February 28, 2023 • 24 mins
In this episode, Mary Purk (Executive Director, AI & Analytics Center at Wharton School of Business, University of Pennsylvania) and Andreas Welsch discuss how AI leaders can help scale AI adoption across their company. Mary shares her learnings on AI adoption in business and provides valuable insights for listeners looking to benefit from both, deep academic and practical experience.
Key topics:
- Learn how to better run your AI programs
- Drive adoption through effective collaboration
- Focus on cross-collaboration for scalable AI adoption
Listen to the full episode to hear how you can:
- Explore and embrace ChatGPT
- Encourage your team to use AI
- Improve data stewardship
Watch this episode on YouTube: https://youtu.be/YmGLgtvEAhA
Questions or suggestions? Send me a Text Message.
***********
Disclaimer: Views are the participants’ own and do not represent those of any participant’s past, present, or future employers. Participation in this event is independent of any potential business relationship (past, present, or future) between the participants or between their employers.
Level up your AI Leadership game with the AI Leadership Handbook:
https://www.aileadershiphandbook.com
More details:
https://www.intelligence-briefing.com
All episodes:
https://www.intelligence-briefing.com/podcast
Get a weekly thought-provoking post in your inbox:
https://www.intelligence-briefing.com/newsletter
Mark as Played
Transcript
Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
Andreas Welsch (00:00):
Today, we'll talk about accelerating AI
adoption in your business, andwho better to talk to about it
than someone who's seeing thisfrom both actually industry and
academia.
Mary Purk.
Thank you so much for joining.
Mary Purk (00:13):
Thank you, Andreas, for having me.
Really appreciate it.
Andreas Welsch (00:17):
Awesome.
Hey, why don't you tell us alittle bit about yourself, who
you are and what you do?
Mary Purk (00:21):
Sure.
I am currently the executivedirector for AI and analytics
for business at the WhartonSchool at the University of
Pennsylvania.
I've been there for four yearsand the way I've received that
great honor to be there is I'veboth been in industry and
academia.
I have big consulting experiencewith Accenture and then data and
(00:43):
analytics experience throughNielsen and information
resources.
But I did do a stint in betweenthat at the University of
Chicago.
And ran a marketing researchcenter there.
So I do know how important it isto bring academic talent with
industry to solve the currentbusiness problem.
So I'm really excited to be inthat intersection and be here
(01:07):
today to talk to you about AIand analytics.
Andreas Welsch (01:10):
Fantastic.
Thanks again.
It's great to hear you have sucha wealth of experience and I'm
sure we'll have an interestingshow and episode.
So for those of you who are justjoining the stream, drop a
comment in the chat if you arealready using tools like
generative AI ChatGPT and so on,and what do you use'em for?
But Mary, maybe should we play alittle game to kick things off?
Mary Purk (01:33):
Okay, let's play a little game.
What's it gonna be?
Andreas Welsch (01:36):
So this game is called In Your Own Words, and
when I hit this buzzer, thewheel will start spinning.
And when they stop, you'll see asentence.
I'd like you to answer with thefirst thing that comes to mind
and why.
In your own words.
And so to make it a little moreinteresting, you'll only have 60
seconds for your answer.
And for those of you watchingthis live, drop your answer and
(01:59):
why in the chat as well.
I'm really curious.
Mary, are you ready for, What'sthe BUZZ?
Mary Purk (02:05):
I am.
Andreas Welsch (02:06):
Okay, then let's get started.
If AI were a fruit, what wouldit be?
60 seconds.
Mary Purk (02:17):
If AI were a fruit, I think it would be an apple.
Only because when I was atschool, apples always remind me
of education.
And teachers are some of thebest role models that I've had
in my life, especially when Iwent to Montessori school.
And so it would be an apple,because not only would it teach
(02:39):
me wonderful things that I canlearn, but also it is healthy
and there is a wide variety ofthem.
Andreas Welsch (02:47):
Fantastic.
And well within time.
With the ice breaker out of theway, let's jump right into the
first question.
Maybe we start with the mostobvious,one because we titled
the episode around acceleratingAI adoption.
But I'm curious, what are youseeing?
What are maybe, first of allsome of the common challenges
for AI adoption in business thatyou see?
Mary Purk (03:08):
There's a couple of different ones, but I'll
highlight maybe four of them.
And this first one, it's likebeating a dead horse, but, I
think it's the most importantchallenge for AI adoption.
And that is, identifying aclearly defined business problem
for AI.
We have to do that all the time.
At our center, when we'retalking to companies, we spend
(03:30):
at least three or four weeksjust explain to understand what
the problems that they're tryingto solve and dissecting that.
So many times, you might decideyou're gonna look at a problem,
but the second part of that isto also look at the data that
you need to solve that problem.
And in doing that, you'll haveto look at and see if the data
(03:51):
is biased or not.
As most companies andindividuals, they say, I wanna
use AI.
You can't use AI to solve yourbusiness problem unless you have
data.
So it's one and the same.
To use AI, you need data.
So the keys to having AIadoption would be the business
(04:13):
problem to solve for a reallyclearly defined business
problem, the data, and then twoother things are your team, you
need a cross cross-disciplinaryteam to do that.
And you need IT and marketingand finance and operations.
It's multidimensional and it'svery complex.
(04:33):
And then finally, what are youruse cases?
What is the use case you'regoing to solve this for within
your organization?
And then if you solve for, canyour company implement it?
Because why would you solve?
Some people just wanna solve tosolve, but why use all those
resources to solve, if you knowthat your company couldn't
(04:53):
necessarily take the solutionand implement it?
So the keys are making sure youknow what that business problem
is.
That you have the data and youhave enough data to solve it.
A multidisciplinary team.
And then what the use case is,and can you in fact use the
solution and implement it inyour organiz?
Andreas Welsch (05:13):
Thanks for sharing.
And to your point, it feels likewe constantly need to read.
But it's very good to hear alsofrom you, these are the key
things that leaders need tofocus on if they want to have
their AI initiatives succeed.
And it also mirrors what I'mseeing others share as well.
(05:35):
So very good to see how wellaligned that is in keeping an
eye on the chat here.
People are still answering withtheir favorite fruit.
If AI wear a fruit, it could bewatermelon.
Yes.
And, Caryn's response about thedurian.
So why don't we move on to, ournext question.
I know when we had ourpreparation call for this
(05:59):
session, we also talked aboutgenerative AI and that there's
basically no way around this atthe moment.
And it really feels like there'sso much talk in the industry
about ChatGPT, generative AI.
I see that a lot of businessleaders are asking their AI
teams to find that holy grail,that use case that really makes
money or save a lot of money.
(06:19):
But I'm wondering, what are youseeing?
What can leaders actually learnand apply to that explosive
interest around generative ai?
Mary Purk (06:27):
Just prior to what you were saying too, is, what do
leaders really have to zero inon?
And in terms of the use case, Ithink I heard you say that
ChatGPT what is a use case andwith that is remembering that
you don't have to solve it to ahundred percent perfection.
It would be best to solve almostto 80% and pilot it and see what
(06:52):
that adoption is.
And then, do a test and learn,test and improve.
That is also really important.
The only caveat I have to nothaving it be a hundred percent
applicable or correct, is thatyou have to make sure your data
is not biased.
(07:15):
And, that you have to almostover correct and make sure
that's why you also have to havea multidisciplinary team to make
sure that data is not biased.
These things sound really I knewthat I knew that.
Well, why but why are they beingsaid?
And then know the motivation forthat.
(07:36):
The motivation for themultidisciplinary team isn't cuz
it's so popular to do right now.
It's a fact.
You need it.
That's your insurance to not putsomething in the marketplace
that all of a sudden blows upand you find out that it's so
biased and then you've lost halfyour customer base.
Your people are your insuranceto making sure that you can have
(07:58):
successful AI adoption andrevenues that would come from
that.
Okay, so now, leaders are.
We have so much dinner with AIChatGPT, what do we do?
We're behind the eight ball andthe bottom, the most simplest
expression I can use is (08:16):
get on board.
Get on board.
It's like you cannot be on thesidelines.
You have to embrace it yourself.
I would encourage you to open aChatGPT account and experiment
with it.
If you haven't, just set asidesome time on your calendar to do
(08:37):
it.
And then have some of yoursignificant others or other
people in your lives encouragethem to use it.
The more people are using it, itwill help shape it for the good
in our society.
I think I heard like over it'sthe fastest growing app we've
ever had.
There's over what, a hundredmillion users on it, and it's
gonna change.
(08:58):
I would say get on board.
Invest in it.
As a leader, you can put outinto your team use ChatGPT to
solve a pain point that you havein your process flow for the
week or a task that you have forthe week.
And then share that with yoursupervisor.
It's do it like, this is anexercise and then as a leader
(09:22):
you're showing how you'recurrent.
This is important to us.
We're going to experiment, we'regonna discover, these are all
the things you can do withChatGPT.
Now's the time to do that.
So that's what I would say as aleader.
If you haven't opened anaccount, open an account, use
it, and then use it personallyand use it in the in business.
Andreas Welsch (09:42):
I think that's a fantastic call to action.
And the reason why I think thatis because it's also so much
more accessible now.
Compare that to six, seven yearsago when we were just climbing
that hype cycle and everybodywas getting excited about
machine learning, AI,self-driving, cars, flying
drones, packages delivered bydrones and all that stuff.
Now you can actually touch itand feel it.
(10:03):
And you can feel it in so manydifferent ways.
Again, not just image, videoaudio transcription,
summarization of text,generation of text, and so on.
So I think that's an excellentcall to action to get onboard.
Mary Purk (10:19):
And it's fun.
There's discovery in it.
But know that ChatGPT is afunction of what everything that
we as humans have put out on theinternet, all of it.
Most of it's true, some of it'snot true.
So we still have to use our owninstincts and other knowledge
points to say we might need tochange that.
(10:41):
It's not all fact.
We just have to remember that.
And then as we're reading thingsor hearing things, we are going
to have to use our own filteringto know what is good and what
maybe we have to take out thatwe don't necessarily believe.
And you might have to use otherresources or other sources to
validate
Andreas Welsch (11:01):
I'm looking at the chat.
I see one message from Ken whosays, you mentioned the notion
of a multidisciplinary team ofpeople.
So do you think that we mightsoon see multiple AI used to
cross-check each other'srecommendations?
Mary Purk (11:17):
Oh, sure.
I, would think that.
We talked about this, remember?
Everyone's so excited aboutusing the application.
We're so busy discovering itthat some people have looked
forward, but this is a verypractical use you just brought
up.
There will be those who haveinvested early and understand
some of the capabilities ofChatGPT who actually develop
(11:41):
applications to further verifyspecific things that come out of
it.
It could be how there'sdifferent applications for a
dictionary and Grammarly andthings like that.
There might be FactCheckerGPT orsomething to that effect.
There will be differentsplinters within ChatGPT.
They'll be even more specificand more narrow in what they do.
Andreas Welsch (12:06):
Fantastic.
I think that's a very goodoutlook and a good summary.
Can you share some examples ofwhat that collaboration looks
like that's successful betweenbusiness leaders and business
teams and technology teams?
Just to get more adoption.
Maybe an additional question tothat is generative AI all of a
(12:27):
sudden so much differentcompared to what we were doing
on November 29th, before ChatGPTwas released.
Mary Purk (12:37):
I'll take more of a simple approach to this in terms
of the collaboration betweenbusiness and technology that
foster AI and machine learning.
There might be some people thatremember, but we needed like CDs
and we needed albums and stuff,but then came along a company
called Pandora and they decidedthey were going to offer many
(13:01):
different songs to people.
A whole library of songs thatwere gonna be available.
And then it was going to becustomized to your taste.
But there was no data aroundthat.
So they then scraped all thesesongs and provided all these
different qualifiers for songsso they could create a library.
Very pure data set that thenprovided very good
(13:22):
recommendations for individuals.
So that just was an explosion inthe music industry, because they
saw that need and thatpersonalization.
But a lot of that was dependenton business needs and technology
to create that data that wasneeded.
Then there was a competitor,Spotify, that saw that and they
said, we're not gonna do that.
That was way too time consuming.
(13:44):
But it's a really good idea.
And they then slowly grew theirown dataset through uses,
through people using the data.
And then that dataset grew andthat's what's gonna happen with
ChatGPT's.
All this it's gonna kept gettingbigger and bigger as we all are
contributing to the data.
And the bottom line is for bothof those organizations, they had
(14:07):
people that understood what theend user needed, but
technologists that understooddata.
We can't talk about generative.
AI without talking about data.
People really have to realizedata fuels the AI algorithm.
And so as we're talking aboutthat, after we get done with all
(14:28):
the novelty, we're gonna startspending more time on
understanding how that data'sgonna be collected.
Who's gonna be continuing tocontribute to the data?
How are we gonna filter outtruth or untruthful data?
And so that would be oneapplication of AI and ML that
brought a whole new industry tous.
We also know what's happenedwith Amazon, good things and bad
(14:49):
things.
Amazon's able to feed us evenbetter things that we might want
from our previous buying historyor previous browsing history.
But then they also got introuble because of some hiring
recommendations when they theyonly used their current data.
And they didn't even think thatmaybe it was biased towards one
gender or one type of skillset.
(15:11):
And they're like, oops, guess wehave to expand it.
So that's why you need bothbusiness and technologists to
look at those algorithms.
But like I said, both of thoseexamples I gave were related to
understanding what data youneeded to get the best output.
Bad things into a process,possibly bad recommendations.
(15:31):
So I think just keeping thatsimple analogy at top of mind is
also very important for bothsides of the teams of the
business and technologists.
Andreas Welsch (15:43):
I see some questions here, and I think
Ramnath is asking did Microsoftopen the floodgates too quickly
without guidelines on GPT-3 andthese kinds of technologies?
But building on your pointaround data, right?
You see this concern of ChatGPTand this kind of technology
answering the user in ways thatmight be creepy, or that might
(16:05):
be perceived as being creepy or,just inacceptable in so many
different ways.
And then when you look at thetype of data, the corpus that
these large language models havebeen trained on, hey, if it's
scraped off the internet inpublicly available forums in and
sources.
Then it's also a mirror and areflection of how we as humans
(16:25):
and as people communicate andthe biases that we hold.
Whether they're conscious orsubconscious or explicit or
implicit.
But I see Mike has aninteresting one here.
He says, what are your thoughtson build versus buy AI in the
enterprise?
What are the commercial factorsthat will govern the decision in
an organization?
(16:45):
And I can see this applied toboth before generative AI really
popped in now after.
Mary Purk (16:55):
I'm, from the Midwest.
And I'm a mom of four, and I'mthe oldest in my family.
So I am super practical.
I am super practical.
Time is of the essence.
I'm a busy person and so peoplealways ask me to do something
cuz I guess a busy person alwaysknows how can get more things
done.
And I would say, as you can seewhere I might be leaning, if
(17:17):
there is an application outthere that meets your.
needs.
I would choose that AIapplication before building it
in-house.
You obviously have to vet theindividuals or the company
that's building that AIapplication, but they are that
much more ahead in theirtechnology and knowledge.
And then the key that theenterprise brings is making sure
(17:41):
that you have the right peopleevaluating that.
And they have not only the focusof what the current situation is
in the marketplace for them, butlooking forward and that they
can clearly define what thebusiness problem or pain points
are that they're using AI tosolve for.
So that's where I would go.
I think there was someone like,what about in the regulatory
industry?
(18:02):
There'll be certain companiesthat cannot buy anything off the
shelf and will have to buildin-house for security reasons.
And I would say in otherinstances, it will have to go
that route.
Andreas Welsch (18:13):
That makes sense.
I think that also brings up thenext question and the last one
for us to end on today.
So if we think about AI adoptionin the enterprise, and there's
AI, now there's generative AI.
There's still this buzz and thisinterest around AI and how can
(18:34):
we get value out of it.
How can we get more adoption inbusiness and IT leaders should
work together?
What do you think should they-business leaders and IT
leaders- focus their teams on inthe next six to 12 months in
setting the expectations when itis about AI adoption in
business?
Mary Purk (18:52):
The first thing, and I think we started out with
this, is experimentation.
I think that you have to startout with experimentation with
your team and you as a leadertell them and be forthright.
If you aren't feeling this, thendon't say it.
But you know that you wannaexperiment with ChatGPT in in
your processes and challengeyour team to come up with the
(19:14):
best ways to use generative AIor if it has to be built
in-house or purchase or such.
But you can use ChatGPT tosimulate that.
I think that would be the firstthing.
So experimentation, second data.
We've had data engineers, datascientists, but every single
person on your team in theorganization needs to know you
are a data steward.
(19:36):
There's certain things we'recompelled to do.
If someone falls down, we go andpick them up.
We need to teach people.
This is also about data.
If there is something bad goingon data, they have to be able to
speak up about it.
It's in kin to be awhistleblower, but you just have
to bring these things up.
If you're at that table and yousee something that's not right
or something that's really good,it's up to you to speak up.
(19:58):
So I think make giving them thepermission to do that.
And then I think for me, theglass is always half full.
You can make lemonade out oflemons.
It's a great.
Time to be in with technology.
There's so much that you can doand it's a level playing field.
I learned FORTRAN and COBOL.
(20:19):
That's not relevant.
Guess there's some programsstill in COBOL there's R and
Python now.
That's gonna be obsolete soon.
But this, technology, I don'tneed to know how to code.
I need to know how to writereally good questions.
Maybe really good logic.
If you could say anything peopleneed to know how to logically
put together a story or abusiness problem to solve.
(20:41):
That's what you need and youneed to be a good editor.
That's what we've been hearing,right?
We don't have to be greatwriters, although I love
writing.
But you got to be a really goodeditor and know what you wanna
communicate.
So giving people really good,positive attitude to go and
explore and discover, butthey're responsible for
contributing and their job willbe there if they learn to use AI
(21:03):
and use these tools.
Andreas Welsch (21:04):
I think that's a very encouraging call to action
also to not only people who arein a formal leadership role, but
actually everybody.
Go try it out.
Experiment, learn and becomepart of the discussion so that
you can have an informeddiscussion as well.
(21:24):
Maybe as we're getting close tothe end of the show, can you
summarize the three keytakeaways for our audience
today?
Mary Purk (21:31):
The three key takeaways.
Remember what fruit you thoughtAI would be.
I think that's important.
And you can tell that at dinnertonight or this weekend if
you're going out.
I think that'd be a greatconversation starter.
Thank you for starting us offthat way.
Two, if you don't have an aChatGPT account.
Open it.
And then, if you're a leader oryou can influence your team,
(21:55):
challenge people to use thetechnology.
And thirdly really I think tome, data has always been very
close to my heart in my career.
And I think it just never goesaway.
So I think making sure that yourcompany is doing better job of
creating data stewardsthroughout the company.
(22:17):
And if I could just add one lastthing, is you don't have to get
it a hundred percent correct tomove the the idea or technology
out of pilot, maybe into whereit could be tested further.
So if you do wait till it's ahundred percent correct, the
opportunity might pass you by.
So risk this, risk reward.
(22:39):
I'm not saying things are gonnabe oh easy now that we have
ChatGPT, but it's gonna be a lotmore interesting and you're
gonna be able to have more freetime.
I'm looking forward to creatingmy own personal secretary now
with ChatGPT.
So those will be my summarypoints.
Andreas Welsch (22:55):
Fantastic.
Thank you so much for sharing.
So folks, we're getting close tothe end of the show today.
Thank you so much for joiningus, and Mary, thanks for sharing
your expertise with us.
Mary Purk (23:05):
Oh, thank you for inviting me.
It was really a pleasure.
Andreas Welsch (23:08):
Fantastic.
Today, we'll talk about accelerating AI
adoption in your business, andwho better to talk to about it
than someone who's seeing thisfrom both actually industry and
academia.
Mary Purk.
Thank you so much for joining.
Mary Purk (00:13):
Thank you, Andreas, for having me.
Really appreciate it.
Andreas Welsch (00:17):
Awesome.
Hey, why don't you tell us alittle bit about yourself, who
you are and what you do?
Mary Purk (00:21):
Sure.
I am currently the executivedirector for AI and analytics
for business at the WhartonSchool at the University of
Pennsylvania.
I've been there for four yearsand the way I've received that
great honor to be there is I'veboth been in industry and
academia.
I have big consulting experiencewith Accenture and then data and
(00:43):
analytics experience throughNielsen and information
resources.
But I did do a stint in betweenthat at the University of
Chicago.
And ran a marketing researchcenter there.
So I do know how important it isto bring academic talent with
industry to solve the currentbusiness problem.
So I'm really excited to be inthat intersection and be here
(01:07):
today to talk to you about AIand analytics.
Andreas Welsch (01:10):
Fantastic.
Thanks again.
It's great to hear you have sucha wealth of experience and I'm
sure we'll have an interestingshow and episode.
So for those of you who are justjoining the stream, drop a
comment in the chat if you arealready using tools like
generative AI ChatGPT and so on,and what do you use'em for?
But Mary, maybe should we play alittle game to kick things off?
Mary Purk (01:33):
Okay, let's play a little game.
What's it gonna be?
Andreas Welsch (01:36):
So this game is called In Your Own Words, and
when I hit this buzzer, thewheel will start spinning.
And when they stop, you'll see asentence.
I'd like you to answer with thefirst thing that comes to mind
and why.
In your own words.
And so to make it a little moreinteresting, you'll only have 60
seconds for your answer.
And for those of you watchingthis live, drop your answer and
(01:59):
why in the chat as well.
I'm really curious.
Mary, are you ready for, What'sthe BUZZ?
Mary Purk (02:05):
I am.
Andreas Welsch (02:06):
Okay, then let's get started.
If AI were a fruit, what wouldit be?
60 seconds.
Mary Purk (02:17):
If AI were a fruit, I think it would be an apple.
Only because when I was atschool, apples always remind me
of education.
And teachers are some of thebest role models that I've had
in my life, especially when Iwent to Montessori school.
And so it would be an apple,because not only would it teach
(02:39):
me wonderful things that I canlearn, but also it is healthy
and there is a wide variety ofthem.
Andreas Welsch (02:47):
Fantastic.
And well within time.
With the ice breaker out of theway, let's jump right into the
first question.
Maybe we start with the mostobvious,one because we titled
the episode around acceleratingAI adoption.
But I'm curious, what are youseeing?
What are maybe, first of allsome of the common challenges
for AI adoption in business thatyou see?
Mary Purk (03:08):
There's a couple of different ones, but I'll
highlight maybe four of them.
And this first one, it's likebeating a dead horse, but, I
think it's the most importantchallenge for AI adoption.
And that is, identifying aclearly defined business problem
for AI.
We have to do that all the time.
At our center, when we'retalking to companies, we spend
(03:30):
at least three or four weeksjust explain to understand what
the problems that they're tryingto solve and dissecting that.
So many times, you might decideyou're gonna look at a problem,
but the second part of that isto also look at the data that
you need to solve that problem.
And in doing that, you'll haveto look at and see if the data
(03:51):
is biased or not.
As most companies andindividuals, they say, I wanna
use AI.
You can't use AI to solve yourbusiness problem unless you have
data.
So it's one and the same.
To use AI, you need data.
So the keys to having AIadoption would be the business
(04:13):
problem to solve for a reallyclearly defined business
problem, the data, and then twoother things are your team, you
need a cross cross-disciplinaryteam to do that.
And you need IT and marketingand finance and operations.
It's multidimensional and it'svery complex.
(04:33):
And then finally, what are youruse cases?
What is the use case you'regoing to solve this for within
your organization?
And then if you solve for, canyour company implement it?
Because why would you solve?
Some people just wanna solve tosolve, but why use all those
resources to solve, if you knowthat your company couldn't
(04:53):
necessarily take the solutionand implement it?
So the keys are making sure youknow what that business problem
is.
That you have the data and youhave enough data to solve it.
A multidisciplinary team.
And then what the use case is,and can you in fact use the
solution and implement it inyour organiz?
Andreas Welsch (05:13):
Thanks for sharing.
And to your point, it feels likewe constantly need to read.
But it's very good to hear alsofrom you, these are the key
things that leaders need tofocus on if they want to have
their AI initiatives succeed.
And it also mirrors what I'mseeing others share as well.
(05:35):
So very good to see how wellaligned that is in keeping an
eye on the chat here.
People are still answering withtheir favorite fruit.
If AI wear a fruit, it could bewatermelon.
Yes.
And, Caryn's response about thedurian.
So why don't we move on to, ournext question.
I know when we had ourpreparation call for this
(05:59):
session, we also talked aboutgenerative AI and that there's
basically no way around this atthe moment.
And it really feels like there'sso much talk in the industry
about ChatGPT, generative AI.
I see that a lot of businessleaders are asking their AI
teams to find that holy grail,that use case that really makes
money or save a lot of money.
(06:19):
But I'm wondering, what are youseeing?
What can leaders actually learnand apply to that explosive
interest around generative ai?
Mary Purk (06:27):
Just prior to what you were saying too, is, what do
leaders really have to zero inon?
And in terms of the use case, Ithink I heard you say that
ChatGPT what is a use case andwith that is remembering that
you don't have to solve it to ahundred percent perfection.
It would be best to solve almostto 80% and pilot it and see what
(06:52):
that adoption is.
And then, do a test and learn,test and improve.
That is also really important.
The only caveat I have to nothaving it be a hundred percent
applicable or correct, is thatyou have to make sure your data
is not biased.
(07:15):
And, that you have to almostover correct and make sure
that's why you also have to havea multidisciplinary team to make
sure that data is not biased.
These things sound really I knewthat I knew that.
Well, why but why are they beingsaid?
And then know the motivation forthat.
(07:36):
The motivation for themultidisciplinary team isn't cuz
it's so popular to do right now.
It's a fact.
You need it.
That's your insurance to not putsomething in the marketplace
that all of a sudden blows upand you find out that it's so
biased and then you've lost halfyour customer base.
Your people are your insuranceto making sure that you can have
(07:58):
successful AI adoption andrevenues that would come from
that.
Okay, so now, leaders are.
We have so much dinner with AIChatGPT, what do we do?
We're behind the eight ball andthe bottom, the most simplest
expression I can use is (08:16):
get on board.
Get on board.
It's like you cannot be on thesidelines.
You have to embrace it yourself.
I would encourage you to open aChatGPT account and experiment
with it.
If you haven't, just set asidesome time on your calendar to do
(08:37):
it.
And then have some of yoursignificant others or other
people in your lives encouragethem to use it.
The more people are using it, itwill help shape it for the good
in our society.
I think I heard like over it'sthe fastest growing app we've
ever had.
There's over what, a hundredmillion users on it, and it's
gonna change.
(08:58):
I would say get on board.
Invest in it.
As a leader, you can put outinto your team use ChatGPT to
solve a pain point that you havein your process flow for the
week or a task that you have forthe week.
And then share that with yoursupervisor.
It's do it like, this is anexercise and then as a leader
(09:22):
you're showing how you'recurrent.
This is important to us.
We're going to experiment, we'regonna discover, these are all
the things you can do withChatGPT.
Now's the time to do that.
So that's what I would say as aleader.
If you haven't opened anaccount, open an account, use
it, and then use it personallyand use it in the in business.
Andreas Welsch (09:42):
I think that's a fantastic call to action.
And the reason why I think thatis because it's also so much
more accessible now.
Compare that to six, seven yearsago when we were just climbing
that hype cycle and everybodywas getting excited about
machine learning, AI,self-driving, cars, flying
drones, packages delivered bydrones and all that stuff.
Now you can actually touch itand feel it.
(10:03):
And you can feel it in so manydifferent ways.
Again, not just image, videoaudio transcription,
summarization of text,generation of text, and so on.
So I think that's an excellentcall to action to get onboard.
Mary Purk (10:19):
And it's fun.
There's discovery in it.
But know that ChatGPT is afunction of what everything that
we as humans have put out on theinternet, all of it.
Most of it's true, some of it'snot true.
So we still have to use our owninstincts and other knowledge
points to say we might need tochange that.
(10:41):
It's not all fact.
We just have to remember that.
And then as we're reading thingsor hearing things, we are going
to have to use our own filteringto know what is good and what
maybe we have to take out thatwe don't necessarily believe.
And you might have to use otherresources or other sources to
validate
Andreas Welsch (11:01):
I'm looking at the chat.
I see one message from Ken whosays, you mentioned the notion
of a multidisciplinary team ofpeople.
So do you think that we mightsoon see multiple AI used to
cross-check each other'srecommendations?
Mary Purk (11:17):
Oh, sure.
I, would think that.
We talked about this, remember?
Everyone's so excited aboutusing the application.
We're so busy discovering itthat some people have looked
forward, but this is a verypractical use you just brought
up.
There will be those who haveinvested early and understand
some of the capabilities ofChatGPT who actually develop
(11:41):
applications to further verifyspecific things that come out of
it.
It could be how there'sdifferent applications for a
dictionary and Grammarly andthings like that.
There might be FactCheckerGPT orsomething to that effect.
There will be differentsplinters within ChatGPT.
They'll be even more specificand more narrow in what they do.
Andreas Welsch (12:06):
Fantastic.
I think that's a very goodoutlook and a good summary.
Can you share some examples ofwhat that collaboration looks
like that's successful betweenbusiness leaders and business
teams and technology teams?
Just to get more adoption.
Maybe an additional question tothat is generative AI all of a
(12:27):
sudden so much differentcompared to what we were doing
on November 29th, before ChatGPTwas released.
Mary Purk (12:37):
I'll take more of a simple approach to this in terms
of the collaboration betweenbusiness and technology that
foster AI and machine learning.
There might be some people thatremember, but we needed like CDs
and we needed albums and stuff,but then came along a company
called Pandora and they decidedthey were going to offer many
(13:01):
different songs to people.
A whole library of songs thatwere gonna be available.
And then it was going to becustomized to your taste.
But there was no data aroundthat.
So they then scraped all thesesongs and provided all these
different qualifiers for songsso they could create a library.
Very pure data set that thenprovided very good
(13:22):
recommendations for individuals.
So that just was an explosion inthe music industry, because they
saw that need and thatpersonalization.
But a lot of that was dependenton business needs and technology
to create that data that wasneeded.
Then there was a competitor,Spotify, that saw that and they
said, we're not gonna do that.
That was way too time consuming.
(13:44):
But it's a really good idea.
And they then slowly grew theirown dataset through uses,
through people using the data.
And then that dataset grew andthat's what's gonna happen with
ChatGPT's.
All this it's gonna kept gettingbigger and bigger as we all are
contributing to the data.
And the bottom line is for bothof those organizations, they had
(14:07):
people that understood what theend user needed, but
technologists that understooddata.
We can't talk about generative.
AI without talking about data.
People really have to realizedata fuels the AI algorithm.
And so as we're talking aboutthat, after we get done with all
(14:28):
the novelty, we're gonna startspending more time on
understanding how that data'sgonna be collected.
Who's gonna be continuing tocontribute to the data?
How are we gonna filter outtruth or untruthful data?
And so that would be oneapplication of AI and ML that
brought a whole new industry tous.
We also know what's happenedwith Amazon, good things and bad
(14:49):
things.
Amazon's able to feed us evenbetter things that we might want
from our previous buying historyor previous browsing history.
But then they also got introuble because of some hiring
recommendations when they theyonly used their current data.
And they didn't even think thatmaybe it was biased towards one
gender or one type of skillset.
(15:11):
And they're like, oops, guess wehave to expand it.
So that's why you need bothbusiness and technologists to
look at those algorithms.
But like I said, both of thoseexamples I gave were related to
understanding what data youneeded to get the best output.
Bad things into a process,possibly bad recommendations.
(15:31):
So I think just keeping thatsimple analogy at top of mind is
also very important for bothsides of the teams of the
business and technologists.
Andreas Welsch (15:43):
I see some questions here, and I think
Ramnath is asking did Microsoftopen the floodgates too quickly
without guidelines on GPT-3 andthese kinds of technologies?
But building on your pointaround data, right?
You see this concern of ChatGPTand this kind of technology
answering the user in ways thatmight be creepy, or that might
(16:05):
be perceived as being creepy or,just inacceptable in so many
different ways.
And then when you look at thetype of data, the corpus that
these large language models havebeen trained on, hey, if it's
scraped off the internet inpublicly available forums in and
sources.
Then it's also a mirror and areflection of how we as humans
(16:25):
and as people communicate andthe biases that we hold.
Whether they're conscious orsubconscious or explicit or
implicit.
But I see Mike has aninteresting one here.
He says, what are your thoughtson build versus buy AI in the
enterprise?
What are the commercial factorsthat will govern the decision in
an organization?
(16:45):
And I can see this applied toboth before generative AI really
popped in now after.
Mary Purk (16:55):
I'm, from the Midwest.
And I'm a mom of four, and I'mthe oldest in my family.
So I am super practical.
I am super practical.
Time is of the essence.
I'm a busy person and so peoplealways ask me to do something
cuz I guess a busy person alwaysknows how can get more things
done.
And I would say, as you can seewhere I might be leaning, if
(17:17):
there is an application outthere that meets your.
needs.
I would choose that AIapplication before building it
in-house.
You obviously have to vet theindividuals or the company
that's building that AIapplication, but they are that
much more ahead in theirtechnology and knowledge.
And then the key that theenterprise brings is making sure
(17:41):
that you have the right peopleevaluating that.
And they have not only the focusof what the current situation is
in the marketplace for them, butlooking forward and that they
can clearly define what thebusiness problem or pain points
are that they're using AI tosolve for.
So that's where I would go.
I think there was someone like,what about in the regulatory
industry?
(18:02):
There'll be certain companiesthat cannot buy anything off the
shelf and will have to buildin-house for security reasons.
And I would say in otherinstances, it will have to go
that route.
Andreas Welsch (18:13):
That makes sense.
I think that also brings up thenext question and the last one
for us to end on today.
So if we think about AI adoptionin the enterprise, and there's
AI, now there's generative AI.
There's still this buzz and thisinterest around AI and how can
(18:34):
we get value out of it.
How can we get more adoption inbusiness and IT leaders should
work together?
What do you think should they-business leaders and IT
leaders- focus their teams on inthe next six to 12 months in
setting the expectations when itis about AI adoption in
business?
Mary Purk (18:52):
The first thing, and I think we started out with
this, is experimentation.
I think that you have to startout with experimentation with
your team and you as a leadertell them and be forthright.
If you aren't feeling this, thendon't say it.
But you know that you wannaexperiment with ChatGPT in in
your processes and challengeyour team to come up with the
(19:14):
best ways to use generative AIor if it has to be built
in-house or purchase or such.
But you can use ChatGPT tosimulate that.
I think that would be the firstthing.
So experimentation, second data.
We've had data engineers, datascientists, but every single
person on your team in theorganization needs to know you
are a data steward.
(19:36):
There's certain things we'recompelled to do.
If someone falls down, we go andpick them up.
We need to teach people.
This is also about data.
If there is something bad goingon data, they have to be able to
speak up about it.
It's in kin to be awhistleblower, but you just have
to bring these things up.
If you're at that table and yousee something that's not right
or something that's really good,it's up to you to speak up.
(19:58):
So I think make giving them thepermission to do that.
And then I think for me, theglass is always half full.
You can make lemonade out oflemons.
It's a great.
Time to be in with technology.
There's so much that you can doand it's a level playing field.
I learned FORTRAN and COBOL.
(20:19):
That's not relevant.
Guess there's some programsstill in COBOL there's R and
Python now.
That's gonna be obsolete soon.
But this, technology, I don'tneed to know how to code.
I need to know how to writereally good questions.
Maybe really good logic.
If you could say anything peopleneed to know how to logically
put together a story or abusiness problem to solve.
(20:41):
That's what you need and youneed to be a good editor.
That's what we've been hearing,right?
We don't have to be greatwriters, although I love
writing.
But you got to be a really goodeditor and know what you wanna
communicate.
So giving people really good,positive attitude to go and
explore and discover, butthey're responsible for
contributing and their job willbe there if they learn to use AI
(21:03):
and use these tools.
Andreas Welsch (21:04):
I think that's a very encouraging call to action
also to not only people who arein a formal leadership role, but
actually everybody.
Go try it out.
Experiment, learn and becomepart of the discussion so that
you can have an informeddiscussion as well.
(21:24):
Maybe as we're getting close tothe end of the show, can you
summarize the three keytakeaways for our audience
today?
Mary Purk (21:31):
The three key takeaways.
Remember what fruit you thoughtAI would be.
I think that's important.
And you can tell that at dinnertonight or this weekend if
you're going out.
I think that'd be a greatconversation starter.
Thank you for starting us offthat way.
Two, if you don't have an aChatGPT account.
Open it.
And then, if you're a leader oryou can influence your team,
(21:55):
challenge people to use thetechnology.
And thirdly really I think tome, data has always been very
close to my heart in my career.
And I think it just never goesaway.
So I think making sure that yourcompany is doing better job of
creating data stewardsthroughout the company.
(22:17):
And if I could just add one lastthing, is you don't have to get
it a hundred percent correct tomove the the idea or technology
out of pilot, maybe into whereit could be tested further.
So if you do wait till it's ahundred percent correct, the
opportunity might pass you by.
So risk this, risk reward.
(22:39):
I'm not saying things are gonnabe oh easy now that we have
ChatGPT, but it's gonna be a lotmore interesting and you're
gonna be able to have more freetime.
I'm looking forward to creatingmy own personal secretary now
with ChatGPT.
So those will be my summarypoints.
Andreas Welsch (22:55):
Fantastic.
Thank you so much for sharing.
So folks, we're getting close tothe end of the show today.
Thank you so much for joiningus, and Mary, thanks for sharing
your expertise with us.
Mary Purk (23:05):
Oh, thank you for inviting me.
It was really a pleasure.
Andreas Welsch (23:08):
Fantastic.
Popular Podcasts
Dateline NBC
Current and classic episodes, featuring compelling true-crime mysteries, powerful documentaries and in-depth investigations. Follow now to get the latest episodes of Dateline NBC completely free, or subscribe to Dateline Premium for ad-free listening and exclusive bonus content: DatelinePremium.com
24/7 News: The Latest
The latest news in 4 minutes updated every hour, every day.
Therapy Gecko
An unlicensed lizard psychologist travels the universe talking to strangers about absolutely nothing. TO CALL THE GECKO: follow me on https://www.twitch.tv/lyleforever to get a notification for when I am taking calls. I am usually live Mondays, Wednesdays, and Fridays but lately a lot of other times too. I am a gecko.