Welcome to this edition of Adobe Think Tank.
I am Daya Nadamuni with Adobe's corporate strategy team.
And with me here today, is Susan Etlinger,
industry analyst with the Altimeter group.
Welcome Susan.
Thank you.
It's great to have you here.
Oh, it's great to be here in this beautiful little space.
Oh, and I'm so excited to be talking to you about,
you know, what AI means for digital strategy
for the customer experience.
Yeah
So , would you tell us a little more
about how you're thinking about it?
Sure, so I come to this from,
sort of a layperson's perspective right.
I'm not an engineer, I have a background in humanities
and I've been doing research into data
and sort of how businesses use data
and now increasingly, algorithms and artificial intelligence
technologies for the past few years
and what strikes me is that, you know,
when we, now that we live in a world where we have,
in many cases sort of algorithms,
machine learning algorithms that sit between
an individual and an organization,
it really changes the experience.
And so, in some cases, we've gotten used to it,
it's not that big a deal.
Like, you think about search or you think about
a recommendation engine, you know,
if you buy something on amazon.
You're used to that, even predictive text in our messaging,
in our messaging tools, like we're kind of used to that
and we laugh about it but we sort of know it's there,
we sort of don't really care so much.
But now that we have tools like
alexa, voice agents, chat bots,
obviously autonomous vehicles coming.
There's this real difference in the way that we interact
with technology and in many ways,
interact with products and services
from the way that we used to do it
and so one of the things that I started to think about,
especially as regards to, you know,
voice and anything that kind of has to do with language,
is that now we're dealing with an entity
that may or may not be human.
And we don't really have ways of knowing that sometimes.
And so that there's a need to be transparent.
Or we're looking at something that looks
kind of a little like that vase on the table
and we're talking to it but we don't know what it knows.
We don't know how much it can help us
we don't know what it's able to do.
We don't know what it know about us.
And so that really kind of creates a few different things;
one is an opportunity to think about design differently.
To think about like, how do you design for
a space in which people don't know what they can ask.
It also creates an opportunity to think about
like how transparent should we be.
You don't want to over explain
and stop the experience and kind of like,
clutter it up with stuff and at the same time,
you don't want to do something
where you surprise somebody in an unpleasant way.
And so, then it gets into questions of
kind of ethics, right?
And ethics are simply just accepted norms of behavior
or you could say, of moral behavior.
But let's just say, in a business context,
accepted norms of behavior; we don't really have
accepted norms of behavior in an AI world.
And so, we have to ask the question;
should it have a gender?
Should we know that it's a robot?
Should it be able to make decisions on our behalf
that are different from the kinds of decisions
that were made in the past?
So that's kind of where this all started to boil up for me.
Well that's super interesting and you're absolutely right
I mean we think of it as Alexa or Siri.
So we personified these intelligences
that we don't stop to think about what's actually behind it.
So, you bring up a very good point about ethics and bias
I mean, there are a lot of concerns around, you know,
how do we, not only design the best possible experience,
but do it in a way that's ethically right?
- Yeah and so this comes out in interesting ways.
So I have a little bit of an addiction
to MIT technology review and I read,
pretty much everything I can get my hands on that comes out
and there have been these situations
that are very well known among sort of,
data scientists and machine learning technologists
but not necessarily in the rest of the world
about things like, you know, the data set
that's used to train translation algorithms
and search algorithms.
It has biases in it and it's not that
the algorithm itself has a bias.
It's that it absorbs all the biases
that occur in the culture that it's been trained on.
So if, for example, we live in a world
in which more women are nurses than men are nurses.
If you're a mathematical model, probabilistic model,
you will then infer that it's more likely
for a nurse to be female and you could
actually mis-gender somebody that way.
Right.
If you are training a voice agent on the kind of voice,
for example, that I have, or in England, let's say,
you know, BBC english, it's less likely
to understand accents that diverge from that.
Will it understand someone from Glasgow?
Will it understand an African American,
you know, somebody who speaks African American
vernacular English?
Will it understand somebody from south Asia,
you know, who's learned English
maybe as a second language or speaks English
but not as often as they speak their first language.
So there are all sorts of different ways
in which we have to be really careful
not to inadvertently disenfranchise people
because of the fact that we think of math
as being inherently neutral.
- No, and that's very true because the unconscious biases
can just slip in and you don't know
until the output sort of gives you
the wrong results or something like that.
That right, that's right and there have been stories.
There's stories about, you know,
gender bias, there's stories about race bias,
both in computer vision and in voice agents
and you know, when you think about it,
one of the challenges here is...
And we've seen this with Facebook
and the Cambridge analytic stuff
where for example, if you are not receiving an ad
because an algorithm has made an inference
that you are not the right audience segment,
the right target for that ad.
But that ad is maybe for a low interest credit card
that would actually be an economic boon to you,
like, that's not fair.
If the person sitting next to you then got that ad.
And so we have to think about issues of fairness.
And that doesn't necessarily mean
we're going to kind of prevent
every bad outcome from happening
but what I'm arguing for is; we have to be realistic
about what we live in and we have to be very intentional
about the world that we want to create in a digital space.
Absolutely, absolutely so we've talked
a little bit about the customer experience.
I'm also curious to hear your thoughts
on how AI enhances our productivity workflows
or business workflows.
Yeah I mean there's some really interesting stories
about this.
So, one of the great things about AI and about math is that,
and about computers and processing,
is that they can just take in a ton more data
than we can as humans.
Even the math savants in the world, and process it.
And so, sometimes they can see patterns that we can't see.
Those patterns could be patterns about
like the kind of actions that precede
a conversion event in a commerce process.
Or it could be a way to shorten
an account opening in a financial services company
from 15 days to two days.
Or it could be a way to find fraud or money laundering
or it could be a way to find a tumor
10 years before it becomes a tumor.
So, like, there's just a tremendous, you know,
yes there are challenges and yes there are risks
but there's a tremendous, I think, social and business
and economic value to these technologies
and it's sort of that old Spiderman adage;
like with great power comes great responsibility.
But we shouldn't also forget that there's great power
and potential in these technologies too.
So if I had to ask you the top two or three things
you'd advise brands and enterprises
to keep in mind as they're thinking about leveraging AI
for their digital strategies, what would those be?
- So I would say, that artificial intelligence
and you know, I'm using the term sort of loosely
because if anybody was listening to yesterday's live stream
they probably know that it's not the easiest term to define.
True, true.
Even if you get a bunch of exp...
Especially if you get a bunch of experts in a room.
But what I would say is that it really...
This is a massive shift and you know,
if we're lucky in our lifetime,
we get to live through one massive shift,
I think this is at least for me, kind of two.
Right, there's the internet, which dramatically changed
the relationship between companies and people
because now it gave people a voice
and now there's this sort of intelligence shift,
you could call it the fourth industrial revolution,
you could call it whatever you like but...
Which now enables organizations to learn from
the data that they have, the experiences that they have
at scale, right?
You have individual learning but now we have
the ability to learn in scale.
So I think the first thing is,
remember this is a massive shift.
It does require culture change.
And while it's perfectly good to start with
proof of concepts and build from there.
I think it's important to have that long term perspective
in the back of your mind.
That's the first thing.
The second thing is; we really are moving
from a rule of space world to a more probabilistic world.
And that also changes the way we set KPIs and measurements
and the way that we set values.
And so precision and accuracy, which are the thing,
you know sort of the six sigma things
that most companies really want.
That's almost not possible in an algorithmic world.
Absolutely, there's a lot more...
So this is a big culture shift, I think
and mind shift for people.
And the third thing I would say is,
never forget that you have to just
get started and be pragmatic, right.
Not every company has the resources
of a Facebook or an Amazon or a Google
or a BuyDO for that matter.
But you can get started with a proof of concept,
with a project that is focused
on something that really matters to you.
You know, don't do something cool in vanity
unless you wanna win a Cannes Lion
you know, that's perfectly good.
But if you really wanna move the needle,
start with trying to optimize a part of your business
that isn't working as well.
That's great, well we're out of time.
So thank you everyone for tuning in to this livestream
of the Adobe Think Tank.
So follow us on Twitter, and tune in again
for other episodes.
#Adobett
Thank you Susan.
For more infomation >> Reverie - Script to Screen (Digital Exclusive) - Duration: 3:01. 
For more infomation >> Agencia de Marketing Digital en Lima - EN VIVO - Duration: 1:50. 
For more infomation >> How I Do My Digital Art / Medibang Tutorial - Duration: 6:07. 
For more infomation >> Digital Dive: Roseanne blaming offensive tweets on Ambien - Duration: 1:42. 
For more infomation >> Sacramento Part Of Pilot Program For Digital License Plates - Duration: 2:22. 
For more infomation >> Digital License Plates Could Track Your Car's Every Move - Duration: 1:41. 
For more infomation >> Western Digital Green SSD - Unboxing - Duration: 3:12. 
For more infomation >> David López - Presentación del Master en Digital Sales - Duration: 2:46. 
For more infomation >> Subvencions per a projectes de transformació digital del comerç de BCN - Duration: 0:57. 
For more infomation >> Sacramento Becomes First U.S. City To Test Digital License Plates - Duration: 0:21. 
Không có nhận xét nào:
Đăng nhận xét