September 5, 2023 • 33 mins
As businesses adopt AI, a new era of problem-solving, innovation, and creative decision-making can be brought to scale. In this episode of Smart Talks with IBM, Malcolm Gladwell and Jacob Goldstein explore the future of AI in enterprise business AI for business with Kareem Yusuf, senior vice president of product management and growth for IBM software. They discuss the advent of foundation models, how AI can transform data storage and decision-making, and how next-generation AI platforms like watsonx from IBM can empower businesses to use AI at scale.
This is a paid advertisement from IBM.
See omnystudio.com/listener for privacy information.
Mark as Played
Transcript
Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
Speaker 1 (00:04):
Welcome to Tech Stuff, a production from iHeartRadio. Today, we
are witnessed to one of those rare moments in history,
the rise of an innovative technology with the potential to
radically transform business in society forever. That technology, of course,
(00:24):
is artificial intelligence, and it's the central focus for this
new season of Smart Talks with IBM. Join hosts from
your favorite Pushkin podcasts as they talk with industry experts
and leaders to explore how businesses can integrate AI into
their workflows and help drive real change in this new
era of AI, and of course, host Malcolm Gladwell will
(00:46):
be there to guide you through the season and throw
in his two cents as well. Look out for new
episodes of Smart Talks with IBM every other week on
the iHeartRadio app, Apple Podcasts, or wherever you get your podcasts,
and learn more at IBM dot com slash smart Talks.
Speaker 2 (01:05):
Hello, Hello, Welcome to Smart Talks with IBM, a podcast
from Pushkin Industries, iHeartRadio and IBM. I'm Malcolm Glabo. This season,
we're continuing our conversation with new creators visionaries who are
creatively applying technology in business to drive change, but with
a focus on the transformative power of artificial intelligence and
(01:27):
what it means to leverage AI as a game changing
multiplier for your business. Our guest today is Kareem Yousef,
Senior Vice President of Product Management and Growth for IBM Software.
Kareem's focus at IBM is on product strategy, thinking about
the roadmap for IBM Software products and how they can
deliver effective and compelling customer experiences with the current boom
(01:53):
and generative AI. Kareem's job is to help businesses figure
out how they can apply artificial intelligence at scale to
help solve big problems and boost productivity at new orders
of magnitude. In today's episode, you'll hear Kareem explain how
AI powered by foundation models can make AI adoption by
enterprise businesses even easier, how generative AI will change the
(02:18):
way businesses process data and make decisions, and how these
considerations influence the design of Watson x, IBM's next generation
AI and data platform. Kareem spoke with Jacob Goldstein, host
of the Pushkin podcast What's Your Problem. A veteran business journalist,
Jacob has reported for The Wall Street Journal, the Miami Herald,
(02:41):
and was a longtime host of the NPR program Planet Money. Okay,
let's get to the interview.
Speaker 3 (02:49):
I'm Jacob Goldstein. I'm one of the hosts at Pushkin
and a correspondent on this show, and I'm delighted to
have you here. Can you introduce yourself?
Speaker 4 (02:57):
Ah? Hi, I'm Kareem Yusuf. I'm this and your vice
president of Product management and Growth for IBM Software. You
can think of me as the chief product officer for
IBM Software.
Speaker 3 (03:07):
Okay, sounds like a big job. We're here today to
talk about AI. We've heard really an extraordinary amount in
the last few months about chat GPT and you know,
particularly in how it's used in the very kind of
consumer facing way. But I'm curious what is the rise
(03:27):
of chat GPT and you know, AI more generally, what
does it mean for business?
Speaker 4 (03:32):
Well, you know, it's if you kind of step back
and think about what really happens. You know, in a business,
you're really talking about a set of processes, right, you know,
activities that represent what a business needs to get done,
whether it's product they produce and then sell or service
that they provide. And inherent to operating the business, I
(03:54):
would say are two very key factors. Data and then
the decisions you make around data and then actually lastly
the processes or activities you do in accordance with that decision.
So if you then think about AI as applied to
business right in that context, well, the first place it
often starts is how do you make sense of a
(04:16):
lot of the data associated with driving the business? And
so AI has always been, in my mind at its
foremost about gaining insights then lead in to supporting decisions,
and ultimately ending at helping to automate the activities that
then are executed as a result of those decisions. So
(04:38):
that's kind of my simple way of thinking of AI,
and we can obviously coloring with examples, but that's my
simplest way of thinking about AI. When you think about
in the business context, gain insights from masses of data
to support decisions and then drive.
Speaker 3 (04:52):
Actions, that's a really helpful framework. And then if we
think about sort of what's happening in the world now
with you know, enterprise businesses NAI, what are you seeing
with enterprise adoption of AI at this moment?
Speaker 4 (05:06):
So we're really talking about almost a tale of two periods.
So let me first of all kind of take you
back before the advent of what I would call generative AI,
and the whole chat gpt to what has been going
on in what I would term the realm of more
standardized machine learning models. A lot of what has been
(05:26):
going on has been very much in the realms of
certain things like anomaly detection or optimization, right, using machine
learning models to do that kind of work, and where
might it apply well, think of anomaly detection in security
software right detecting threats based upon different events flowing through
(05:47):
or in enterprise asset management software monitoring equipment and detecting
anomalies within their behavior, or even in IT automation software
once again detecting anomalies based upon what's going on with
various IT events and then tasks that should occur. Optimizations
often play around in the realm as you might imagine
(06:09):
to solve problems of resource optimization, whether you think of
that in the context of application resource management for IT
or in the context of supply chain. These have been
very classical applications of machine learning AI to really make
sense of the data and provide a basis to drive decisions. Now,
(06:30):
what is characterized by all those examples have given and
the state of the art of that kind of technology
has always been it's very task specific. So there was
a air quotes, if I may, kind of limitation in
the sense that the tak it had to be very
(06:50):
task specific. And so we've seen a lot of broad
based adoption within the enterprise, right, but it's very very
task specific. As you might imagine. Now, what has happened
recently and has been brought to the four has been
this discussion of generative AI, which is powered by a
very specific innovation, this notion of foundation models. And in
(07:14):
the simplest way to think about it, it's about training
this large model that can then be refined to various tasks.
And the easiest one that everybody recognized at the moment
is the notion of a large language model, a model
that has an understanding of a lot of the elements
(07:37):
of a language such that it can be refined to
a variety of tasks. Write an essay, answer a question,
singer songs, so on, answers so forth. I like to
liken the power if you like, and this will speak
to the why everybody is so excited about it. Why
would argue at an inflection point? I like to liken
(07:58):
it to teaching a child the alphabet. When you teach
a child and alphabet, it's a set of letters. Right,
Let's call that our foundation model. But over time that
knowledge of the alphabet is tuned to read a book,
write an essay, do a composition, create a song, write
a poem, write an invoice. You understand what I mean, right,
(08:21):
And so from one foundation model you can support multiple
targeted tasks as opposed sticking with the analogy to having
a model for reading, writing, thinking of doing a poem,
doing an essay, so on and so forth. And so
in the enterprise context, that means that we're now talking
(08:42):
about being able to unlock even additional value at scale
because of the nation of nature foundation models and their
appeal to generative use cases. Generative in this case meaning
creation of new content.
Speaker 3 (08:58):
So let's talk about what's in X. IBM recently announced
what's an X. Just first of all, what is that?
What is what's an X?
Speaker 4 (09:06):
Well, what's an X refers to our is our brand
for our platform, the WhatsApp platform for really taking advantage
of generative AI within the enterprise, within business. And so
when you begin to think about what does that mean,
while it leads you to the components of what's an
X and to a set of use cases. So let
(09:27):
me paint a few quick pictures for you here. What's
an X first of all, is about enabling our customers
to manipulate models against their task, manipulate these foundation models
against their task. Our belief is that the world is
a multi model world, right and especially when you think
(09:48):
about it in the context of business. Models are going
to come from various sources, the ones we supply, the
ones out there in open source, and so of you.
But there are activities you need to do around these
models to as I said, apply them to your use case.
And we'll talk about use cases in a bit. So
what's next. Dot AI is that environment that build a
(10:09):
tool if you like, for being able to do those
manipulation of models to meet your specific use case. Thinks
that people will recognize in the field prompt engineering, prompt tuning,
fine tuning, those kinds of activities which are all the
manipulation of models to meet your use case. Yeah. The
second component is dot data, So what's the next? Dot
(10:30):
data is essentially a next generation data store is based
upon something referred to as an open data lakehouse architecture
that helps to bring together the data that's needed to
actually do the AI. In this case, when you think
about manipulating a model, a foundation model, you're generally using
some data to prompt it, tune it, to train it
(10:51):
to your use cases. And so we provide a very
open data store that allows all manner of data and
formats to be brought through. Today you that and the
third component is what's next up governance And this is
all about the framework and the toolkit required to apply
the right governance principles across doing this kind of work,
(11:14):
because when you're deploying AI within the enterprise, governance is
actually important, right, It's critical to understand why is your
data coming from? What data did you add in? How
is your model performing? Are you able to keep an
appropriate audit trail of your activities for your own internal
policy and compliance needs or for regulatory needs as well.
Speaker 3 (11:36):
So this platform, the system that you're describing, I'm curious,
how is it different from the you know, the generative
AI options that you know we've all been hearing about
sort of in the press.
Speaker 4 (11:48):
Well, I think it really comes down to the ethos
or the principles that first of all drive the work
that we're doing. The first I would fixate on is
being open. Right. We fundamentally believe that to do this
kind of work within the enterprise, you need an open
platform that, as I said, is open to all manner
(12:09):
of models from all sources. It's one of the reasons
why we announced our partnership with hugging Face to make
sure that our clients can gain access to open source
innovation within the platform to do their work.
Speaker 3 (12:22):
And hugging Face, to be clear, is sort of the
open source AI kind of hub.
Speaker 4 (12:27):
That's right, that's correct. Yes, it's a marketplace hub for
all kind of ecosystem coordinator for open source models. And
I believe there's a lot of innovation going on out there. So,
first of all, open becomes important. The second targeted So
our focus is very much on enabling these business use cases, right,
(12:51):
And you might say what kind of use cases are
we talking about? I give you three very quick ones
that with our customers are focused on a lot of
focus around and enhancing customer service use cases. Think of
this as chatbots or digital assistance that are further trained
in more and more information about what the company has
(13:11):
to offer, or could be internal policies, external policy, so
on and so forth. This means a platform that makes
it really easy to bring your own data to train
and tune the model, while protecting your own data as
extremely important for the enterprise right. Another important use case
(13:32):
seeing a lot of focused on what i'd call AI
based orchestration or automation of task whereby think about like
an HR professional as an example, going through a job
requisition is able to interact with multiple systems via a
very simple chat interface and have work dynamically sequenced to
(13:52):
support them in doing their task. That once again requires
a notion of working with models and technology in a
way that in many ways can be unique to how
a business wishes to work and indeed, in various cases
can embody what they consider their their secret source or
their differentiated advantage. So once again, a platform that recognizes
(14:13):
that and designed for business that's not the same scope
or frame of reference for a consumer platform. And then
you know, we're also seeing a lot of work around
cod generation, application modernization, you know, and people enhancing their skills.
So targeted becomes really important. Mentioned open and I mentioned
targeted targeted to the business to the use cases that
(14:36):
they need to do. Underpinning that then it's trusted. So
everything I gave you in those targeted use cases talk
about handling enterprise proprietary and specific data. We are trusted
in this regard right. We have been serving the business
for many, many a year, and we are designing our
(14:57):
platform and even our principles and way of creating to
recognize and enable that. Both in terms of the work
we do around the governance framework and transparency that you're
able to gain and apply, but even in the way
we allow our platform to be deployed in multiple kind
of locations, of footprints consumed as a service on a hyperscaler,
(15:19):
running your own private footprint on prem or your cloud footprint.
All of these need to be brought together to meet
the needs of an actual enterprise business. My last comment
is where I think we're fundamentally differentiated is we're really
about empowering our customers to take advantage of AI to
(15:41):
unleash the intelligence, capabilities productivity of their own business. This
isn't about, oh, we've established a bunch of APIs that
you can ask questions. This is about how do you
craft what you need for your business to deliver different
shaped value to your customers, shareholders, employees with all the
(16:06):
appropriate protections as well. And so there's a lot of
focus on what we've done with the platform and the
tool set to enable that, to enable what we like
to call AI value creators, not just consumers of AI value.
Speaker 3 (16:20):
When you were talking about basically enterprise adoption of AI,
you use the word trust, and I'm curious, you know,
what does trust mean in the context of AI and
the enterprise.
Speaker 4 (16:37):
I would kind of deconstruct trust along these k avenues.
If AI is about giving you insights to help you
support decisions, how do you trust what insights it's provided?
What data did it use? What did it consider based
(17:01):
upon that data that therefore led to the insight provided?
Why is this important? Why this notion of trust? Well, One,
you're about to make a decision, so you want to
understand the basis for a decision. It's no different than
(17:22):
me asking you something and then saying, okay, can you
explain you're working? Right, that would be a notion of
trust that we establish and a very natural interaction as humans, right,
we do it all the time, right, So there is
that element. The other reason why it becomes important if
you're applying AI into business processes and therefore how your
(17:43):
business works. You want to make sure that you know
what biases are built in to any decision or not
or if the AI the model in effect is drifting
away from kind of the parameters that you would want
it to remain within, right or go trust and so
(18:07):
in many ways, that's one big aspect of trusting the
technology because you're applying it into decisions you need to
make every day, and you need to know in very
simple terms how it works and how it is working.
The element of trust that I think is important in
this discussion. Who are you getting your AI from? That's
(18:32):
very important to us as a company here at IBM. Right,
given we serve business, that trust becomes extremely important and
what are the elements of that trust? What are the
customers trying to understand? Well, first and foremost, what's your
ethos around AI? We're very clear on the customer's data
(18:53):
is their data when they tune or refine those models
to meet their use cases. That is all THEIRS actually
provide the ability for them to do that in very
isolated and protected ways as they choose, and we never
use that data without explicit opting and permissions. Right, customers
(19:13):
might say Oh yeah, use this so that you can
make a generally overall better model. But it's full awareness,
full transparency that is important. That's a trust of who
you're doing business with. So that's how I think about trust.
How do you build systems you trust? And are you
working with people you trust?
Speaker 2 (19:35):
I find Kareem's point about trust when it comes to
data to be so important because as powerful as AI
tools can be, their helpfulness is dependent on how trustworthy
the data is. Humans will have to decide if our data,
our decision making, and our AI insights live up to
the vision we hope to achieve in business. As Green
(19:55):
and Jacob continue the conversation, Jacob asks some more practical
questions about how businesses can adopt AI into their own processes.
Speaker 3 (20:05):
Let's listen, how can businesses move toward integrating AI as
part of their core business model instead of, you know,
sort of as an add on on the periphery.
Speaker 4 (20:17):
It's funny, you know. My simple answer to that is
it's actually the simplest thing in the world to do.
By thinking about your business, thinking about your elements of differentiation,
and then thinking about how AI can help you extend
(20:37):
expand those Right, what do you want to be known for.
I picked a very simple use case of customer service interaction.
Almost every business needs to do that and wants to
do it better, and so it becomes a way to stop.
But then as you begin to work your way through,
you think about various automation of business processes. You think
about decisions that need to be made right or how
(20:58):
can individuals be helped, how can they made more productive?
I think always becomes a very important one. Right, So,
and you can apply this in many context a financial
analyst looking at reams of data and trying to derive insights.
How do you leverage AI to make that financial analyst
even more powerful? And so that's how I advise you, know,
people to always look at it. Think about your task,
(21:20):
think about your business processes, think about where help is
needed or where new value could be unlocked, and then
you're applying AI as a tool to achieve that end.
Speaker 3 (21:31):
One of the themes we return to on this show
a lot is creativity and the relationship between technology and creativity.
And I'm curious how you think that AI can help
people be more creative at work.
Speaker 4 (21:50):
I think AI can help people be more creative at
work by automating the mundane to unlock your mind to
be able to focus on higher value. You know, I've
used a couple of times I've talked about deriving insights
from data right to drive informed decisions. If you can
use AI to gather a lot more insights into one place,
(22:14):
then you could typically do yourself or more manually, you'd
have to like write it down, look at six spreadsheets,
copy from here to there. Then you actually have more
time to look at that data, digest those insights, and
think about what do I need to do with these
as a business, which direction do I want to go?
I think of its freeing us up to do more
(22:35):
of what we actually as humans do extremely well, which
is actually that creative thinking exactly simple terms. Why do
we use a calculator to do arithmetic? It's not that
we cannot necessarily knock it out ourselves. But if you're
trying to balance your checkbook, to use an old phrase
or dare I say, what's a check but so modernize that.
(23:02):
If you're trying to check your expenses for the month
and your performance against budget, yes you could print out
all your statements, circle everything and add it all up.
Or you could begin to use technology to improve that experience,
so you can get more time to think about what
(23:24):
really am I learning from my spending patterns and what
do I want to do about it. It's a very
simple personal example, but I think it's fundamentally what we're
talking about here, and that's always been in my mind,
the promise of technology freeing us up to actually apply
ourselves to higher value thought and higher value problems.
Speaker 3 (23:46):
So we've been talking basically about the present so far,
and I'm curious if if you think about the future
and you think, you know, medium to long term, how
do you think AI is going to transform business? And
you know, how can people now, business leaders now prepare
for what's coming.
Speaker 4 (24:05):
So to an earlier comment I made, I do really
think that we are at an inflection point with the
advancement of the technologies of AI. I talked about foundation models.
We definitely at the cusp of being able to address
(24:25):
use cases at scale that were more challenging before, and
so I do think the future looks like a lot
more generative AI surfacing within the enterprise and within business
processes and manifesting in interesting ways. I think it's almost
(24:47):
a given that any piece of software right think, whether
you think of it in terms of an application or
you think about it in terms of you know, the
interact with the website will have conversational enabled interfaces from
the analyst saying give me the latest reports for the
last three months, you know, typing that or saying it
(25:10):
versus the right click file blah blah. I think you're
going to see that change in interaction to more conversational interaction.
I think, particularly chat based.
Speaker 3 (25:22):
We forget that the graphical user interface is just a metaphor, right,
It's not like the way computers work. It's just an interface.
And if chat is a better interface, people will use chat.
Speaker 4 (25:32):
And I think we're going to see that rarely explode.
And that's powered by a lot of this generative AI
work because it becomes for it to feel natural, for
it to be as informed, to readily, as I said,
link things to get and orchestrate. That's a big part.
So I think I see that happening and the appropriate
or associated productivity on locks you begin to see with
(25:53):
that will just change what kind of decisions, the ease
with which we can make more and more formed business decisions.
And so for me, it's that rolling out at scale,
touching everything, procurement, hr think about the advent of the
spreadsheet and how many different roles it just ended up
(26:18):
touching and everybody can use or does user spreadsheeting business
in some shape, size or form. So I think of
this as AI at scale. And so what it therefore
means from as you said, getting prepared, Well, it's all
about gaining first of all, the right understanding of the technologies.
And part of what a lot we'll be talking about
(26:40):
necessary ingredients began to be well, where do I want
to apply it first? What data do I need to
bring together to readily support that? What unlocks what new value?
And I think it's going to be like this rollout right,
you got to start with this project and then there's
another project, and very soon it will be so much
it will be ubiquitu just in the way it supports
(27:02):
the work we need to do. That it will just
speak to a new way of us working that is,
when you now look back, will be pretty different from
how we work today. You see the seeds today, but
I would argue, think of that now like fully bloomed.
It's a forest, not a not a flowerbed. You know, yeah, yeah, yeah, great.
Speaker 3 (27:25):
One other one other sort of loose thread I want
to I want to return to UH and that's that's governance. Right,
you talked about governance and maybe just just to help
sort of set the table, like you mentioned it in
a broadway but narrowly, what does governance mean in the
context of IBM's work on enterprise AHI.
Speaker 4 (27:46):
I think, as the word tries to suggest, it is
about having the way to govern one's activity in this realm,
which really speaks to policies, rules and frameworks within which
(28:10):
to understand all of that. Now, before we dive in
the direction of regulation, which is where people often go,
policies can be all internal. So think about it this way.
If I say to you, when I build AI, I
(28:30):
do not use my customer's data. Is their customer's data,
Then from a governance perspective, I need processes that ensure
I know what data I'm using and I can prove
to myself just first of all internally, forget about anybody else,
that I'm actually adhering to the policies I've laid out.
(28:53):
That in my mind is a lot of what governance
is about, and in the context of AI, it always
tends to I think structure around three key areas data
where did it come from? And what did I do
with it? And how did I apply it? And where
did I use it? And then usage, what do I
expect this model to do? Is this model still performing
(29:16):
the way I think it should be performing? What are
my processes to address whether they answered that question is
yes or no? And manage that through. And then importantly
so this is then to bridge to regulation. If you
take a look at what's going on in the world
of AI regulation and our point of view on this,
(29:38):
by the way, is that you actually regulate the use cases,
not the technology. Then from a governance perspective, how are
you able to clearly understand, track and account for what
use cases you are leveraging AI for? And then back
to my earlier comments how that AI.
Speaker 3 (29:57):
Is performing and when you talk about how do you
make sure that you have the governance you need without
inhibiting innovation?
Speaker 4 (30:06):
I think what is key and this is key A
key design point for what we're doing with What's the
next is how you make governance seamless institute versus another
activity that you do right. And so our goal is
to try and drive that kind of seamless interactions or
(30:31):
value add in terms of governance, so that when oh,
let's pull through the history right of everything we've done here,
or what prompts we've created, or what data we've used,
it's kind of already there, right, and so you can
feel free to be innovating and testing out your different
prompts and all that stuff, or bringing in your data
(30:54):
sets without saying, oh, before I do that, I need
to make sure I run this checker. And now you
can kind of bring it systems kind of automatically categorizing it,
and then you can go in a lead very five,
validate or explore, say I'm no longer going to take
this path based upon these facts. I think the more
we can make it more of a natural extension of
(31:14):
the activities that need to be done, the more we
can make it then just a part of what needs
to be done. And as you're to your point, gain
our governance needs or supports the governance needs of our
customers without stifling the innovation of the individuals at the
glass trying to think through I iteratively think through new
(31:35):
value ways to do work excellent.
Speaker 3 (31:39):
Let me ask you are there things I didn't ask
you that I should? Are there things you want to
talk about that we didn't talk about.
Speaker 4 (31:45):
I think we covered quite a lot true it. Oh No,
I think we we covered the bases there.
Speaker 2 (31:54):
Earlier, Green mentioned that we are at an inflection point
in AI technology. Implementing a in business will get easier,
and AI platforms like Watson x can empower even the
largest enterprise businesses to reinvent the way they run. As
Greem said, in the same way the spreadsheet took over
business operations, the adoption of AI at enterprise scale could
(32:18):
be just as ubiquitous. It's not an overstatement to say
that a new era of work may be upon us.
I'm Malcolm Gladwell. This is a paid advertisement from IBM.
Smart Talks with IBM is produced by Matt Romano, David
jaw Nisha Venkat and Royston Deserve with Jacob Goldstein. We're
(32:42):
edited by Lydia gene Kott. Our engineers are Jason Gambrel,
Sarah Brugaier and Ben Tolliday. Theme song by Gramoscope. Special
thanks to Carli Migliore, Andy Kelly, Kathy Callahan and eight
Bar and the eight Bar and IBM teams, as well
as the Pushkin marketing team. Smart Talks with IBM is
a production of Pushkin Industries and Ruby Studio at iHeartMedia.
(33:07):
To find more Pushkin podcasts, listen on the iHeartRadio app,
Apple Podcasts, or wherever you listen to podcasts.
Welcome to Tech Stuff, a production from iHeartRadio. Today, we
are witnessed to one of those rare moments in history,
the rise of an innovative technology with the potential to
radically transform business in society forever. That technology, of course,
(00:24):
is artificial intelligence, and it's the central focus for this
new season of Smart Talks with IBM. Join hosts from
your favorite Pushkin podcasts as they talk with industry experts
and leaders to explore how businesses can integrate AI into
their workflows and help drive real change in this new
era of AI, and of course, host Malcolm Gladwell will
(00:46):
be there to guide you through the season and throw
in his two cents as well. Look out for new
episodes of Smart Talks with IBM every other week on
the iHeartRadio app, Apple Podcasts, or wherever you get your podcasts,
and learn more at IBM dot com slash smart Talks.
Speaker 2 (01:05):
Hello, Hello, Welcome to Smart Talks with IBM, a podcast
from Pushkin Industries, iHeartRadio and IBM. I'm Malcolm Glabo. This season,
we're continuing our conversation with new creators visionaries who are
creatively applying technology in business to drive change, but with
a focus on the transformative power of artificial intelligence and
(01:27):
what it means to leverage AI as a game changing
multiplier for your business. Our guest today is Kareem Yousef,
Senior Vice President of Product Management and Growth for IBM Software.
Kareem's focus at IBM is on product strategy, thinking about
the roadmap for IBM Software products and how they can
deliver effective and compelling customer experiences with the current boom
(01:53):
and generative AI. Kareem's job is to help businesses figure
out how they can apply artificial intelligence at scale to
help solve big problems and boost productivity at new orders
of magnitude. In today's episode, you'll hear Kareem explain how
AI powered by foundation models can make AI adoption by
enterprise businesses even easier, how generative AI will change the
(02:18):
way businesses process data and make decisions, and how these
considerations influence the design of Watson x, IBM's next generation
AI and data platform. Kareem spoke with Jacob Goldstein, host
of the Pushkin podcast What's Your Problem. A veteran business journalist,
Jacob has reported for The Wall Street Journal, the Miami Herald,
(02:41):
and was a longtime host of the NPR program Planet Money. Okay,
let's get to the interview.
Speaker 3 (02:49):
I'm Jacob Goldstein. I'm one of the hosts at Pushkin
and a correspondent on this show, and I'm delighted to
have you here. Can you introduce yourself?
Speaker 4 (02:57):
Ah? Hi, I'm Kareem Yusuf. I'm this and your vice
president of Product management and Growth for IBM Software. You
can think of me as the chief product officer for
IBM Software.
Speaker 3 (03:07):
Okay, sounds like a big job. We're here today to
talk about AI. We've heard really an extraordinary amount in
the last few months about chat GPT and you know,
particularly in how it's used in the very kind of
consumer facing way. But I'm curious what is the rise
(03:27):
of chat GPT and you know, AI more generally, what
does it mean for business?
Speaker 4 (03:32):
Well, you know, it's if you kind of step back
and think about what really happens. You know, in a business,
you're really talking about a set of processes, right, you know,
activities that represent what a business needs to get done,
whether it's product they produce and then sell or service
that they provide. And inherent to operating the business, I
(03:54):
would say are two very key factors. Data and then
the decisions you make around data and then actually lastly
the processes or activities you do in accordance with that decision.
So if you then think about AI as applied to
business right in that context, well, the first place it
often starts is how do you make sense of a
(04:16):
lot of the data associated with driving the business? And
so AI has always been, in my mind at its
foremost about gaining insights then lead in to supporting decisions,
and ultimately ending at helping to automate the activities that
then are executed as a result of those decisions. So
(04:38):
that's kind of my simple way of thinking of AI,
and we can obviously coloring with examples, but that's my
simplest way of thinking about AI. When you think about
in the business context, gain insights from masses of data
to support decisions and then drive.
Speaker 3 (04:52):
Actions, that's a really helpful framework. And then if we
think about sort of what's happening in the world now
with you know, enterprise businesses NAI, what are you seeing
with enterprise adoption of AI at this moment?
Speaker 4 (05:06):
So we're really talking about almost a tale of two periods.
So let me first of all kind of take you
back before the advent of what I would call generative AI,
and the whole chat gpt to what has been going
on in what I would term the realm of more
standardized machine learning models. A lot of what has been
(05:26):
going on has been very much in the realms of
certain things like anomaly detection or optimization, right, using machine
learning models to do that kind of work, and where
might it apply well, think of anomaly detection in security
software right detecting threats based upon different events flowing through
(05:47):
or in enterprise asset management software monitoring equipment and detecting
anomalies within their behavior, or even in IT automation software
once again detecting anomalies based upon what's going on with
various IT events and then tasks that should occur. Optimizations
often play around in the realm as you might imagine
(06:09):
to solve problems of resource optimization, whether you think of
that in the context of application resource management for IT
or in the context of supply chain. These have been
very classical applications of machine learning AI to really make
sense of the data and provide a basis to drive decisions. Now,
(06:30):
what is characterized by all those examples have given and
the state of the art of that kind of technology
has always been it's very task specific. So there was
a air quotes, if I may, kind of limitation in
the sense that the tak it had to be very
(06:50):
task specific. And so we've seen a lot of broad
based adoption within the enterprise, right, but it's very very
task specific. As you might imagine. Now, what has happened
recently and has been brought to the four has been
this discussion of generative AI, which is powered by a
very specific innovation, this notion of foundation models. And in
(07:14):
the simplest way to think about it, it's about training
this large model that can then be refined to various tasks.
And the easiest one that everybody recognized at the moment
is the notion of a large language model, a model
that has an understanding of a lot of the elements
(07:37):
of a language such that it can be refined to
a variety of tasks. Write an essay, answer a question,
singer songs, so on, answers so forth. I like to
liken the power if you like, and this will speak
to the why everybody is so excited about it. Why
would argue at an inflection point? I like to liken
(07:58):
it to teaching a child the alphabet. When you teach
a child and alphabet, it's a set of letters. Right,
Let's call that our foundation model. But over time that
knowledge of the alphabet is tuned to read a book,
write an essay, do a composition, create a song, write
a poem, write an invoice. You understand what I mean, right,
(08:21):
And so from one foundation model you can support multiple
targeted tasks as opposed sticking with the analogy to having
a model for reading, writing, thinking of doing a poem,
doing an essay, so on and so forth. And so
in the enterprise context, that means that we're now talking
(08:42):
about being able to unlock even additional value at scale
because of the nation of nature foundation models and their
appeal to generative use cases. Generative in this case meaning
creation of new content.
Speaker 3 (08:58):
So let's talk about what's in X. IBM recently announced
what's an X. Just first of all, what is that?
What is what's an X?
Speaker 4 (09:06):
Well, what's an X refers to our is our brand
for our platform, the WhatsApp platform for really taking advantage
of generative AI within the enterprise, within business. And so
when you begin to think about what does that mean,
while it leads you to the components of what's an
X and to a set of use cases. So let
(09:27):
me paint a few quick pictures for you here. What's
an X first of all, is about enabling our customers
to manipulate models against their task, manipulate these foundation models
against their task. Our belief is that the world is
a multi model world, right and especially when you think
(09:48):
about it in the context of business. Models are going
to come from various sources, the ones we supply, the
ones out there in open source, and so of you.
But there are activities you need to do around these
models to as I said, apply them to your use case.
And we'll talk about use cases in a bit. So
what's next. Dot AI is that environment that build a
(10:09):
tool if you like, for being able to do those
manipulation of models to meet your specific use case. Thinks
that people will recognize in the field prompt engineering, prompt tuning,
fine tuning, those kinds of activities which are all the
manipulation of models to meet your use case. Yeah. The
second component is dot data, So what's the next? Dot
(10:30):
data is essentially a next generation data store is based
upon something referred to as an open data lakehouse architecture
that helps to bring together the data that's needed to
actually do the AI. In this case, when you think
about manipulating a model, a foundation model, you're generally using
some data to prompt it, tune it, to train it
(10:51):
to your use cases. And so we provide a very
open data store that allows all manner of data and
formats to be brought through. Today you that and the
third component is what's next up governance And this is
all about the framework and the toolkit required to apply
the right governance principles across doing this kind of work,
(11:14):
because when you're deploying AI within the enterprise, governance is
actually important, right, It's critical to understand why is your
data coming from? What data did you add in? How
is your model performing? Are you able to keep an
appropriate audit trail of your activities for your own internal
policy and compliance needs or for regulatory needs as well.
Speaker 3 (11:36):
So this platform, the system that you're describing, I'm curious,
how is it different from the you know, the generative
AI options that you know we've all been hearing about
sort of in the press.
Speaker 4 (11:48):
Well, I think it really comes down to the ethos
or the principles that first of all drive the work
that we're doing. The first I would fixate on is
being open. Right. We fundamentally believe that to do this
kind of work within the enterprise, you need an open
platform that, as I said, is open to all manner
(12:09):
of models from all sources. It's one of the reasons
why we announced our partnership with hugging Face to make
sure that our clients can gain access to open source
innovation within the platform to do their work.
Speaker 3 (12:22):
And hugging Face, to be clear, is sort of the
open source AI kind of hub.
Speaker 4 (12:27):
That's right, that's correct. Yes, it's a marketplace hub for
all kind of ecosystem coordinator for open source models. And
I believe there's a lot of innovation going on out there. So,
first of all, open becomes important. The second targeted So
our focus is very much on enabling these business use cases, right,
(12:51):
And you might say what kind of use cases are
we talking about? I give you three very quick ones
that with our customers are focused on a lot of
focus around and enhancing customer service use cases. Think of
this as chatbots or digital assistance that are further trained
in more and more information about what the company has
(13:11):
to offer, or could be internal policies, external policy, so
on and so forth. This means a platform that makes
it really easy to bring your own data to train
and tune the model, while protecting your own data as
extremely important for the enterprise right. Another important use case
(13:32):
seeing a lot of focused on what i'd call AI
based orchestration or automation of task whereby think about like
an HR professional as an example, going through a job
requisition is able to interact with multiple systems via a
very simple chat interface and have work dynamically sequenced to
(13:52):
support them in doing their task. That once again requires
a notion of working with models and technology in a
way that in many ways can be unique to how
a business wishes to work and indeed, in various cases
can embody what they consider their their secret source or
their differentiated advantage. So once again, a platform that recognizes
(14:13):
that and designed for business that's not the same scope
or frame of reference for a consumer platform. And then
you know, we're also seeing a lot of work around
cod generation, application modernization, you know, and people enhancing their skills.
So targeted becomes really important. Mentioned open and I mentioned
targeted targeted to the business to the use cases that
(14:36):
they need to do. Underpinning that then it's trusted. So
everything I gave you in those targeted use cases talk
about handling enterprise proprietary and specific data. We are trusted
in this regard right. We have been serving the business
for many, many a year, and we are designing our
(14:57):
platform and even our principles and way of creating to
recognize and enable that. Both in terms of the work
we do around the governance framework and transparency that you're
able to gain and apply, but even in the way
we allow our platform to be deployed in multiple kind
of locations, of footprints consumed as a service on a hyperscaler,
(15:19):
running your own private footprint on prem or your cloud footprint.
All of these need to be brought together to meet
the needs of an actual enterprise business. My last comment
is where I think we're fundamentally differentiated is we're really
about empowering our customers to take advantage of AI to
(15:41):
unleash the intelligence, capabilities productivity of their own business. This
isn't about, oh, we've established a bunch of APIs that
you can ask questions. This is about how do you
craft what you need for your business to deliver different
shaped value to your customers, shareholders, employees with all the
(16:06):
appropriate protections as well. And so there's a lot of
focus on what we've done with the platform and the
tool set to enable that, to enable what we like
to call AI value creators, not just consumers of AI value.
Speaker 3 (16:20):
When you were talking about basically enterprise adoption of AI,
you use the word trust, and I'm curious, you know,
what does trust mean in the context of AI and
the enterprise.
Speaker 4 (16:37):
I would kind of deconstruct trust along these k avenues.
If AI is about giving you insights to help you
support decisions, how do you trust what insights it's provided?
What data did it use? What did it consider based
(17:01):
upon that data that therefore led to the insight provided?
Why is this important? Why this notion of trust? Well, One,
you're about to make a decision, so you want to
understand the basis for a decision. It's no different than
(17:22):
me asking you something and then saying, okay, can you
explain you're working? Right, that would be a notion of
trust that we establish and a very natural interaction as humans, right,
we do it all the time, right, So there is
that element. The other reason why it becomes important if
you're applying AI into business processes and therefore how your
(17:43):
business works. You want to make sure that you know
what biases are built in to any decision or not
or if the AI the model in effect is drifting
away from kind of the parameters that you would want
it to remain within, right or go trust and so
(18:07):
in many ways, that's one big aspect of trusting the
technology because you're applying it into decisions you need to
make every day, and you need to know in very
simple terms how it works and how it is working.
The element of trust that I think is important in
this discussion. Who are you getting your AI from? That's
(18:32):
very important to us as a company here at IBM. Right,
given we serve business, that trust becomes extremely important and
what are the elements of that trust? What are the
customers trying to understand? Well, first and foremost, what's your
ethos around AI? We're very clear on the customer's data
(18:53):
is their data when they tune or refine those models
to meet their use cases. That is all THEIRS actually
provide the ability for them to do that in very
isolated and protected ways as they choose, and we never
use that data without explicit opting and permissions. Right, customers
(19:13):
might say Oh yeah, use this so that you can
make a generally overall better model. But it's full awareness,
full transparency that is important. That's a trust of who
you're doing business with. So that's how I think about trust.
How do you build systems you trust? And are you
working with people you trust?
Speaker 2 (19:35):
I find Kareem's point about trust when it comes to
data to be so important because as powerful as AI
tools can be, their helpfulness is dependent on how trustworthy
the data is. Humans will have to decide if our data,
our decision making, and our AI insights live up to
the vision we hope to achieve in business. As Green
(19:55):
and Jacob continue the conversation, Jacob asks some more practical
questions about how businesses can adopt AI into their own processes.
Speaker 3 (20:05):
Let's listen, how can businesses move toward integrating AI as
part of their core business model instead of, you know,
sort of as an add on on the periphery.
Speaker 4 (20:17):
It's funny, you know. My simple answer to that is
it's actually the simplest thing in the world to do.
By thinking about your business, thinking about your elements of differentiation,
and then thinking about how AI can help you extend
(20:37):
expand those Right, what do you want to be known for.
I picked a very simple use case of customer service interaction.
Almost every business needs to do that and wants to
do it better, and so it becomes a way to stop.
But then as you begin to work your way through,
you think about various automation of business processes. You think
about decisions that need to be made right or how
(20:58):
can individuals be helped, how can they made more productive?
I think always becomes a very important one. Right, So,
and you can apply this in many context a financial
analyst looking at reams of data and trying to derive insights.
How do you leverage AI to make that financial analyst
even more powerful? And so that's how I advise you, know,
people to always look at it. Think about your task,
(21:20):
think about your business processes, think about where help is
needed or where new value could be unlocked, and then
you're applying AI as a tool to achieve that end.
Speaker 3 (21:31):
One of the themes we return to on this show
a lot is creativity and the relationship between technology and creativity.
And I'm curious how you think that AI can help
people be more creative at work.
Speaker 4 (21:50):
I think AI can help people be more creative at
work by automating the mundane to unlock your mind to
be able to focus on higher value. You know, I've
used a couple of times I've talked about deriving insights
from data right to drive informed decisions. If you can
use AI to gather a lot more insights into one place,
(22:14):
then you could typically do yourself or more manually, you'd
have to like write it down, look at six spreadsheets,
copy from here to there. Then you actually have more
time to look at that data, digest those insights, and
think about what do I need to do with these
as a business, which direction do I want to go?
I think of its freeing us up to do more
(22:35):
of what we actually as humans do extremely well, which
is actually that creative thinking exactly simple terms. Why do
we use a calculator to do arithmetic? It's not that
we cannot necessarily knock it out ourselves. But if you're
trying to balance your checkbook, to use an old phrase
or dare I say, what's a check but so modernize that.
(23:02):
If you're trying to check your expenses for the month
and your performance against budget, yes you could print out
all your statements, circle everything and add it all up.
Or you could begin to use technology to improve that experience,
so you can get more time to think about what
(23:24):
really am I learning from my spending patterns and what
do I want to do about it. It's a very
simple personal example, but I think it's fundamentally what we're
talking about here, and that's always been in my mind,
the promise of technology freeing us up to actually apply
ourselves to higher value thought and higher value problems.
Speaker 3 (23:46):
So we've been talking basically about the present so far,
and I'm curious if if you think about the future
and you think, you know, medium to long term, how
do you think AI is going to transform business? And
you know, how can people now, business leaders now prepare
for what's coming.
Speaker 4 (24:05):
So to an earlier comment I made, I do really
think that we are at an inflection point with the
advancement of the technologies of AI. I talked about foundation models.
We definitely at the cusp of being able to address
(24:25):
use cases at scale that were more challenging before, and
so I do think the future looks like a lot
more generative AI surfacing within the enterprise and within business
processes and manifesting in interesting ways. I think it's almost
(24:47):
a given that any piece of software right think, whether
you think of it in terms of an application or
you think about it in terms of you know, the
interact with the website will have conversational enabled interfaces from
the analyst saying give me the latest reports for the
last three months, you know, typing that or saying it
(25:10):
versus the right click file blah blah. I think you're
going to see that change in interaction to more conversational interaction.
I think, particularly chat based.
Speaker 3 (25:22):
We forget that the graphical user interface is just a metaphor, right,
It's not like the way computers work. It's just an interface.
And if chat is a better interface, people will use chat.
Speaker 4 (25:32):
And I think we're going to see that rarely explode.
And that's powered by a lot of this generative AI
work because it becomes for it to feel natural, for
it to be as informed, to readily, as I said,
link things to get and orchestrate. That's a big part.
So I think I see that happening and the appropriate
or associated productivity on locks you begin to see with
(25:53):
that will just change what kind of decisions, the ease
with which we can make more and more formed business decisions.
And so for me, it's that rolling out at scale,
touching everything, procurement, hr think about the advent of the
spreadsheet and how many different roles it just ended up
(26:18):
touching and everybody can use or does user spreadsheeting business
in some shape, size or form. So I think of
this as AI at scale. And so what it therefore
means from as you said, getting prepared, Well, it's all
about gaining first of all, the right understanding of the technologies.
And part of what a lot we'll be talking about
(26:40):
necessary ingredients began to be well, where do I want
to apply it first? What data do I need to
bring together to readily support that? What unlocks what new value?
And I think it's going to be like this rollout right,
you got to start with this project and then there's
another project, and very soon it will be so much
it will be ubiquitu just in the way it supports
(27:02):
the work we need to do. That it will just
speak to a new way of us working that is,
when you now look back, will be pretty different from
how we work today. You see the seeds today, but
I would argue, think of that now like fully bloomed.
It's a forest, not a not a flowerbed. You know, yeah, yeah, yeah, great.
Speaker 3 (27:25):
One other one other sort of loose thread I want
to I want to return to UH and that's that's governance. Right,
you talked about governance and maybe just just to help
sort of set the table, like you mentioned it in
a broadway but narrowly, what does governance mean in the
context of IBM's work on enterprise AHI.
Speaker 4 (27:46):
I think, as the word tries to suggest, it is
about having the way to govern one's activity in this realm,
which really speaks to policies, rules and frameworks within which
(28:10):
to understand all of that. Now, before we dive in
the direction of regulation, which is where people often go,
policies can be all internal. So think about it this way.
If I say to you, when I build AI, I
(28:30):
do not use my customer's data. Is their customer's data,
Then from a governance perspective, I need processes that ensure
I know what data I'm using and I can prove
to myself just first of all internally, forget about anybody else,
that I'm actually adhering to the policies I've laid out.
(28:53):
That in my mind is a lot of what governance
is about, and in the context of AI, it always
tends to I think structure around three key areas data
where did it come from? And what did I do
with it? And how did I apply it? And where
did I use it? And then usage, what do I
expect this model to do? Is this model still performing
(29:16):
the way I think it should be performing? What are
my processes to address whether they answered that question is
yes or no? And manage that through. And then importantly
so this is then to bridge to regulation. If you
take a look at what's going on in the world
of AI regulation and our point of view on this,
(29:38):
by the way, is that you actually regulate the use cases,
not the technology. Then from a governance perspective, how are
you able to clearly understand, track and account for what
use cases you are leveraging AI for? And then back
to my earlier comments how that AI.
Speaker 3 (29:57):
Is performing and when you talk about how do you
make sure that you have the governance you need without
inhibiting innovation?
Speaker 4 (30:06):
I think what is key and this is key A
key design point for what we're doing with What's the
next is how you make governance seamless institute versus another
activity that you do right. And so our goal is
to try and drive that kind of seamless interactions or
(30:31):
value add in terms of governance, so that when oh,
let's pull through the history right of everything we've done here,
or what prompts we've created, or what data we've used,
it's kind of already there, right, and so you can
feel free to be innovating and testing out your different
prompts and all that stuff, or bringing in your data
(30:54):
sets without saying, oh, before I do that, I need
to make sure I run this checker. And now you
can kind of bring it systems kind of automatically categorizing it,
and then you can go in a lead very five,
validate or explore, say I'm no longer going to take
this path based upon these facts. I think the more
we can make it more of a natural extension of
(31:14):
the activities that need to be done, the more we
can make it then just a part of what needs
to be done. And as you're to your point, gain
our governance needs or supports the governance needs of our
customers without stifling the innovation of the individuals at the
glass trying to think through I iteratively think through new
(31:35):
value ways to do work excellent.
Speaker 3 (31:39):
Let me ask you are there things I didn't ask
you that I should? Are there things you want to
talk about that we didn't talk about.
Speaker 4 (31:45):
I think we covered quite a lot true it. Oh No,
I think we we covered the bases there.
Speaker 2 (31:54):
Earlier, Green mentioned that we are at an inflection point
in AI technology. Implementing a in business will get easier,
and AI platforms like Watson x can empower even the
largest enterprise businesses to reinvent the way they run. As
Greem said, in the same way the spreadsheet took over
business operations, the adoption of AI at enterprise scale could
(32:18):
be just as ubiquitous. It's not an overstatement to say
that a new era of work may be upon us.
I'm Malcolm Gladwell. This is a paid advertisement from IBM.
Smart Talks with IBM is produced by Matt Romano, David
jaw Nisha Venkat and Royston Deserve with Jacob Goldstein. We're
(32:42):
edited by Lydia gene Kott. Our engineers are Jason Gambrel,
Sarah Brugaier and Ben Tolliday. Theme song by Gramoscope. Special
thanks to Carli Migliore, Andy Kelly, Kathy Callahan and eight
Bar and the eight Bar and IBM teams, as well
as the Pushkin marketing team. Smart Talks with IBM is
a production of Pushkin Industries and Ruby Studio at iHeartMedia.
(33:07):
To find more Pushkin podcasts, listen on the iHeartRadio app,
Apple Podcasts, or wherever you listen to podcasts.
TechStuff News
Hosts And Creators
Oz Woloshyn
Karah Preiss
Popular Podcasts
Stuff You Should Know
If you've ever wanted to know about champagne, satanism, the Stonewall Uprising, chaos theory, LSD, El Nino, true crime and Rosa Parks, then look no further. Josh and Chuck have you covered.
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
Las Culturistas with Matt Rogers and Bowen Yang
Ding dong! Join your culture consultants, Matt Rogers and Bowen Yang, on an unforgettable journey into the beating heart of CULTURE. Alongside sizzling special guests, they GET INTO the hottest pop-culture moments of the day and the formative cultural experiences that turned them into Culturistas. Produced by the Big Money Players Network and iHeartRadio.