All language subtitles for subtitle1 English

Yeah.

Hi. Hi. Yeah, nice to meet you. Nice to meet
you as well. Yeah. I'm John Kim. Yeah, I'm the

communication leader in today
is the first session.

Talk about our life and our interest. I'll
do my best. Yeah, cause you introduce yourself briefly.

So my name is Chris Ha. I'm currently a capital
fellow and I'm part of the Cat Digital, which is a

business
unit that.

Owns and manages a digital product. Yeah, within
the Caterpillar. Yeah. And I've been working in

Caterpillar for 22 years, 20 years and before
that, three years at the utility company called Duke

Energy. And currently I'm I live in Champaign and I
have a a wife who works at the university as

well as a
daughter who's in college.

One medial stay in character for 23 years.
Have you thought about our jobs at all?

Pardon me, Have you, have you thought about,
like, getting another job? Yeah, sure. I mean, I've,

I've,
I've.

Has some opportunity to explore, yeah, other things,
but the frankly Caterpillar has been giving me

a lot of great opportunities in a lot of
different area and most of the project I was fortunate

enough to to
to feel about.

That
uh.

Very, very passionate about the
work that I've done. Yeah, so.

Although there might have been you know some,
some sometimes that other opportunity, I decided to

just to stay at Caterpillar because I felt like
it's the right, you know, choice for me and my

family as well as my career. So I've been very
happy with the Caterpillar so far. Yeah, so far I am

very happy to. I kept trying to calculate last year,
last year and but I think it's it's a great

company and it gives me
a ton of opportunities now.

So let's talk about
what you do in cab.

So I am a a technical leader within
CAD digital, yeah, and my job is to.

Provide thought leadership
in some of the.

Enterprise big initiative within the for the
for the enterprise such as the condition monitoring

electrification and more recently
in artificial intelligence space yeah.

So I made my you know role is that individual
as an individual contributor is to to to ensure that

that we're providing
solid and robust technical.

You know technical, technical basis for our digital
product as well as to explore maybe the latest

AI technology where
we can provide.

You know, a
good, a solid.

Analytics, yeah, for to to
gain, you know, business value.

By applying them to a real practical Caterpillar
applications. Yeah, when I was looking for a job

opportunity in Kiev, I came across some apps
developed by Cal which is called Starlink and Master

Solution. Those two apps were
developed by you and your team.

Not, not really. The the the mind star that you're
talking about is by the. Actually the the IT was

a.

To currently ICS, the integrated component solutions
division as well as the resource industry

mostly we have a team in Australia, yeah that that
does this. The CAT digital is not the the one

that's behind it. However,
we do have some.

UH, projects that we're going
to be collaborating with the.

The autonomous and and automation group which actually
owns Mindstar technology as well as the the,

the the commercial group and the MRI side of
it which is resource industry. So we have some

integration and collaboration we hope to do in the
future, but it might start was not part of the

CAT digital I was
into actually being into.

In terms of still myself till I am a
stock investor and everyone was talking about like Tesla

autonomous driving, so
compared to those?

The big companies, right, How far can is behind
or ahead like in terms of technology? So I'm not

the expert at it, but I know
that we've been using, you know, autonomous.

You know, perception. Yeah. Type of computer vision
technology for a long time, yeah. And we've

had, you know, the, the big mining truck
at the mine site automatically, you know, driverless

mining truck. It's been, you know, doing for a
long time and millions of tons of, you know, the

payload
has been.

Carried by those those mindset we have.
So from that perspective you know automation autonomous

truck especially has been you know in a Caterpillar
for for a long time now. You know the the

technology wise that that might be slightly different
but I'm sure that we at least from our

industry perspective not the moving industry, heavy
machinery industry we're we're one of the best

if not the
best in that area.

So currently like yeah, so we make some profits
by selling the software where it's just well the

mind star is actually a software system as well
as hardware, right. So it's a hardware with the lot

of hardware that has to be installed and but we
also have a software that supports it and we have

they have an entire team that has to be
dedicated to, to support that both at the the dealership

and customer level as well as the
team that that supports that from the.

Mostly from our eye side of it then put
it back end technologies developed by I would think that

ICS side. Yeah. So we don't sell the software as
it is that we sell it with the machine. And yeah,

I think you have to have hardware in order to
solve the software. You have to have a lot of

equipment right to perception, you have to actually
have to you know the Lidars and all those

things installed and those things. So those other those
other the must have and then I'm sure we we

do also have as a part of the
package right there you have to you have to.

The software, but I don't know whether we actually
sell as a software by itself without any type of

a support, right? Yeah. As I understand it,
Tesla in those other multiple companies, they said it

is a subscription basis. So to use it like
you have to subscribe to the system service. Yeah, this

one, I'm sure there is a subscription.
Mindstar is also various different things. There's mindset

command, there's my * fleet, there's Mindstar
Health. So it's not just the one.

Nine star, there's a different level of different things
than my star can do. So yeah, but that's

not really my area of expertise. So what was
your flagship like application that you did? So, OK,

so those are what we do right now. The
condition monitoring side of it is something that we do

analytics
to.

Ohh
find out.

Whether there would be
anything wrong with the?

Asset in advance so that you avoid downtime with maybe
we can sell parts in advance or we can we

can reduce unplanned you know maintenance So. So we
try to predict failure in the future so that we

can you know ask our customers to to plan ahead so
that they don't have to to to to suffer with the

downtime because especially in the you
know resource industry the downtime is the.

The
key.

And better, better manage their their machine. Yeah,
the uptime, right, that that you don't want

all of a sudden machines to go down so that
you can really do the work, right. So you see some

symptoms right from the machine because machine has a lot
of sensors and what we try to do is to

warn them, you know, ahead so they can
go and fix it before we actually break down.

Uh, I came across the news article on the demo
like UH, you are getting a huge reward and and

something about data analytics and what was it
about. I read articles but I didn't quite understand

what you actually did for. Yeah. So I think
the the word that you're talking about is an annual

award. It's called the Business of Data Symposium
and they're given to innovative idea that was

done by the
North American analytics.

Yeah,
and.

We had some stiff competition such as a Verizon,
but it was Verizon yeah, yeah but but that thing

they're like four
other for their.

Companies that were nominated for that, but the lot we
were lucky that we got the word and we were

very proud and I did analytics, analytics, but
this particular one was about a what's called ACE,

an analytics crowdsourcing environment. So when people
want to do analytics and there are many

different people that has to collaborate. So it's
more like a contribution model of a different

people creating the analytics
model and able to quickly.

Validate.

And deploy and we created this environment so
that the time that start from creating the model,

testing the model then deploying it usually take months to
do it. Now it's it's in, it's in days or

weeks, yeah. So we had a huge improvement
and and streamlining that process the collaboration. So

it's a it's a collaboration especially also with
not just among the data scientists or engineers.

That the data scientist has to get a feedback
from the subject matter expert. But these people may

not be coders right? So we can just work
with them in an environment where engineers work in. We

have to have a very a collaborative environment
like website. So it's a web application that they

can work on. So that when when developer creates and
try to test it then you get a feedback from

the.

The subject matter expert right away and and you.
It's more of a both their validation tool as a

communications tool and it had improve our workflow process
by 5 to 10 X fold. I love always using

the Microsoft Teams when I'm working with my
coworkers, so this system that you're talking about

is.

All that different from the Microsoft Teams that we
use. Yeah well Microsoft team is a very good

collaboration tool but this is very specific to sort of
analytic side of it. So, so in order for if

you're developing some sort of software or some sort
of a coding, right and then let's say the

Python base and and lot of people work on a
Jupiter notebook and and and they when they look at the

answers and they
they try to create.

Another code or try to script something to visualize the
data tab to you know if you have to. I

don't know whether your software engineers, but when
you have a Python code that's a 200 lines.

Like 90% of them is just getting the data, how
to get the data and show the result. But the actual

logic, Yeah, right. And so that logic is only
10% of the time, yeah. So we're trying to avoid

having data scientists really focus on the one that's
a non value added work and have them already

set up so they can focus on the just
the larger part of it and visualization. And the data

extraction is already sort of a with the button that
you can get. But at the same time if you look

at the the
result, right, it's not.

That you as a data scientist or engineers has
to validate, it has to validate by some other people

like experts, right. They're not necessarily the corner.
So if you give them Jupiter notebook to do

or some sort of static picture to do
that's really not integrative or inter interactive, right. So

they they can come in through a website link
with that they can provide the feedback and yeah so

that type of a
very easy to use tool.

But also with the people who are not expert on
the upper right, we're not expert in the in the

coding, they can come in collaborate. So that process
has been worked out really well. So it used

to be taking a lot of time. Now the
the workflow has been improved by 5X to 10X.

So you yourself or like programmer coder and I don't
anymore but yeah I used to do a little bit but

it's we have a much smarter once you're younger
generation who are better than me right. So I'm

more of a provide them guidelines and maybe
directions and and you know provide my expertise and

products and and and and the business side of it I
used to be 1/4 too like I used to program and

and there was one
problem I was really.

Out of but after a few years, if I look at
the code that I wrote myself I don't understand. I know

my code. Yeah. Yeah. These days you know why
things are all different. Syntax is different than we

have lots of language that that's more of a
higher level language that you know people don't have

to really write a lot of lines. So we have
a lot of you know younger generation who can definitely

do the better coding but so we just
have to hire the right people. So we.

That digital, yeah, the the we have a
lot of software engineer, application developer et cetera. So

programming
skills.

One of the fundamental skills in computer science
and those IT industries, but these days with the

AI and like
the automatic codings like.

And it's not necessarily you think that in
the future like it's the AI intelligence artificial

intelligence really wipe out or intermediate programmers are
well and I know that you can ask

ChatGPT for instance to write a simple
code right. So there will do that.

Thing that you know the simple code like that can
be done, but when you wanted to do a really

professional level right, production level code that there
always has to be at yet right that the

person has to sort of
take a look at the.

As to how those things are done, but I I
guess the the process can be improved, right? Process can

be yeah like translation for instance. Like if you were
to have a chat chippy to do 90% of the

work, but person to chat, you know what the error
is. That would be much faster than have a manual

person a person do it manually right? So it it
would be similar type of thing. Things like AI will

always be trying to look
to to to to replace.

Uh, but I guess not entirely though. Yeah, not
entirely, Not entirely. I think we still have the

change
management.

Nothing is is is a big deal from the
people perspective even even if let's say you've mentioned

about the
Tesla, right? So.

There's possible that everybody can just to ride on
a car and they will do automatic but I don't

think that people will become most people will
comfortable doing that right. Yeah. So you still

have to be override by the the the manual
and that might take a long time to actually everybody

even though technology might be there and and has own
era of a 99.99% of the time they would be

safe but I don't think people will adapt to it
because it's more of a how feel, you know human are

comfortable with
it so similar.

So AI, AI technology might be already there or
it's it's getting there that's going to take a

little bit of time,
but what's important is that.

Especially our younger generations and and as well
as us that you know markets changing all the

time job markets right job and and we just have
to be aware of the all the changing in a much

faster pace these days how to be aware of the
change. Yeah and then we have to to start leveraging

these tools. Yeah that's because that's who has somebody
has to leverage it right. So if you if you

keep up with this thing and always try
to reinvent yourself of of using the tool.

Like this, I think you'll be OK if that's that's
my opinion. We don't have to actually have to be

programmable, but you know how to use the
program, like like checkpoint, like we use Google every

day. Yeah. So knowing
how to use the program.

And not knowing is makes makes a huge difference.
In the future, just yeah, know how to use the

tools. I
agree, yeah.

Yeah. Have you used the CHP? Yes, we actually
have a have a couple of different projects that we're

trying to. So our executives right are very highly interested
in chat JPT and we I happen to be a

part
of the.

A ICOE Center of Excellence team. So we
are looking to to leverage something like ChatGPT for

Caterpillar application, but there's
a little bit of a.

Caviat that the chat deputies are very general
tool, yeah, leverages most of the you know, public

information to train itself. But if from the
industry perspective, you know things like AI, they

wanted to use it to to gain competitive competitive
advantage, right. Which made in order to do a

competitive advantage that you cannot just do use
any public data. You have to start using your

private proprietary data. But you also
don't want this to be get out.

So that other people can
use it. Umm, so that.

Protection of IP and and and legal issue and
trained model to be staying in Europe sort of a

industry in order to make sure that they
have a competitive advantage might be the one that

industry focused on So which means is that legal
people to people has to be involved in in doing

this thing and and usually when legal teams involved
it takes a little bit of time to make sure

that everything is protected So while the technology
might out there while the use case might be

identified
that.

It might take a little bit of time for it
to be really productionized and and and make the real

impact.
Hmm.

So I was very curious, right, like when we
think about the search engine always Google is the king

of the search engine industry like it's always,
but how come Microsoft suddenly like well Microsoft

is an investor of this opening, right. So this
is just the other companies are also doing this

thing, but it's more like a think like a the
human. So a lot of their conversational type of things

and and I think the interface that
we're going to be doing from here.

One would be more like a human like rather
than just type in a certain word or moral conversational

like. Right. So I think what's the what's
the maybe difference between what Google did with the

this generative AI which is the chat JPT is that
there would be a lot more human like I user

interface. So that that's what they really, really
like. Right. So yeah you you, you know the

little bit of nuances of if you
think about like even translation for instance.

You know is a French French is a little
different than you know the Canadian French. Yeah if you

wanted to to to write a story but from a, you
know a more of a a children type of story might be

different than you know adult type of novel, right.
I think this type of a nuances is something

that the chatty PT can can can do anywhere where
the human can you know experience a little bit of

that
interface.

And
even.

Even for now that you you've been, you've
been experiencing with the, you know, the, the, the.

Like
a?

What is that Amazon. Amazon. Yeah. The the how
they ask questions like saying things like that so a

lot of interfaces more like this conversational stuff
right. So they're Google the the search you

know the if you really explain about the long
sentences that that I don't think that's how people

usually search it but with the
chat T right that you you can.

Even write the entire like two or three sentences
then it'll still process it and try to give you

the best answer. But
what's important that it can.

It's still based on the training
data set. I know there are some.

Different things that they can actually learn from. Yeah,
but still I I had the other day I ask

about myself, right The Who is correct. Chris
Hyatt Caterpillar and that gave me, you know,

completely wrong answer because I don't have a
lot of public information that's out there for chat

TPT to train about me. So you just need
to be careful, right. It's not there's some amazing things

to chat you can do, but a lot of times
also it also gives you completely wrong answer or may even

lie.

About it yeah. And and or no answer. So it
it, you know right now it's any technology that goes

through what's called hype curve. Yeah, so right now
we're at that peak of the hype or getting

getting there as a hype. So they're all surely
be like a you know unexpected type of a.

The I'm sorry the unreasonable or expectation there
will be there people will be disappointed there

would be then then then that that that they'll
come down from the high expectation then they'll be

yeah there will be yeah then then slowly is going
to get up there once they find people find the

right use cases and you'll play plateau out to a
some sort of a good use cases for people can

actually
either gain.

Value out of it. But right now it's that's
going through that big hype cycle it in my opinion.

Yeah. So other big tech companies like Amazon and
Google why they were just standing by. Well no.

I'm I'm pretty sure that they're also doing the
research of course. Yeah. They have their own you

know own version of a what's called generative
AI which is like a a chat. Yeah.

It's just that, you know Chappie was one of the
the better ones that sort of came out. Yeah. And

I'm sure other, I mean Google tried it that
they didn't really success in the last ad. But yeah,

I'm sure
there's other other.

You know the the Technet tech company that's doing
this right now? They have been doing it. It's

not charged particles came out all of a sudden, right?
There's a GPT 2 dot O 2.5 then open sources

that's based
on, right?

Ah damn,
it's a church.

Just to interface them like if there were and then
and then. More of a a demo version of what but.

There has been there yeah the GT2 dot O1 one
that O that there that this has been being being

studied throughout and being developed yeah yeah I know
you have the case like the when Apple made

an iPhone the technology they use were
already existed so it's already interface revolutionized the

interface to make people understand their interface and
so I when I think of the chat.

To continue down, it's kind of something
similar case like it's just make people.

Yeah, I think, I think they did also improve on
you know making more much more human like type of

yeah type of a a logic and and conversational like.
So it wasn't just about the the UI side of it,

but it had an improvement over time, OK. So that
maybe they felt like, hey, this is good enough to

really make a big splash. So they they advertise it
right. They just came out. So a lot of people

grabbed to it, but people have been studying
these type of thing for for a while.

So you'll think Google down out of big tech
companies will come out with their own version of what

Google already came out, right? So they did some.
I think they didn't they they tried to to.

Demonstrate their own version to AI or or
advertisement. It didn't. It wasn't very successful at

the
time, but.

I'm I'm pretty sure they'll come up with the next
version, I don't know. So it's that this is still

a
A.

Topic for for for for different tech companies to
really. Yeah. To grab onto and and and make sure

that that they're also
bring their own version, yeah.

When I think about like those kind of AI and
big programs, all the big tests in the US can manage

to do that. So if a smaller content like
South Korea and it's nothing more about in terms of

population
but those.

Better developed company and the countries can survive
how they can use the the AI technology to

for
their industry.

You
mean like?

About yeah Labor and those I'm sure they're they're
using it and and I'm sure there are lots of

there was there was
I mean Korea is a.

Heavy hitter right among the the the tech community
throughout the globe. So I don't think that we

should underestimate I mean certainly you
shouldn't we shouldn't we shouldn't under yeah

underestimate their capability is you know the Samsung
and and all these tech company and I know

yeah they're they're using probably on a daily
basis to to to own their marketing and their

advertisement sure are they
going to come up with.

Your
own aversion.

I mean, if not they're already using their own
version for their development side of it. Yeah, you

know they they they may not advertise it or try not
to sell it. But I am sure there's a lot of

people there doing AI within that company
already think that technology itself is not very

difficult
to.

No,
I mean.

Technology is out there, there there's the people
always developing it, right. So it's open source,

it's open source and there's a community that that
that always been due so that the that there are.

You know exploring. So so
I'm I'm I'm sure that.

Companies like
you know.

Taco or or
neighbor or these?

Sort of what you would think as a more of
a Google like yeah. So yeah, I'm pretty sure that they

already have their own version of AI. And and
I mean there's a reason why Google was not successful

in Korea because neighbor or Cocoto has their
own version, which probably Korean film is a much

stronger of getting what they want versus Google,
right. They have, I think they have like blogs

and yeah, so, so there's
a so you have to.

Understand the market also and they try to
customize yeah customize that their search engine for

for
these.

Market, right. And then seems
like that was the case.

If you look at most of the other companies
in other countries and globally, right Google is used.

Almost, you know, dominantly. Yeah. Used. Except. Except
Korea. Yeah. And maybe China as well. But

there, there is a reason why neighbour and
Kakatoe is doing better. Yeah. Justin. South Korea.

Yeah. Yeah. Because. So they have a very
good version. Same thing, actually. Even for Uber, right.

Uber. Uber is not very popular in Korea. There
was a legal issue. But it's just that story. Yeah,

they got kicked out basically
by the government. Yeah. OK Yeah.

It's not the technology here. It's not about cultural
issue was by the law. Yeah well that's the

that that is that is a yeah that's that's a
side thing. I don't know why that other company other

countries cannot
do that.

Yeah, there's deeper story. I
didn't look into that, but yeah.

Yeah, in the US like the government paid off
the like the taxi drivers drivers that by allowing.