- Welcome to this addition of the Adobe Think Tank.
I am Daya Nadamuni, with Adobe's Corporate Strategy Team
and with me here today is Chris Benson.
He is chief scientist for AI, artificial intelligence,
and machine learning with Honeywell SPS. Welcome Chris.
- Thank you very much
Great to have you here.
- It's an honor to be here.
So, okay,
so what do you think
CEOs and other leadership should be thinking about
and collaborating on around AI,
at this moment in time?
- Well we're at a very special moment,
in terms of what technology is doing to business.
We really need to change the way we're
going to serve our customers and produce the products
and services that we need to fulfill
the customer experience that they're expecting from us.
Doing that we have this moment where we've gone through
the digital revolution and as part of that
we're hitting this point where AI has finally,
after decades, become a reality.
And in doing that, it is, the speed of business
is accelerating at such a rate,
that we need to start changing
the way that we're approaching it.
I'll give a few examples here.
We suddenly have found ourselves,
after many false starts in AI,
to where we have more data than we know what to do with
and it's growing exponentially all the time.
We call it big data.
We also have compute capability that is in the cloud,
it's out on the edge and so there's the
ability to do calculation that's never been there before.
And so suddenly we have this ability to take
this new tool set, which is a bunch of algorithms
that we call AI and with some specific tool kits,
things like deep learning, that we can now apply
to customer experience and get pretty amazing outcomes
compared to things that we've done in the past.
And so, this is a moment where senior executives
at companies big and small need to be thinking
how can they use these tools to better
enable those customer experiences.
One of those is to recognize that
we're in an age for the first time ever, where
we don't know exactly what the next step is,
what the future holds.
It's moving too fast and it's changing too often.
So it really requires that you organize your
company around a learning structure.
So you need to be a learning organization,
you need to be able to think like a startup
and fast fail your way into a great path
for the customer experience that you're trying
to create for your customers through
your products and services and so I'm always
urging people to, where ever they're at in
their career, whether they are fairly new or old,
to take their cup and pour it out
and accept that they need to refill it.
And it's not a one time thing, it's something they
need to start doing more and more
often as things accelerate.
In doing that find the best customer experience
that your customers need and how to apply
these specific tools for that.
Given the need for continuous learning,
and the fact that we don't know what,
you know we can't see around the corner
quite as well as we would like to.
What are some best practices around
integrating, you know, deep learning
into your enterprise architecture?
What advice would you have for brands and enterprises?
- Yeah, well one thing is to keep it in perspective.
There's a lot of hype around AI
and deep learning right now.
It really, it's a specific tool box that
is very good at specific problems.
And so the first thing is to recognize is,
that it's not magic and it's not the
solution for all problems.
And you need to begin evaluating
your business challenges against
all the tool kits in software development
and data science, in general,
and find the ones that are,
that lend themselves to AI and deep learning solutions.
Which tend to be very complicated problems
that other tools have not been sufficient
in solving for you, and so–
Then you, kind of,
rapidly experiment your way into saying,
Do I have an AI tool in my AI toolbox
that can solve this problem,
where as I may not have been able to solve it before.
Or I can solve it, at much higher fidelity.
So it's a very practical and pragmatic approach
instead of a big buzz word, you know,
magic kind of approach.
Think of it, think of your deep learning team
as another software development slash data science team.
They are going into a project, they evaluate it,
make sure they're the right team for it.
The tools that they think are appropriate,
are in fact and you validate that.
And then you take a small bite out of that problem
and go try and solve it with this
and see what you get.
One of the challenges, you don't want to try to go
make, take too big a chunk, like anything,
even outside of AI, too early.
Take small things, when you have quick fails,
learn from them, find the right path,
and find your way into success.
So be very practical, be very pragmatic
and that's how you'll find success in it.
- So, which brings up a good point.
Are there specific classes of problems
or types of problems that would lend themselves
better to deep learning than something else?
- There are, so and some of them are
being widely used right now, I mean,
some of the very obvious ones are
natural language processing.
That field has been around for a long time,
but deep learning is starting to revolutionize it,
in that, there are–
there is a lot of variability in it.
There is a lot of nonlinearity in it.
These tools are really good at
solving problems with those characteristics.
Another thing would be machine vision.
There is a type of AI called a
convolutional neural network,
which is very popular right now
for driving machine vision solutions
and the reason is, is that neural networks
are really good at finding hard
to detect patterns in data that
human beings or lesser data science tools
are not so good at finding.
And so if this sounds like a problem that
your company might be facing then say,
okay well we tried it with smaller data science tools,
things that are not quite as sophisticated,
maybe this the right time to try a neural network
and see if that works for you.
- Okay, so find the right solution
to what you want to apply or find
the right problem to what you want apply
the deep learning solution?
- That's correct.
- I think a lot of it is, sort of, customer facing,
you know large volumes of data,
where you can really try to solve
the business problem.
- Absolutely
- And improve the customer experience.
- Absolutely and specifically looking for those
problems that your other tools,
you have tools in you tool belt if they have not
been sufficient to solve it,
that's when you want to turn to neural networks
and say, is there an architecture here,
within this world that might help me get there.
- That sounds, very intriguing and full of possibility.
- Yeah
- Especially for brands and enterprises.
- There is an investment you have to make.
- Right
- To be able to accomplish that
and that because it's computationally intensive,
there is a lot of math, fundamentally
it's all math under the cover.
So you need to be able to have people
that understand that, train them up.
Whether they be existing data scientists
or mathematically inclined software engineers
or whether you're bringing people from the outside.
There is a huge demand for this
and you also need to get your infrastructure in place.
Not only for the calculation itself,
which might be in the cloud
or it might be a super computer that's
designed for deep learning
or some other
collection of computing devices
that you've put into place for this.
But, beyond that you need to get your
data infrastructure as well too
because data strategy, which may evolve with that,
is pretty crucial.
I think if there is one way to finish this is up, it's that
I've seen companies that are successful
in this space, really focus on a data strategy
that supports their AI strategy
and that can evolve with that AI strategy.
They figure out how to get all that data
they're gonna need for AI training
in place and usable as, before they go into the AI training.
- Perfect.
We talked a lot about data strategy
and we talked a lot about algorithms and machine learning.
When it comes to creativity, what's your optimism
around some of the skills,
that are pretty much human,
and irreplaceable, no matter how advanced
an AI algorithm or a machine learning algorithm could be?
- Well you've just identified what,
maybe, the most important one and that is creativity.
And if there is–
Computers, AI driven computers
can be very good at attacking
problems that are very task specific.
We've seen instances in the news
where you'll have a very specific type
of thing where a computer gets better at it than a person
or a computer and a person together, you get better at that.
The place where I think humans will have a firm
place of there own, without computers, is in creativity.
That's not to say that we will never see
facets of creativity in computers,
but that's something we're really good at.
Making those associations in our brain,
our memory system, it's something that is,
going to take a long time for AI to ever get to.
- That's fantastic.
Well, we're at the
end of this episode of the Adobe Think Tank.
Please tune in again for more episodes.
Follow us on twitter #abobett.
Thank you Chris,
- Thank you very much,
- It was a pleasure to have you here.
- It was my delight, thank you very much.
For more infomation >> Will the May jobs report trigger a shift in Fed behavior? - Duration: 5:11. 
For more infomation >> Cute cartoon cats have a black twist that you definitely will not see (part 3) - Duration: 3:35. 
For more infomation >> HAPPENING: Trump's Next Pardon Will ROCK Obama to HIS CORE!! - Duration: 2:51. 
For more infomation >> Will the Russell 2000 be impacted by Trump's tariffs? - Duration: 2:00. 
Không có nhận xét nào:
Đăng nhận xét