Name: AI through the noise: Beyond the hype
Uploaded: 2023-12-13T10:22:58.466Z
Duration: 639.2

AI through the noise.mp4

AI through the noise.vtt

ENGLISH

{{00:00:07}}
My name is Paul Hindle, Editor at
FinTech Futures, and we are here today

{{00:00:10}}
at the FinTech Futures AI Insights
event at our office in London.

{{00:00:13}}
With the AI hype cycle in full swing,
we thought it would be great to kick

{{00:00:16}}
the event off with a panel looking
really beyond the hype of AI and

{{00:00:19}}
looking at the real world use case of
the technology in financial services

{{00:00:23}}
and the impact that's being made there.

{{00:00:25}}
I was joined on stage by three
AI experts, so you can hear

{{00:00:28}}
some of their insights now.

{{00:00:35}}
Yeah, so our firm specializes in
data extraction from PDF documents.

{{00:00:40}}
So many of our customers are
using our software to read

{{00:00:43}}
things like bank statements.

{{00:00:44}}
So they may have millions of bank
statement pages that they need to

{{00:00:48}}
analyze in order to make a loan, a
commercial loan to an SME, for example.

{{00:00:52}}
Now that process is just
impossible to do manually.

{{00:00:55}}
You can't have people reading two million
pages of bank statements every year.

{{00:00:59}}
This is too much work, it takes
too long, and it's too expensive.

{{00:01:02}}
So we're using generative AI models to
extract that information automatically.

{{00:01:07}}
So I think we're seeing three
key things around generative AI.

{{00:01:11}}
They are, one, talent.

{{00:01:13}}
So how do you manage the HR processes
of onboarding talent, and then managing

{{00:01:17}}
their experience within the organization.

{{00:01:19}}
in as friction free way as possible.

{{00:01:21}}
That's the first one.

{{00:01:22}}
The second one is customer service.

{{00:01:24}}
How do you automate that customer
service whilst making it cheaper

{{00:01:27}}
to serve, but also a real pleasure
to experience for that end user.

{{00:01:32}}
And then the third theme is what
we call application modernization.

{{00:01:36}}
And this is really where we are.

{{00:01:38}}
Converting old technical data or old
legacy technology into more modern

{{00:01:42}}
technology using generative AI, and
things like automating the testing

{{00:01:46}}
process as well as converting old
code to new and more modern code

{{00:01:51}}
which is far easier to manage.

{{00:01:53}}
So the three key things.

{{00:01:54}}
That's what we're seeing across
the financial services sector.

{{00:02:04}}
Mainly it's speed.

{{00:02:05}}
It's also accuracy.

{{00:02:07}}
Because humans are very
bad at doing this task.

{{00:02:10}}
Humans are actually really bad at manually
extracting information from documents.

{{00:02:14}}
They tend to make mistakes, they get
bored, obviously the machines don't

{{00:02:17}}
get bored, so their accuracy is really
good, their accuracy is much higher than

{{00:02:21}}
humans, and ultimately it's much cheaper.

{{00:02:24}}
You don't need as many people doing
this task, and it's a boring task,

{{00:02:27}}
and people just don't want to do it.

{{00:02:29}}
So the benefits of AI are boundless,
I think we are still very early in

{{00:02:34}}
this nascent market, especially of
generative AI, so I don't want to stay

{{00:02:39}}
we know exactly where this is going
to go, but I think in terms of we're

{{00:02:43}}
heading to where the net interest margin
businesses major banks is, a challenge.

{{00:02:51}}
So keeping costs down and getting
unnecessary costs out of businesses

{{00:02:56}}
is an absolute business imperative.

{{00:02:59}}
But also improving the customer
experience at the same time.

{{00:03:03}}
I am genuine of the belief that this
technology is that sort of nirvana,

{{00:03:07}}
which enables you to both bring costs
down, will make experience go up.

{{00:03:13}}
So I think that's the key sort of
benefit that I would say I'd make.

{{00:03:24}}
So for a long time, I think people have
thought that AI is only really suitable

{{00:03:29}}
for automating repetitive tasks, or
tasks which don't require any creativity.

{{00:03:32}}
I now think that with the advent
of generative AI, that's not true.

{{00:03:37}}
That these models can now do very
highly complex, do very complex work.

{{00:03:44}}
They have very high cognitive
capabilities, and they can actually

{{00:03:47}}
be better at humans at doing some of
the creative tasks that we set up.

{{00:03:50}}
I think that that what happens when OpenAI
came out with Chapter 23 back in November

{{00:03:57}}
last year was it appears to publish public
consciousness of what this stuff was.

{{00:04:03}}
Now this stuff is a mix of
phenomenally clever maths and

{{00:04:07}}
engineering and sheer computing power.

{{00:04:11}}
So those two things coming
together in this era are what

{{00:04:14}}
really is fueling this boom.

{{00:04:16}}
So I think when banks and other financial
institutions are looking for problems

{{00:04:21}}
to solve with generative AI, I would
encourage them to think about it, as

{{00:04:26}}
you think about any other technology,
what business problems are you trying to

{{00:04:30}}
solve, and then generative AI may help
solve that, rather than walking around

{{00:04:34}}
with a hammer and everything to it.

{{00:04:42}}
So the new thing now with generative AI
is that you need to, you probably need to

{{00:04:46}}
send your data outside of your firewall.

{{00:04:49}}
So some, a company like OpenAI or one
of the other large tech companies,

{{00:04:53}}
that didn't used to be the case.

{{00:04:55}}
You used to be able to use state
of the art models in house.

{{00:04:58}}
You didn't need to send your data outside.

{{00:04:59}}
Now you probably can.

{{00:05:00}}
If you make the imports of GPC4 or another
model from Atropic or from from Google.

{{00:05:07}}
So this means you've
got data privacy issues.

{{00:05:08}}
So now there's a big tension between
using in house open source models and

{{00:05:13}}
sending your potentially sensitive data
outside your firewall to another company.

{{00:05:17}}
And that, these are questions
which have been yet to be.

{{00:05:22}}
In regulated and financial services
environments it's, it is a challenge to

{{00:05:27}}
get this stuff into production and to
get this stuff in front of customers.

{{00:05:31}}
I think the way that I would
encourage any customer to think

{{00:05:34}}
about this is cool in front, right?

{{00:05:37}}
So start small and then get
bigger as you understand it more.

{{00:05:41}}
Bring your risk teams with you, don't
treat them as a gate at the end.

{{00:05:45}}
But really build a core muscle
about how to deliver this stuff into

{{00:05:48}}
production with small use cases and
then scale up would be my encouragement.

{{00:05:54}}
And also choose the right tooling,
choose the right teams, choose

{{00:05:57}}
the right professionals around
you to help you really scale this.

{{00:06:01}}
I'm fortunate to lead a team that have
done this now for the best part of two

{{00:06:05}}
years at Generative AI, which is really
an ancient history of Generative AI.

{{00:06:09}}
And we've learned a huge
amount in that period.

{{00:06:12}}
As well as IBM has wider tooling
around lots of things like governance,

{{00:06:15}}
which helps you, for example, govern
these Generative AI models, which

{{00:06:20}}
is not an insignificant challenge
for regulated environments.

{{00:06:31}}
I think something that people don't
want to talk about, I'm not sure if

{{00:06:33}}
it's a negative, depends on which
perspective you stand on, I think

{{00:06:36}}
there are going to be job losses,
particularly in the white collar.

{{00:06:40}}
There are, the evidence is that
generative AI can automate a

{{00:06:43}}
lot of white collar tasks, and
ultimately lots of white collar jobs.

{{00:06:48}}
We don't know what's going to happen.

{{00:06:49}}
It may be that new jobs arise, lots
of people hope that, but the truth

{{00:06:53}}
is we don't know what's going to
happen, and it may be that there

{{00:06:56}}
are Large numbers of job losses.

{{00:06:57}}
I think we need to be careful as a society
to make sure that we don't mitigate that.

{{00:07:02}}
I think the negatives are just, it can
be quite frustrating when you can see

{{00:07:07}}
the opportunity that it can afford you.

{{00:07:09}}
But actually the pace of
change is remarkably slow when

{{00:07:12}}
you are heavily regulated.

{{00:07:14}}
So really trying to get those quick
wins on the board and release value.

{{00:07:19}}
Whilst you're going through the
process of governance and change

{{00:07:22}}
management in a large regulated bank.

{{00:07:26}}
I think that's really where
the frustration sets in.

{{00:07:29}}
But keep your eye on the prize,
bear with it, and the prize

{{00:07:33}}
will definitely be worth it.

{{00:07:42}}
I can only say I only live
in one small AI world.

{{00:07:45}}
We deal with document processing,
intelligent document processing.

{{00:07:49}}
Previously, we've always done raw
data extraction for documents.

{{00:07:53}}
Someone gives us a balance sheet or an
income statement, and we extract some

{{00:07:56}}
numbers from those financial statements.

{{00:07:59}}
Now, with the newest models,
we can do a lot more than that.

{{00:08:03}}
We can actually do financial analysis.

{{00:08:05}}
We can extract that information
and do something with it.

{{00:08:08}}
We can make projections,
sensitivity analysis, hypothesis

{{00:08:11}}
testing, and our ambition is
much bigger than it used to be.

{{00:08:16}}
More.

{{00:08:17}}
We've moved on from just raw
extraction of data to true financial

{{00:08:23}}
analysis and that's really exciting.

{{00:08:26}}
I think the negatives are just, it can
be quite frustrating when you can see the

{{00:08:30}}
opportunity that it can afford you, but
actually the pace of change is remarkably

{{00:08:34}}
slow when you are heavily regulated.

{{00:08:37}}
So really trying to get those quick
wins on the board and release value

{{00:08:41}}
whilst you're going through the
process of governance and change

{{00:08:45}}
management in a, large Regulated Bank.

{{00:08:48}}
I think that's really where
the frustration sets in.

{{00:08:51}}
But keep your eye on the prize.

{{00:08:53}}
Bear with it.

{{00:08:54}}
And the prize will definitely be worth it.

{{00:09:05}}
I don't want to predict the future.

{{00:09:07}}
I think it would be disingenuous
for me to pretend that I know

{{00:09:09}}
what the future is going to bring.

{{00:09:11}}
We're still in very early days.

{{00:09:12}}
Hasn't even been any years
since Jack GPC was released.

{{00:09:17}}
And we don't really know
what's going to, or how this

{{00:09:20}}
technology is going to be used.

{{00:09:21}}
This was still We're still
playing with the technology.

{{00:09:24}}
We're still trying to understand its
capabilities and its limitations.

{{00:09:28}}
So it's going to take us some time.

{{00:09:29}}
And I don't think anyone can say
where this technology is going to

{{00:09:33}}
take us over the next few years.

{{00:09:34}}
And that's that's why it's so exciting.

{{00:09:36}}
So I think there's going to be this idea
of assistants or agents, which, yeah,

{{00:09:40}}
if you think about your Consumer using
one of these large language models is

{{00:09:47}}
like a goldfish conversation where you
have a queue, you call question and then

{{00:09:50}}
an answer, question and then answer.

{{00:09:51}}
I think where we'll start seeing is
these idea of agents or assistants

{{00:09:56}}
which chain together those thoughts.

{{00:09:58}}
and then are able to execute a workflow.

{{00:10:00}}
So you can have a procurement agent, you
can have a legal agent, you can have a

{{00:10:04}}
fraud detection agent, and you ask it
in your natural language about to do a

{{00:10:09}}
task, to do a job, to complete a workflow.

{{00:10:12}}
It can then understand that within the
context of the environment that you're

{{00:10:16}}
in, and then go and fulfill that task.

{{00:10:18}}
And I think that's quite an
exciting thing that we're starting

{{00:10:21}}
to move towards in the industry.

{{00:10:24}}
Thank you so much for watching.

{{00:10:25}}
Once again, my name is Paul
Hindle, editor at FinTech Futures.

{{00:10:27}}
Make sure to check us
out at FinTechFutures.

{{00:10:29}}
com where you can keep up to date
with all the latest FinTech news.

My name is Paul Hindle, Editor at
FinTech Futures, and we are here today

at the FinTech Futures AI Insights
event at our office in London.

With the AI hype cycle in full swing,
we thought it would be great to kick

the event off with a panel looking
really beyond the hype of AI and

looking at the real world use case of
the technology in financial services

I was joined on stage by three
AI experts, so you can hear

Yeah, so our firm specializes in
data extraction from PDF documents.

So many of our customers are
using our software to read

So they may have millions of bank
statement pages that they need to

analyze in order to make a loan, a
commercial loan to an SME, for example.

Now that process is just
impossible to do manually.

You can't have people reading two million
pages of bank statements every year.

This is too much work, it takes
too long, and it's too expensive.

So we're using generative AI models to
extract that information automatically.

So I think we're seeing three
key things around generative AI.

So how do you manage the HR processes
of onboarding talent, and then managing

their experience within the organization.

How do you automate that customer
service whilst making it cheaper

to serve, but also a real pleasure
to experience for that end user.

And then the third theme is what
we call application modernization.

Converting old technical data or old
legacy technology into more modern

technology using generative AI, and
things like automating the testing

process as well as converting old
code to new and more modern code

That's what we're seeing across
the financial services sector.

Because humans are very
bad at doing this task.

Humans are actually really bad at manually
extracting information from documents.

They tend to make mistakes, they get
bored, obviously the machines don't

get bored, so their accuracy is really
good, their accuracy is much higher than

humans, and ultimately it's much cheaper.

You don't need as many people doing
this task, and it's a boring task,

So the benefits of AI are boundless,
I think we are still very early in

this nascent market, especially of
generative AI, so I don't want to stay

we know exactly where this is going
to go, but I think in terms of we're

heading to where the net interest margin
businesses major banks is, a challenge.

So keeping costs down and getting
unnecessary costs out of businesses

But also improving the customer
experience at the same time.

I am genuine of the belief that this
technology is that sort of nirvana,

which enables you to both bring costs
down, will make experience go up.

So I think that's the key sort of
benefit that I would say I'd make.

So for a long time, I think people have
thought that AI is only really suitable

for automating repetitive tasks, or
tasks which don't require any creativity.

I now think that with the advent
of generative AI, that's not true.

That these models can now do very
highly complex, do very complex work.

They have very high cognitive
capabilities, and they can actually

be better at humans at doing some of
the creative tasks that we set up.

I think that that what happens when OpenAI
came out with Chapter 23 back in November

last year was it appears to publish public
consciousness of what this stuff was.

Now this stuff is a mix of
phenomenally clever maths and

So those two things coming
together in this era are what

So I think when banks and other financial
institutions are looking for problems

to solve with generative AI, I would
encourage them to think about it, as

you think about any other technology,
what business problems are you trying to

solve, and then generative AI may help
solve that, rather than walking around

So the new thing now with generative AI
is that you need to, you probably need to

So some, a company like OpenAI or one
of the other large tech companies,

You used to be able to use state
of the art models in house.

You didn't need to send your data outside.

If you make the imports of GPC4 or another
model from Atropic or from from Google.

So this means you've
got data privacy issues.

So now there's a big tension between
using in house open source models and

sending your potentially sensitive data
outside your firewall to another company.

And that, these are questions
which have been yet to be.

In regulated and financial services
environments it's, it is a challenge to

get this stuff into production and to
get this stuff in front of customers.

I think the way that I would
encourage any customer to think

So start small and then get
bigger as you understand it more.

Bring your risk teams with you, don't
treat them as a gate at the end.

But really build a core muscle
about how to deliver this stuff into

production with small use cases and
then scale up would be my encouragement.

And also choose the right tooling,
choose the right teams, choose

the right professionals around
you to help you really scale this.

I'm fortunate to lead a team that have
done this now for the best part of two

years at Generative AI, which is really
an ancient history of Generative AI.

And we've learned a huge
amount in that period.

As well as IBM has wider tooling
around lots of things like governance,

which helps you, for example, govern
these Generative AI models, which

is not an insignificant challenge
for regulated environments.

I think something that people don't
want to talk about, I'm not sure if

it's a negative, depends on which
perspective you stand on, I think

FinTech Futures

Event Session

Ai through the noise is an interview with Dr Martin Goodson, CEO of Evolution AI AND Michael Conway, Executive Partner, Data, Ai and technology Transformation Service Line Leader at IBM. We look beyond the hype of Ai that has become a massive topic in Fintech and try to find out what the real deal is in the future with this technology. Exclusively held at the Fintech Futures insight into Ai event.