say you
46:58
know what else is funny is that
47:00
a lot of this is not data that you gave
47:02
them or that even exists
47:04
the algorithms are so strong now that
47:06
their predictive capabilities
47:08
they can tell you my political
47:11
persuasion although i may never have
47:14
shared that they can tell you my sexual
47:16
preference they can tell you
47:17
all kinds of things that they can derive
47:20
based on
47:21
really strong analytics so you know it's
47:24
it's funny we
47:25
we assume that there's some data set
47:28
there that's completely accurate
47:30
they can do marketing on the basis of a
47:32
prediction algorithm that may or may not
47:34
be accurate
47:35
but they can still make a lot of money
47:36
on it and that's that's going to get
47:38
really weird because i can stop you from
47:40
using my data
47:41
but i don't know that i can stop you
47:43
from predicting how i'll do things just
47:45
because i'm a
47:46
x year old black man who lives in
47:48
reynolds
47:50
well this is i mean this is this is
47:51
exactly and and i think there's a
47:53
there's a concept here when it comes to
47:55
again how we explore
47:56
the call it the propensity i won't even
47:58
say bias the propensity for something
48:00
bad to happen
48:01
and it's what we call in my business
48:03
it's called signal data
48:04
right so to your point doug there is
48:06
data that is like explicit that you know
48:08
i request something i download something
48:11
then there's all this telemetry all
48:12
these signals that are out there and
48:14
part of you know what you do to build
48:15
good recommendation algorithms is you
48:17
understand okay can i aggregate a
48:19
variety of signals
48:20
to determine the next best action for an
48:22
individual so i may never have seen you
48:24
before
48:24
but the way you show up to my site the
48:26
way you move your mouse the things that
48:28
you look at the longest is all sort of
48:30
informing me
48:30
so that's a very you know kind of a cool
48:33
way to
48:34
build these predictive models but it's
48:36
also an opportunity for us to look at
48:38
each individual signal that we're using
48:41
to make these decisions to determine
48:42
bias i used a very basic example for my
48:44
company and again we weren't worried
48:46
about
48:46
offending anybody we were worried about
48:48
not having the most effective algorithm
48:49
for conversion
48:50
and that was whether or not we're going
48:52
to take color into consideration
48:54
because if you think about it in a
48:55
shopping ecommerce context when you're
48:57
building a recommendation model
48:59
color can be incredibly important it can
49:01
become irrelevant
49:02
very quickly and how we weight that
49:05
field will
49:06
create bias within that model so just
49:08
looking at that one signal color we can
49:10
have a conversation around whether this
49:11
can have a negative impact
49:13
we should be doing this around all of
49:15
the signals now when you get into things
49:16
like deep learning and vectors and where
49:18
we the
49:19
computer now has gotten not only good at
49:21
these performing these operations but
49:23
actually determining where new signals
49:25
might occur
49:26
and pulling those in we also have an
49:28
additional challenge but again you can
49:30
go back to
49:31
what is the the unit that we are trying
49:33
to make sense of and what's our
49:34
intention
49:35
and have it have just a healthy
49:36
conversation about what could go good or
49:38
bad as a result of that
49:39
and i think that could alleviate or at
49:41
least prevent
49:42
a lot of these mistakes from occurring i
49:45
mean that's such an important point that
49:46
i want to touch on
49:47
uh lauren you and i talked about this
49:49
before in the prep that
49:50
uh netflix got in trouble with just this
49:52
right they went
49:53
and they they were they were they were
49:55
doing exactly what you said well
49:57
uh signaling and looking at that and
49:59
then making recommendations for what
50:01
people want to see they still do this
50:03
now well you had one woman who was from
50:05
a very conservative family and she was
50:07
catholic and
50:08
didn't want didn't want necessarily
50:09
everybody to know that she was lgbt
50:11
right and they but they they outed her
50:13
they said oh look we know for the lgbt
50:15
and they came up
50:16
and they looked and then you know our
50:17
family looks at this like wow like how
50:19
did you know that and
50:20
because of what you talked about but i
50:21
think there has to be the conversation
50:24
not only can we do this should we do
50:27
this
50:28
and if we're going to do this how do we
50:30
do this in an intentional way to cause
50:32
the least amount of harm as we have more
50:33
and more conversations about technology
50:36
that conversation always has to be there
50:38
because if not you'll have
50:40
more consequences than that but that's
50:41
one obvious obvious obvious example of
50:44
how things can go wrong and you cause
50:46
harm and that was likely unintentional
50:48
but again they were just following
50:49
following the signals and there's no
50:51
but there's no it but there's no process
50:53
in place to look at should we do this
50:55
what are the harmful effects and how are
50:57
we how will we account for that
50:59
uh that you can run across issues like
51:01
that and i think we will continue to run
51:03
across challenges like that as we go
51:04
along
51:05
as we can as we really talk about the
51:06
ethics with this tough with
51:08
with with the technology too that has to
51:10
be part of the conversation
51:11
um so the big deal with this uh with
51:14
netflix and it's kind of scary is that
51:16
it and it ended up in a lawsuit was that
51:18
um
51:19
they had this public contest and they
51:22
set out this public anonymized data set
51:24
and so they removed what they thought
51:26
was any personally identifying
51:28
information
51:28
and from this researchers were still
51:30
able to identify specific individuals
51:33
just from their history on netflix and
51:36
what we don't think is any identifying
51:38
information like our name our zip code
51:40
where we live things like that
51:42
and that's what is so scary because your
51:45
personal data like
51:46
even if you're not giving specifics
51:48
about yourself
51:50
there's so much that can be found out
51:52
and identified
51:53
to you um based on just like what you're
51:56
talking about
51:56
well like how you go about certain
51:58
things and how you interact with
52:00
different tools so for both of you
52:02
do you have any like processes or
52:03
systems in place that you
52:06
to help you ensure that these like pro
52:08
that your data and
52:09
your technologies are unbiased
52:14
well i'm curious to hear doug's answer
52:15
from our perspective because we serve
52:17
clients we're not doing our own data
52:19
capture our own data collection so we're
52:21
typically working with them as part of
52:23
their strategy
52:24
um and so most you know retailers um
52:27
yeah we do work with some governments
52:29
some healthcare
52:30
and we'll partner with part of their
52:31
privacy strategies and their data
52:33
strategies to determine
52:34
you know what is the appropriate level
52:36
of information that we want to process
52:38
really curious to hear how doug thinks
52:40
about this but before doug goes to that
52:42
let's do you have a pro but do you have
52:44
when you think about how you approach it
52:46
with them do you do you
52:47
have a certain process to make them
52:49
think about how to view this in a way so
52:51
you don't get the
52:52
you know you don't get the you know have
52:53
to pick on microsoft or whatever you
52:55
don't have a microsoft example like
52:56
is there a process you may take your
52:58
clients through to consider as they are
53:00
collecting and going through the process
53:02
you got to remember we're playing a
53:03
pretty small piece in the overall
53:05
picture right so most of these brands
53:06
and these retailers we we
53:08
dealers we work with are typically at a
53:09
minimum of about a billion dollars a
53:10
year in revenue so they're very large
53:12
they're omni-channel
53:14
and so we're coming in to help you know
53:16
with with a piece of their
53:17
call it data technical strategy for
53:19
their primarily their digital properties
53:21
right we are seeing some people bring
53:22
these into the stores and
53:23
and those kinds of things and when those
53:25
in those particular situations you know
53:26
we don't
53:27
impact a whole lot we're not deciding
53:30
you know visual treatments we're not
53:31
deciding who the
53:32
stock photos should picture we're not
53:34
deciding how you know inclusion gets
53:36
represented in those situations
53:37
what we're here to do is to help you
53:39
effectively sell more right so we're
53:41
helping you convert more
53:42
customers by getting them information
53:43
that they want we're empowering your
53:45
merchandisers to make those human
53:47
decisions to decide hey
53:48
the recommendation engine says this but
53:50
you know what if a person is looking at
53:51
these hiking boots and looking at this
53:53
pull i guarantee you i can sell them
53:54
this hat
53:55
you know that's something that like a
53:56
human being who has some domain
53:58
understanding and who's a hiker can make
54:00
those decisions right
54:01
so i just i bring that up not to say
54:03
that to give ourselves any excuses we're
54:05
still responsible for the decisions that
54:06
we make
54:07
it's just typically where we come in a
54:10
lot of that has already been
54:11
established on the other hand side of
54:13
the house for us like i said we work
54:14
internally with companies to help with
54:16
their own information
54:17
there i think there's a lot less risk
54:19
because what we're looking at there is a
54:21
lot more just domain specific
54:23
information if you're an oil and gas
54:24
company we're classifying
54:25
your maps and your research and we're
54:27
doing some image processing and we're
54:28
making that stuff retrievable
54:30
um typically does not fall into the the
54:33
the bucket of where you know some human
54:35
biases can
54:36
impact sort of the the social experience
54:39
that i have
54:40
if anything it just makes me less
54:41
effective at my job when you fail to
54:43
surface information that i'm looking for
54:47
yeah and and so i i started it out today
54:50
by talking about how
54:52
typically in a small local government we
54:54
are not big
54:55
contributors to new technologies like we
54:58
don't build it
54:59
we tend to work with an innovator like
55:01
will's company
55:02
um but that being said we are doing that
55:04
when it comes to this blockchain
55:05
experiment
55:06
and this is what we think is the wave of
55:09
the future so
55:11
we uh grew up in a way that we did not
55:14
have
55:15
specific methods to prevent that and now
55:17
that data is exploding and our
55:19
information about
55:20
citizens is exploding we really need a
55:22
new platform a new
55:23
way of thinking about it and that is the
55:25
hope for that technology
55:27
that being said we also take a similar
55:29
philosophy to kind of what will is
55:31
saying
55:32
we did a machine learning exercise in
55:34
some data that we were not really
55:36
allowed to look at
55:37
and examine uh we have some tax data
55:41
that
55:42
the majority maybe two or three people
55:44
in the city are really
55:45
allowed by law to even be reviewing and
55:48
but we wanted to do some predictive
55:49
analytics on it
55:50
what we managed to figure out what to do
55:52
is to take our algorithm
55:54
our machine learning model and put it on
55:56
the inside of that firewall
55:58
so i don't need to see it and i i'm
55:59
really excited about this potential
56:02
that our our privacy involves people
56:06
looking at someone's information in
56:08
using it for a nefarious purpose
56:09
but if we can take the technology and
56:11
place it within the data set
56:13
and generate the uh insights and
56:17
and those kinds of things they become
56:19
useful
56:20
while still being anonymous and so i
56:22
want to continue thinking and working
56:24
about that as well
56:25
the the third thing that i want to make
56:27
sure i also capture here
56:29
is we often get into talking about the
56:31
dark side and the negativity
56:33
about how these analyses and these
56:35
insights they just seem to emerge
56:37
because this technology is so efficient
56:40
at being able to now tell all sorts of
56:42
things about people
56:43
there are positive things that we could
56:45
do with this
56:46
while we're trying to protect ourselves
56:48
and i'm going to say something
56:49
controversial about china now
56:51
and their social score which we all
56:54
appropriately hate
56:56
and we think it's a terrible invasion of
56:58
privacy that
56:59
you know people could get a social score
57:00
on some of these basis but i've also
57:02
seen other perspectives talking about
57:04
how
57:05
you know if you're the unbanked and you
57:08
do not have a credit score
57:09
you're not in a position to take
57:11
advantage of other signal world
57:13
that other people have like a bank
57:15
account like owning property like a
57:16
driving record
57:18
like those kinds of things um and so
57:21
without i shouldn't even talk about
57:23
china but without going down in that
57:25
direction
57:26
how can some of these signals be used by
57:28
people
57:29
in a disadvantaged situation to require
57:33
credit
57:34
to get a bank account to get a better
57:37
score to get
57:38
uh some credibility in getting a job and
57:40
those kinds of things if those signals
57:42
could indicate a certain amount of
57:44
reliability and trustworthiness
57:46
you know we should continue thinking
57:49
about
57:50
how some of this data could be used for
57:52
the least of these
57:53
to build a profile about themselves that
57:56
they don't have
57:57
the 900 smartphone and the
58:00
the location tracking that would
58:03
otherwise build that profile for them
58:05
yeah i
58:05
actually completely agree with that is
58:06
you you sparked some thoughts in my head
58:08
one
58:09
um you really can't opt out of the
58:11
system anyway i tell people this like
58:12
look we gotta
58:14
you can try if you're opting out of
58:16
technology you're opting out of a whole
58:17
bunch of opportunities you're opting out
58:19
of
58:19
really powerful things so like yeah so
58:21
like you're you're you yeah your
58:23
society thank you there's a better way
58:24
to stay you're opting out of society
58:25
essentially so
58:26
it's not to think you can just be
58:28
anonymous it's probably not going to
58:30
work the better
58:31
the better idea is like like you know i
58:33
know there's a lot of worry about facial
58:34
recognition
58:35
frankly the more inputs they get of
58:37
people of color the more accurate it
58:38
becomes that's one example so i know
58:40
that's kind of controversial but
58:42
it's not like they're going to stop
58:43
using facial technology it's not going
58:44
to happen so
58:45
what we have to do is figure out how we
58:46
inject ourselves in the process not
58:47
remove ourselves from the process
58:49
because that is not going to do anything
58:51
except
58:51
removing from society i think that's a
58:53
good example uh beyond that though
58:55
there's also
58:56
you know there are there are great
58:57
opportunities to figure out how to use
58:59
data
59:00
to empower people and we've seen
59:02
examples of how that can play
59:04
out so i i do think there are a way
59:06
there are there are ways to really look
59:08
at this to say well how can we use data
59:10
to make more accurate decisions in
59:11
health how can we use data in a way that
59:14
makes better decisions with our criminal
59:15
justice system
59:17
we know that these things can and do
59:19
work but i know there is a resistance
59:21
and i understand the resistance
59:23
uh from people wanting to worry about
59:25
their data i tell people
59:27
you really don't have any privacy right
59:28
now so the issue is
59:30
i mean if i'm really honest with you the
59:31
issue is how do we want to better affect
59:34
society moving forward yeah and how can
59:37
we help people
59:39
and how can we use this data to help
59:40
more people because at the end of the
59:42
day
59:43
when people are crying about their
59:45
freedoms and things like that
59:46
as they do it on facebook i got news for
59:48
you we know already so like
59:50
it's like from a cell phone talking
59:52
about how vaccines are going to track
59:54
you
59:54
um i will say i mean one of my one of my
59:56
favorite projects
59:57
is a um is the matahari project out of
59:59
india and this was um
60:01
a big data project that was all around
60:03
identity you know india obviously had a
60:05
massive population
60:07
people living in various situations and
60:09
you know a lot of folks who are who are
60:11
needing
60:12
help are hard to identify so part of
60:14
this project was could we create a
60:15
system to provide
60:17
effectively like a a digital identity
60:19
for every single citizen so we now
60:22
understand the demographic of this you
60:24
know multi-billion person country
60:26
and you know india has about 60 billion
60:28
dollars of social programs that go in
60:30
every single year but not a lot of
60:32
visibility into where they need to be
60:34
distributed the most so the result of
60:35
this project that having all these
60:36
signals and having all this information
60:38
was really to help them make much more
60:39
informed decisions
60:40
on where they could invest these social
60:42
programs so i love things like that
60:44
i love hearing doug talk about you know
60:45
signals i think about this in hiring all
60:47
the time right when we think about
60:49
you know if i were to hire somebody off
60:50
of a resume off a paper off of where
60:52
they went to school
60:53
one you can make a lot of really bad
60:54
hires that way but two you miss out on a
60:56
lot of things so when i'm looking at
60:58
somebody's background there's a lot of
60:59
signals that i look for
61:00
overcoming adversity you know if you
61:02
were the first person in your
61:04
your family to go to college you came
61:05
from a single family household you had
61:07
to work to put yourself through college
61:08
but you got that degree
61:09
then you got your first job then you put
61:11
your bat yourself back into night school
61:12
to get your mba
61:13
like that is grit determination forget
61:16
about you know whether or not i'm you
61:18
know feeling
61:19
like your story is pulling on my
61:20
heartstrings i see an individual who is
61:22
performant
61:23
who is serially successful and so again
61:26
this is where i just
61:26
when doug was talking about credit it
61:28
made me think okay if that's how i think
61:29
about hiring
61:31
what makes you decide whether somebody's
61:32
fiscally responsible
61:34
it's not just whether or not they're in
61:35
debt as we know debt can sneak up on you
61:37
from a million
61:38
different ways for a million different
61:39
reasons life and death sometimes health
61:41
right
61:41
but there are plenty of other things
61:43
that we can interpret so i think it's
61:45
it's giving me a lot of encouragement in
61:46
hearing this conversation and thinking
61:48
about what's possible obviously we talk
61:49
about some of these pitfalls but there's
61:51
a lot of awareness
61:52
of these pitfalls as well i think lauren
61:53
alluded to that as well so it's an
61:55
exciting time and
61:56
participation is key and i i would say
61:59
as to your
62:00
the thought i missed that missed me that
62:01
i wanted to talk about is one of our
62:03
speakers one of our partners solo funds
62:05
is doing actually just that
62:07
was it was founded by rodney williams
62:09
and travis holloway they got a lot of
62:11
their funding actually here initially in
62:12
cincinnati
62:13
but solo funds give theirs they're
62:15
essentially disrupting the payday loan
62:17
industry
62:18
and they give loans for i think under a
62:20
thousand dollars
62:22
and they help re-evaluate looking at
62:24
different signals for
62:25
what a credit worthy borrower is and
62:28
they help
62:28
they let lenders and borrowers set their
62:31
own terms
62:32
and the i mean the only thing they they
62:34
they do is they make sure that
62:36
you don't set a i think above 10 or
62:38
something like that but beyond that you
62:40
can set your own terms
62:41
and they also i think ensure a lot of
62:43
the borrowers up to 90 percent of their
62:44
loans and they've
62:45
they've had really really they have
62:46
really low default rates and they've
62:48
been able to help and what they're
62:49
trying to do is build another
62:51
kind of digital footprint for what what
62:53
a credit score could be
62:55
and then hopefully take that to uh
62:57
larger banks but i think they they're
62:58
scheduled to do about 20 million in
63:00
loans this year
63:01
because they kind of picked up with
63:02
kovitz so um so it's it's so that is
63:05
that is already happening solo funds you
63:06
could
63:07
you you've heard from travis holloway
63:08
earlier today if you saw our
63:10
our our our family feud event you're
63:12
going to get a chance to network with
63:12
him later
63:13
so these things are being done and i
63:14
want to encourage people this is why
63:16
you know travis started and you know had
63:19
no had
63:19
he didn't have a bunch of money he just
63:20
had a back he had a background in
63:22
finance they went out and they were able
63:23
to do this so
63:24
there are opportunities and i know
63:26
lauren you're about to get to this as we
63:28
get ready to wrap up to talk about where
63:31
you and i'll just ask this question
63:33
where do you where would you advise
63:35
people that are maybe
63:36
just getting started in business or want
63:39
to know more about
63:40
ai and data what are some resources
63:43
articles
63:45
books that might be just some good
63:46
guidance and a place for them to start
63:48
as they want to kind of learn and and
63:50
develop in this process
63:52
i mean from my perspective i'm trying to
63:54
remember how i learned i mean you know
63:56
uh i i would say i am an advocate
63:59
for uh going back to basics
64:03
and as i mentioned earlier that
64:05
algorithms predate
64:06
computers right i mean the actuarial
64:08
tables of the insurance
64:10
industry you know those are algorithms
64:13
right algorithms are
64:14
on on a whiteboard you can write one out
64:17
you need to understand that power and
64:20
the computer
64:21
and computer science and the scripting
64:23
of it and the language
64:25
becomes just a tool and and sometimes we
64:28
raise up
64:28
our computer tools our computing tools
64:31
to a level that they don't belong there
64:33
we need to understand the concepts of
64:35
the power of
64:36
compound interest for example is an
64:39
algorithm right
64:40
and you need to understand that on
64:41
pencil and paper so i do recommend that
64:43
people study economics and study
64:45
uh you know kind of social sciences and
64:47
those kinds of things
64:48
and then uh learn some aspect of
64:51
technology
64:52
you don't need to be a master at it but
64:54
you should learn a language or to
64:57
learn something about scripting and
64:58
databases and
65:00
and and play around with it there are
65:02
tools that you can download or even use
65:04
in the cloud
65:05
or you can just try a couple of scripts
65:07
out and get a feel for it
65:09
you may take that further or you may
65:10
have learned enough at that point so
65:13
everybody does not need to be a computer
65:14
scientist but that's enough for you to
65:16
start having a
65:17
conversation and understand enough about
65:20
oh i can see how if my assumptions were
65:23
abc
65:24
this algorithm is just going to take off
65:26
and it's gone forever
65:27
so that's the approach i would take
65:30
i want to just double down on this not
65:32
everybody has to be a computer scientist
65:34
i mean i get
65:34
uncomfortable with this sort of pressure
65:36
around stem especially into our
65:38
communities
65:39
um you know it's important and i'm glad
65:41
we're raising awareness but we also need
65:43
to understand that
65:44
we need we need all kinds of backgrounds
65:46
we need all kinds of input and to the
65:47
points that we've been making around how
65:48
artificial intelligence is in some ways
65:50
lowering the bar of contribution because
65:52
you don't have to be
65:53
a phd i think it actually increases the
65:56
demand
65:57
for those art history majors for those
66:00
philosophy majors for people who have
66:01
different ways of thinking that can
66:03
inject
66:04
that thinking into how we're doing this
66:05
so while i completely completely
66:08
subscribe to the idea that everyone
66:09
should conceptually
66:11
understand how software works how
66:14
systems
66:14
work now the implementation detail to
66:16
doug's point whether you know the
66:17
language whether you know the procedure
66:19
to me that's a question of whether or
66:21
not you find that interesting and if you
66:23
do then by all means go all in and i
66:25
will i will champion and support you but
66:26
it is by no means a requirement for
66:28
success in this new world that we're in
66:30
in the society that we're in and we need
66:32
people from all walks of life with
66:34
different ways of thinking contributing
66:36
to this especially now
66:37
this isn't 20 years ago when we were
66:39
just trying to get the first things
66:40
online
66:41
right this is the time now where that
66:42
human intuition that that constructive
66:44
thought
66:45
is just critical so i celebrate more
66:47
diverse backgrounds coming into the
66:48
industry
66:49
and encourage people to find things
66:50
you're passionate about first and
66:52
foremost
66:53
you know stem is just one way to
66:54
contribute i'll close on this but
66:56
there's a workshop i love to run with
66:57
young kids where i bring them in
66:58
and we go through we bring a product to
67:00
market and we from conception
67:02
to how we're going to package it to how
67:04
we're going to build it to how we're
67:05
going to put it into like a best buy
67:06
store it's usually like a video game
67:07
console or a music player
67:08
and my whole purpose of that exercise is
67:10
when we get done we look at it and we
67:11
say you know what
67:12
we spent like five minutes talking about
67:14
what we were going to build
67:15
that was such a little piece of this
67:17
whole thing we talked about promotion we
67:19
talked about selling we picked somebody
67:21
in the team who was going to go talk to
67:22
best buy and make sure that we got that
67:24
great
67:26
setting that we wanted and so it's just
67:28
helping people understand that there is
67:30
so much contribution and so many
67:32
different backgrounds that come
67:33
into this and we should be celebrating
67:35
that and encouraging them
67:38
as we get ready to go i'll just i'll do
67:40
something about
67:43
you told me is like you talked about how
67:45
you look at problems like what
67:46
what problem are you trying to solve and
67:48
what signals relate to that problem
67:50
and how will people be motivated by that
67:51
and i look at it like if you're trying
67:53
to solve
67:53
if you know a problem you're trying to
67:54
solve think of it that way and think of
67:57
it from that level that instinctive
67:58
level
67:59
and not necessarily the technical level
68:01
because i think people get lost in the
68:02
technical and lose sight of the big
68:03
picture so
68:04
you got to do both absolutely so i think
68:07
that's good i know we're wrapping up
68:08
here let me just say uh
68:10
your questions and comments we're going
68:11
to take i think lauren will be taking
68:13
your questions
68:13
you get a chance to interact with both
68:15
doug and will but for everybody we
68:17
appreciate your time
68:18
and check out the podcast disruption now
68:20
as we continue to disrupt common
68:22
narratives and constructs i want to
68:23
thank
68:23
this whole panel for coming you guys
68:24
were awesome thank you really great
68:26
content
68:26
tons of fun thanks a lot thank you see
68:29
you later