Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
Speaker 1 (00:04):
Welcome to Tech Stuff, a production from I Heart Radio.
This season of Smart Talks with IBM is all about
new creators, the developers, data scientists, c t o s,
and other visionaries creatively applying technology in business to drive change.
They use their knowledge and creativity to develop better ways
(00:26):
of working, no matter the industry. Join hosts from your
favorite Pushkin Industries podcasts as they use their expertise to
deepen these conversations, and of course Malcolm Gladwell will guide
you through the season as your host and provide his
thoughts and analysis along the way. Look out for new
episodes of Smart Talks with IBM on the I Heart
Radio app, Apple Podcasts, or wherever you get your podcasts,
(00:49):
and learn more at IBM dot com slash smart talks. Hello, Hello,
Welcome to Smart Talks with IBM, a podcast from Pushkin Industries,
I Heart Radio and ib M. I'm Malcolm Glamo. This season,
we're talking to new creators, the developers, data scientists, ct
os and other visionaries who are creatively applying technology in
(01:13):
business to drive change. Channeling their knowledge and expertise, they're
developing more creative and effective solutions, no matter the industry.
Our guest today is Nicholas Renaut, Senior Data science and
AI technical specialist at IBM. Nicholas's job is to help
companies formulate a data strategy that streamlines the way they
(01:36):
do business and prepares them to use sophisticated AI technologies.
But beyond his day to day, Nick is also a
content creator on YouTube, where his channel has over a
hundred thousand subscribers. His videos explain computer science concepts in
a way beginners can understand, and he often demonstrates how
(01:56):
to use machine learning and data science to solve novel problems.
On today's show, How Nicholas Learned Data science from the
bottom up, the fundamentals of data management, and how an
innovative data strategy can help businesses create novels solutions, Nick
spoke with Ronald Young Jr. Host of the Pushkin podcast Solvable.
(02:21):
Along with being a frequent contributor to NPR, Ronald also
hosts and produces the podcast Time Well Spent and Leaving
the Theater. Okay, let's get to the interview. So tell
me a little bit about how you got into data
and when you found out like the power that it
really harnesses. Do you have a story or anything that
(02:43):
kind of like when you first pique your interest in data.
My first interaction with data and with coding was actually
when I was around about eleven years old, So this
was really just getting started with just looking at spread sheets.
So my dad would come home and after working a
(03:04):
nine or five job, he actually started working with investing
in stocks and doing value based trading that way. I'll
always remember I walked up to his desk one time
and he said, Nick, if there's one thing that you
should learn, I'm seeing all these people work on these
things called macros in spreadsheets, and these people like wizards
(03:27):
inside of my business. I know that you're still you're
still in high school, but I really think you should
learn this stuff. And I started doubling in some Excel
spreadsheets and started just recording macros and tweaking stuff, and
that that's where it all started. But from there, it's
it's always been a recurring vein throughout my career that
I've done some sort of wizardry with data, whether it
(03:51):
be coding or business intelligence or data views. It's it's
always had a bit of a strain throughout throughout whatever
I've done, whether based start ups so YouTube or what
I'm doing now at IBM. Your dad was right. Let
me just say that, because that's someone who's trying to
put together a spreadsheet just to manage my personal finances.
Trying to look up the formula to actually bring a
(04:13):
value from what what uhet to another is enough of
a struggle for me. So I'm glad to do. Really
is it's like it's absolutely is uh so like knowing
that you know, this was how you started getting into spreadsheets.
You know you're looking at stocks and all of that. Um,
can you talk to me about how you found out
the importance of data literacy, how you begin to value
(04:37):
understanding what the numbers meant and what power that could have.
I got a cadet ship at one of the big
four accounting firms and started out as an orditor there,
which is pretty much data focus. So I saw that
these numbers ultimately fed into a significantly big at PA Chill,
(05:00):
which was a formal annual report, and numbers being wrong
in an annual report can move markets, right, those numbers
need to be absolutely bang on. But I think that
is sort of where it started. Where it really culminated
was when I started doing some work at the Reserve
Bank of Australia. And those numbers don't just impact the
(05:24):
metrics for a particular organization, they impact the entire countries metrics.
Getting those numbers wrong on a particular chart, or getting
them right on a particular chart can move entire organizations,
or can shift an entire country. It's kind of crazy
what the value that doing things correctly with data has.
(05:45):
So when you're presenting a metric, you have to ensure
that you are portraying the appropriate message. It's not just
about the wrong number, because correlation does not necessarily imply causation.
So understanding what it is that you're saying is so
so important, and it is so much more powerful now
that we've got so much more data available at our fingertips.
(06:06):
It's really easy to go and grab a bunch of
metrics and go, hey, I'm going to grab this data
from over here, and grab that data from over here
for a measured together. Hey, look, these two lines follow
the same trend. They must be related. Do you find
yourself ever looking at data points and say those this,
How do I don't understand this chart? Why did they
where did they pull this from? Do you find yourself
doing that a lot of your regular life. Oh yeah,
(06:27):
that There's there's some great charts out there as well
that you always see, and they they plot like the
number of Nicolas Cage movies against the g d P
of Bolivia or something, and it's like, well, they're going
in the same direction. They must have some relationship. But
people can really quickly look at a picture and go
and make an assumption about what that is saying without
(06:48):
actually interpreting. Hey, are these on the same scales? Are
they what time period is being displayed? What am I
actually looking at here? And I find myself doing this
more and more often when I just see a child
them like, hold on, let's just not make any assumption.
What is this chart actually trying to say? What is
it actually trying to portray? Because you can lie with
(07:09):
statistics if you know what you're doing. It is they're
so powerful and people can gloss over them so quickly.
We've got attention spans that are so much shorter these
days that it can be very very easy to take
away the wrong message. So you also produce content across
(07:29):
various platforms, including YouTube and your personal blog. UH as
a content creator, how did you get started in that
field and what type of content are you creating? Yeah,
that's a crazy story, right. So I always wanted to
get into tech and said, hey, I'd really really like
to work for IBM. I saw what they were doing
with Watson, and I'm like, why are people talking about
(07:52):
this more? And I had no affiliation with with IBM
at the time, and I'm like, well, this is so cool.
There used to be this thing or or this service
available on the cloud platform called Personality Insights, and you
could plug in a little bit of text and from
that piece of text, it would analyze that particular person's
(08:12):
personality based on the Big five personality traits. And there
actually used to be this demo app where you could
cook it up to a Twitter account, so I could
pass through Oprah's Twitter account or Lebron's Twitter account and
it would actually analyze their profiles. And this is so cool.
It was nuts, and I was like and a lot
(08:34):
of people don't know how to use this. So that
was quite possibly one of the first true toils that
I made on YouTube, and actually used a bunch of
videos that I made following after that too. Finally land
a job at IBM. I actually spammed a bunch of
links in my resume and my coverle that I was like, hey,
I'm already working with this stuff and I could do it.
(08:57):
And the person that hired me, she actually said that
that was like such an amazing way to portray what
what you love about what you do, that that that
had such an influencing factor in actually getting the job.
But yeah, I did it because one the tech was
so cool and I thought it was so interesting and
so powerful, and yeah, eventually that helped me land that job.
(09:22):
So you do a lot of tutorials where you're you're
breaking down complex topics to kind of a wider audience.
Why is that important for you to do? Yeah? I
think one of the amazing things about knowledge is it's
one of the things that you can give away and
never lose, right, And I think one of the trickiest
(09:44):
things about the whole data science and machine learning field
is that it can be pretty tricky to get started,
and sometimes we get hung up with learning from the
bottom up right, And there's nothing wrong with learning fundamentals
and learning foundations and really getting stuck in. But in
(10:06):
order to stick with something, you have to find it interesting.
So if you can see the end result and then
work your way back up and work out how that's worked.
Then it is so much more attractive because you get
that instant gratification and go, hey, I've just built this
machine learning app that is able to decode sign language.
It's so cool. Now I'm going to go and work
out the tech behind it. Admittedly, not everyone goes and
(10:28):
works out the tech behind it, but what I'm trying
to do is make it so that more people can
get involved and get started with it. Lately, I've been
doing these things called code that challenges, and they're kind
of crazy, right, but I love doing them. So I
have to build entire machine learning or data science applications
(10:48):
without looking at any reference code, stack overflow, or looking
at any documentation within fifteen minutes. So it is literally
just a trial by fire. I'll have my phone, I'll
set a time, and I'm like, all right, guys, we're on.
Like the edit is literally just cording NonStop and me
explaining on the go. But it allows people to see
(11:10):
and explain my thought process as I'm developing it. Um.
That's obviously super fun, right because it's highly engaging and
it shows people that, hey, you can get cited in
this relatively quickly. Nicholas is the kind of person whose
passion for data science is so great it spills over
from his professional life onto his YouTube channel. But when
(11:32):
he's not making videos, he's using that same expertise to
help his clients make their businesses work better. At IBM,
Nicholas works with businesses to formulate a data strategy, preparing
them to get the most out of technology like machine
learning or deep learning. He explained to Ronald why thinking
(11:52):
critically about the data it generates can help a company
run more efficiently. So there's a quote that you've used
in your presentation say their firms are trying to become
insights driven, but only one third report succeeding. What is
the role of creativity in the successful one third and
how are you at IBM helping to increase that number.
(12:15):
I remember going to a talk by our previous CEO,
and she said that there's a large number of organizations
that are just experimenting with random acts of digital So
they're just testing out some of these news technologies are
seeing kind of what's possible. But the ones that are
truly being successful are the ones that are getting there
(12:37):
the data ready their data strategy in play. They're the
ones that are starting to collect their data. They're starting
to get it ready and organized. They're starting to take
a look at it and starting to iterate and prototype
and in a structured manner. They're starting to roll this
stuff out. The journey to get something as sophisticated as
(13:00):
machine learning into production is a lot more difficult than
I think people realize because you're now building a box
that has its own rules. You haven't defined those rules yourself,
So how do you explain that when something goes right?
But how do you explain when something goes wrong? And
having governance around that is absolutely critical, which is really
(13:23):
whether the data strategy does come into play. So let's
let's get into more business focused data strategies. Why is
it so important to have a data strategy in place
to fuel AI modeling and how does data literacy play
a role in getting value from these models. We've got
algorithms left, right and center these days, but I think
(13:45):
the thing that people forget is that you can't use
any of these algorithms unless you've got data. So ensuring
that you have a structure in place too one collect
your data to organ is it three, analyze it, and
then or infused to machine learning or deep learning into it.
(14:07):
Is absolutely critical because if you don't collect it, you
can't do anything with it. If you don't organize it,
you can't discover what you've actually got, what the quality
looks like. You don't analyze it, you don't know whether
or not you can trust it. Um and then he
infused is always like the icing on the cake, right
to the machine learning, the deep learning, all the cool
buzzwords that people throw around. That is like the last step,
(14:31):
and it is always the coolest step. But you can't
ever get to that last cool step unless you've gone
through that the hard work that that's come before. Let's
like expand a little bit on the pain points for
companies when they're developing or implementing a data strategy. What
do those pain points look like? Honestly, the biggest pain
(14:52):
point that I see organizations actually the top two that
I see them coming back to over and over again,
and is collecting and organizing their data. So let's say,
for example, you've got a manufacturing type organization and what
(15:13):
they want to do is they want to improve the
production quality on a particular manufacturing line. So ideally, if
they see that they've got defective products on the manufacturing line,
they want to get rid of those sooner rather than
later because they don't want to be shipping him out
to the customer going through the whole warranty and claims
(15:34):
process that just costs a ton of money. So they're like, well,
it would be great to use some computer vision or
some deep learning to detect when we've got defects on
the product line, and then we can grab those and
rip them out. Somebody along the line is like, great,
let's go and do it. The first stumbling block that
you're going to trip up at is, hold on, do
(15:56):
you have any images of defective products from example cameras
that are looking at that production line. So if you
haven't gone and collected images of that or video of that,
there is no way in hell that you can actually
go and build that system to improve your organizational productivity.
(16:16):
So knowing well in advance what data you're likely to
need is absolutely critical. It is the first step in
the data science life cycle. So collecting, understanding, and exploring
your data is the absolute first step. The second one
is a little bit more interesting. So let's say, for example,
(16:38):
you sort of want to get in on the craze
that is data science or machine learning, and you bring
on a data science team. The next biggest stumbling block
that I find a lot of organizations trip up on
is discovering their data. They've got a ton of data,
but nobody knows what they've got. So being able to
(16:59):
find so to discover, rate, review, and rank that information
is paramount because you'll have people come in and go okay.
So a line managers approached me and said that we
want to take a look at our top performing customers
and we want to build a retention strategy so we're
not losing customers anymore. Well, your data scientists is then
(17:22):
going to go, well, do we have data of customers
that have left previously. If you can't easily search and
find out what you've got, that makes it pretty hard
to go and build those models. So collecting, organizing, and
discovering really absolutely critical, but that they can be a
little bit tricky to handle in a large number of organizations.
(17:42):
What kind of supporting technology and new solutions do we
need to meet growing data management issues? It really comes
down to two a few things. So ensuring that you
can one collect the types of data that you're looking at.
So I think when people think of data, they're always
thinking of hate it's just going to be a bunch
of spreadsheets. It might just be stuff that we can
(18:04):
throw into a database. But there is so much more
out there. Right, there's video, how do we store that?
How do we hold that? There is images, there's natural text.
Like we're just talking about ensuring that you've got appropriate
processes in place to be able to store holding catalog
that I think is absolutely critical. We talked a little
(18:27):
bit about data cataloging and the need to be able
to search and discover that data that is absolutely paramount.
Once you've got it collected, how do you find it?
What is IBM's unique approach to facilitating access to data
within companies. So one of the biggest things, and one
(18:48):
of the my favorite things that I get to work with,
is a particular tool set, right, and this tool set
is called cloud Path for Data. So, without getting too
pitchy that the absolutely amazing thing about this is that
those stages that I was talking about, right, so collect, organized, analyzing, infused,
it actually helps facilitate each one of those stages, right,
(19:09):
So you can actually collect, store, and hold your data
in a secure and government place. You've got data catalog
in capabilities which allows you to search. Like one of
my favorite things is that you might have a data set, right,
So I might be a data scientist, and then we
might have another data scientist on the team. I can
have a data set inside of there, and I can
(19:30):
actually rank it and add comments and go, hey, just
be wary of this column with lot certain features that
you need to be mindful of, and that provides additional
metadata understand what is what my data actually looks like
and and things that I should be mindful for. So
I'm I'm Joe employee. How can data be helpful to me?
(19:53):
Great question? So, I mean data is impacting everyone, right,
whether you you like it or not. Um and more
often than not, what you're going to find is that
you can improve whatever it is that you do by
by looking at that data, whether it's let's take an
organization out of it. If you use sleep trackers, you
(20:17):
can begin to see when you're sleep or when you're
getting good quality sleep versus when you're getting bad quality sleep.
If you start to collect additional data points like hey,
am I drinking enough water during the day, am I
doing certain things like looking at my phone just before
I go to bed? Are these things influencing my sleep?
(20:37):
And is that causing a negative impact on my quality
of life? So that's taking a broader view of it.
But when you step into a team or a business view,
data can can make your life a billion times easier.
If you know that there's a particular issue in a
system earlier on in a data pipeline, before something crosses
(21:00):
your desk, you might go and say, hey, look, if
we just changed how we collected these pieces of information,
if we just transformed what we actually did with it,
this is going to streamline my entire workflow and and
help me out. But not only that. Right, So I
work a little bit with the automation team, and they're
really big on robotic process automation. Let's say you're doing
(21:21):
something each and every single day. You're copying a far
from here to there. You're grabbing some information from a website,
You're throwing it into a form, and you have to
do that twenty times a day. There are tools that
can automate that entire process for you, and they're smart.
They're not just looking at where you're clicking on the page.
They're looking at what applications you're opening, They're looking at
(21:42):
what fields you're pulling data out of. You can automate
those entire workflows. That means that you don't have to
do that repetitive, kind of boring work that you don't
really want to do. You can palm that off and
do the robot and do the stuff that you actually
really want to get involved in. As Nicholas said, the
way a company leverages this data has an impact on
(22:02):
every level of the business. Data informs how we do
our jobs day to day and how we plan for
the future. Having an open mindset about data makes it
easier for a business to come up with creative solutions.
In the next part of their conversation, Ronald asked Nicholas
how data science and creativity come together. So let's talk
(22:25):
a little bit more about creativity. We talked a little
bit about your YouTube channel, UH and how you use
that to help people get started with data science. What
does creativity mean to you and do you see your
work as creative? I definitely say my work as creative,
and I think creativity is truly thinking outside of the
(22:48):
box and looking at just different ways of doing things.
I think the biggest thing that I try to embody
is having an open mindset and really in never being
willing to shut something down or not look at a
particular solution or option, because you really never know where
(23:10):
a particular solution might come from. If you look at
where some of the advancements in that the medical field
are coming from, it's because they're being open to new ideas,
new materials, new ingredients, new recipes, new technologies. Having an
open mindset really helps improve that that that ability to
(23:32):
solve complex problems. And I think for me, creativity is
really just having that that open mindset. Tell me a
little bit about how you approach novel problems. What do
you do when you get stuck? I think the most
important thing I really like when I push myself to
do something that I've personally never done before, and a
(23:55):
lot of the time that yields new solutions to problems
that that that might be really difficult to solve. It
doesn't necessarily need to be using this particular set of techniques.
It's what else can we do to solve this problem?
And sometimes like it'll be staring you in the face
and you'll just have no idea until you go, hey,
(24:16):
I'm going to throw everything out of the box and
just give it a crack and see what is possible. Um.
But sometimes it does require that that little bit of
grit to to push yourself to see just what is possible.
And I think that's when I've come up with some
of my favorite things that I've ever done, so something
(24:36):
that I'm trying to adopt in my in my daily life.
And I'm reading a lot more about stoicism and philosophy,
and I'm seeing that you kind of really just got
to push through sometimes to to see what what's on
the other side. We talked a little bit earlier about
how folks can take bits of data and kind of
(24:56):
tell their own story with it, especially if they if
they know the story that they're trying to tell. But
let's talk about using that for good. How does creativity
play a role in data storytelling. I think there's just
so much good that you can do with data that
if you have that in your core ethos, then the
(25:20):
world's your oyster, right. I always come back to my
favorite project that I've ever done, and that was using
computer vision to try to decode sign language. It is
by no means a state of the art model. But
I forget hold on why is never nobody ever approached
this or at least shared how they've tried to do it.
And I've kind of just had to get real creative
(25:40):
and trying to build that I had. I literally spent
weeks just trying to install stuff then trying to get
it running on my computer before I even got anywhere
near building that particular model. And and it's super hardcore
in terms of trying to get it set up. But
there's so many opportunities for good, whether that's improve accessibility
(26:01):
to certain technologies, improving the quality of life for people
that could benefit from us using data a little bit better.
There's a large body of work with a bunch of
different data scientists where they're actually building language translation models
for languages which aren't hyper popular or aren't as widely
(26:23):
spread as we might see in our day to day lives.
If you look at India, there are a turn of dialects.
If you look at even where my parents from Mauritius,
there's there's a whole completely separate dialect where if you've
never heard it before, you were like this just slang French,
(26:44):
but no it's it. It's like um, it's its whole
separate language that obviously allows or improves the ability for
people to to tap into data and do a little
bit of good. But there's so much I mean, people
are using medical image data to improve medical segmentation and
improve diagnoses. There's just so much amazing work that that's
(27:08):
happening in that space. There is obviously the temptation or
used data for bad, but I'd like to think that
the large majority of the community are really trying to
use it for good. You started talking about a little
bit just now, But what are some future trends and
challenges and future topics or projects you're excited about anything
(27:29):
in particular looking real further forward, what I'm super excited
about And I still don't know how it's necessarily going
to impact me, whether or not that's going to change
my experience as a developer or not. That we've got
quantum computers coming right, there's a ton of work that's
happening in that space. It's going to radically shift how
(27:51):
large a machine learning model we're able to create, how
fast we're able to train them. I'm just excited to
see what happens in that space. I'm not a quantum
physicist by any means, but I'm still excited to see
what I'll be able to do with him in the future.
I love that, as you'all contingent filt this technology, you're
(28:12):
excited to play with it after it's built, which I'm
I don't want to have to build it. Nicholas for
not thank you so much for a talk of me today.
It's been an absolute pleasure. Thank you so much for
your insightful questions. It's it's been awesome. Ronald Nick made
a point that I think is important to remember when
(28:34):
it comes to technologies ability to improve our businesses, or
make our jobs easier, or even do social good, A
thoughtful data strategy is always the first stepping stone. Without
good data, using machine learning or artificial intelligence to create
innovative solutions becomes much much harder. Our technology gets more
(28:58):
sophisticated every day, but that doesn't mean we should lose
sight of the fundamentals. If we want to get the
most out of smarter technologies, better business decisions, more optimized technology,
fresh and unexpected insights, we're going to need smarter data strategy.
On the next episode of Smart Talks, with IBM the
(29:21):
power of Salesforce to transform the customer experience. We talked
with Phil Weinmeister had a product for Salesforce America's at IBM,
consulting about transforming digital experiences with the power of Salesforce
and IBM. Smart Talks with IBM is produced by Matt Romano,
David jaw, Royston Deserve, and Edith Rousselo with Jacob Goldstein.
(29:46):
Were edited by Sophie crane Are. Engineers are Jason Gambrel,
Sarah Brugare and Ben Holliday. Theme song by Granmascope. Special
thanks to Carlie mcglory, Andy Kelly, Kathy cal Hand and
the eight Bar and IBM teams, as well as the
Pushkin marketing team. Smart Talks with IBM is a production
(30:07):
of Pushkin Industries and I Heart Media. To find more
Pushkin podcasts, listen on the I Heart Radio app, Apple Podcasts,
or wherever you listen to podcasts. Hi'm Malcolm Gladwell. This
is a paid advertisement from IBM.