This is the sixth, and final episode in a series dedicated to all things A.I. In this episode, Tae Royle, Head of Digital Products APAC from Ashurst Advance Digital is joined by Tara Waters, Partner and Head of Ashurst Advance Digital.
Tae Royle:
Hello and welcome to Ashurst Legal Outlook. This is the sixth and
final episode in a series dedicated to all things Artificial
Intelligence. My name is Tae Royle head of digital products from
Ashurst did that digital and today I'm joined by Tara Waters
partner and head of Ashurst Advanced Digital based out of our
London office. In this series, we've discussed the ethical
considerations of Artificial Intelligence, copyright, patents and
AI trade secrets. Naturally we come to the question of what's
next? In Lewis Carroll's second novel, Alice enters Wonderland
by climbing through a mirror. There, she finds that everything is
not quite as it seems. And some things are very odd indeed. In
today's episode, we'll explore some things about AI that
are very odd indeed. As we ask ourselves, what is the future of AI
in legal, and how can lawyers best prepare themselves? Hi Tara, and
thanks for joining me. Before, we explore the future of AI within
the legal profession. I was wondering whether maybe you could set
the scene and touch on some of the current challenges that are
facing the legal community in the adoption of Artificial
Intelligence more broadly?
Tara Waters:
Firstly, there remains an overall lack of understanding of what
artificial intelligence is, what it means, what it does and how it
works. And I think with what comes with that is a little bit of a
fear of the unknown and also a lot of times a mismatch of the
expectations, the promise of AI versus the reality and what the
current solutions are capable of doing and doing well. Part of the
problem of that, I think, is ultimately the lack of access to big
data in the industry. A lot of our data is confidential data.
Tara Waters:
It's private data, it's client data. And certainly for law
firms, client data is sacrosanct. We don't share that. We
don't allow access to that. That stays locked as tightly as
possible behind firewalls and data silos. And as a result of that,
though, what happens is we don't have a concept of big data in
this industry in the same way we've seen in other industries
and ultimately for AI, really to deliver on some of the promise of
what we've seen elsewhere. You need access to data and a level
of data that we just haven't achieved yet in our industry.
Therefore, I think what we need is industry level discussion about
how do we enable some of that access?
Tara Waters:
How do we enable new solutions to come in and solve the problems
that we have? And I think that we're really at the beginning
stages of that journey at the industry level. I think finally
there's a differentiation in my mind anyway, of legal tech
versus technology for legal. And I think sometimes we're trying
to solve for the very narrow use cases, which in some back to that
first point around the mismatch of the expectations versus reality
and the ability to actually deliver the true promise that AI
has.
Tae Royle:
That's really interesting Tara, because I've heard data
scientists tell me that their single largest challenge is access to
quality data. And I can see that in the legal industry that would
be even more difficult than other industries. Are you aware of any
industry initiatives or approaches in order to resolve this
challenge?
Tara Waters:
I am. Yes. I do think what we're seeing now is an increased
amount of collaboration happening within the industry, not just as
amongst legal services providers, but also new providers and
universities, research organizations, data research organizations.
I think a great example of that would be in the UK. Our UK innovate
grant program ran a tender last year for participants to come
together and form consortium to help solve some data access issues
and our firm, Ashurst has actually participated in one of those
collaborations with an up-and-coming data tech company. And that is
involving participants from a multitude of regulated industries who
all share this data access problem along with universities and
research companies and their own new technology. We are absolutely
seeing the industry recognize the need for there to be
collaborations and discussions around this problem.
Tae Royle:
You also mentioned legal tech as distinct for tech for legal,
specifically in the AI space. Is it your thought/recommendation
that we should be looking at some of the Machine Learning tools
that are more general application and seeing how potentially they
could be applied to legal use cases? Or did you have something else
in mind?
Tara Waters:
No, I think it's exactly that. And I would love to see more
involvement of what we refer to as the big tech companies, the
Googles, the Amazons, et cetera of the world, coming into our
industry and applying their technology to our problems. I don't
think there's any reason why we should only be looking at use
of AI and Machine Learning within our industry without knowledge
and input from the real experts who are doing phenomenal things
with Machine Learning and other industries. If you look at that,
the healthcare industry and the ability now of Machine Learning to
be applied in diagnosing cancers. And I know that that was a
journey and it took a long time for it to reach the level of
accuracy that doctors felt comfortable with what was being done.
And we also, I think as a result saw that doctors weren't being
replaced, but certainly the ability of the technology enhanced
their ability to actually work with their patients, treat their
patients, diagnose issues with their patients.
Tae Royle:
I agree with that completely and incidentally, I was reading an
article today about how, part of the reason why we were able to
develop vaccines so quickly for Corona virus was techniques
utilizing AI that have only been developed in 2020 regarding the
folding of proteins and as such, this led to the acceleration of
the ability to provide the vaccines that are now being rolled out
around the world. There's huge leaps and bounds being made in
AI/Machine Learning on a month by month basis at the moment. It is
a really exciting field and there are lots of opportunities to
bring some of those learnings into legal. I like now to turn to
specifically the label field and what we're seeing going on in
legal, in Artificial Intelligence, what areas do you think
we're doing well? What are some of the successes?
Tara Waters:
I think that the areas where we've seen real success has
definitely been around looking at and using documents, reviewing
documents, extracting information, sorting through vast volumes of
documents. Those are areas that touch upon, for example,
e-discovery, which has become a discipline in and of itself over
the past 10 years. And it is growing quite rapidly as an area that
we're expanding our capability. And when we look around and we
see the industry really jumping behind e-discovery, AI based
solutions. And also then the classic example, of course, in terms
of document review, where you're able to just process vast
volumes of information and data and locate it and pinpoint exactly
the data that you're looking for. Pulling that out, extracting
it, and then pushing that on into a different process where
you're able to take that next step, whether that's analysis
or generation of further information. And that's something that
I think within the industry we can really do at scale, we know
works well. Our clients expect us to be using AI for those types of
activities and tasks. That's definitely widely accepted.
Tae Royle:
E-discovery is a really interesting space for me because
e-discovery's getting involved in so many different formats
other than just the traditional review of email and documents. For
example, you've got information that's coming out on social
media and an area that I found really quite fascinating was I had
something come across my desk about six months ago, we had a client
who came and spoke to us. They had about 30,000 hours worth of
footage from the inside of a warehouse. And they wanted to know
whether we could use AI to scan through the 30,000 hours as part of
a particular investigation.
Tae Royle:
Now, in that particular instance, we weren't really able to
help them, not withstanding that we had some imaging tools that we
could work with, but effectively the problem set that was put to us
was, we want to know if there's any suspicious activity going
on in the warehouse and we weren't able to get any much greater
definition around what those activities might look like. One of the
challenges with AI is if you're going into a large data set and
you're trying to find something, you need to know what
you're looking for. I think that leads on nicely though. There
are some really brilliant use cases in artificial intelligence
where we can deliver great legal outcomes for our clients. Where do
we go from here? What do you think are the next steps and what sort
of achievements might we be able to reach in the next couple of
years?
Tara Waters:
One of the areas that I'm particularly excited about is
application of federated learning within the industry. We've
actually started a collaboration with a data tech company, and
we've been talking to a range of collaborators. We have one
potential use case and proof of concept where we've brought
together a group of law firms. And we have a second similar proof
of concept, but actually it's a couple of law firms, a couple
of corporate clients. And this is the really exciting apart. And
some of the legal regulators in the UK who are expressing an
interest in this technology. Federated learning is ultimately about
how do you achieve that big data view and allow Machine Learning to
start to really provide extra value? What it entails is it allows
multiple parties to come together and ultimately pool their data in
a way that you can then get a much wider and more aggregated view
of the insights.
Tara Waters:
But what's the technology that we're testing is proposing
to do is to allow each data owner or data holder to maintain the
confidentiality and separateness of their data. There's no
movement of data, there's no actual sharing of data. We
don't have to worry about those confidentiality and security
concerns that we inherently have as a law firm. But what we can do
is allow access to a carefully constructed and designed algorithm
to come to our data sites, to extract the insights that we've
agreed, the algorithms should be extracting, not actually
extracting specific data from the contracts themselves or documents
themselves and pulling them out and then allowing a aggregated and
anonymized view of those insights, rather than us being able to see
what's in our own documents, we can actually pull the insights
from a range of documents from across the industry, including the
clients, and get a much more powerful view and much more exciting
level of learnings that can be applied in a wide variety of ranges,
including in the public interest as well as obviously for private
interest.
Tara Waters:
I think that probably takes us to the next step, which is how do we
leverage more open source and industry solutions in the industry as
we discussed a little bit before in those collaborations with big
tech and getting comfortable with open source, which as a general
proposition and anyone that does transactional law will know that,
that's a key due diligence topic when anyone is using open
source technologies and the ability to do that and the risks
inherent in not using code that was completely written by you. I
think there's probably still some steps we need to take in
order to get comfortable around that, but certainly the ability to
leverage what's being built by other people, leverage the vast
amount of expertise on this subject outside of our industry, I
think is going to be really critical to us making proper
advancements.
Tae Royle:
It seems like this would require a fair amount of upfront
investment to get up and running. Do you have any insights to share
around how do we persuade key stakeholders to invest in Artificial
Intelligence and these sorts of initiatives? What sort of messaging
do you think resonates with them most, if you feel that there's
an important AI initiative that should be carried out by the firm
or alternatively, if your in-house council, how might you approach
that?
Tara Waters:
I think that's an important question. And I think every
organization right now is on a trajectory of transformation of some
kind that's being digitally enabled. And so each organization
will have its own approach to that transformation program.
What's really is making sure the leadership and executive team
understands the importance of research and development and
investment in order to actually achieve the aims of transformation.
And what that means in most cases is that you're not going to
get access to a huge pot of money and of investment and risk
capital, and the same way you might, where a corporate spot up a
venture arm, for example, who seated with a set amount of money and
is able to go run around and play and invest in, and find those new
solutions to bring back into the business. I think what's
really important though, is that the leadership starts to
understand a bit better. How do you invest in and how do you accept
risk and see risk as that as an opportunity and not simply as a
risk and money out the door?
Tae Royle:
The title of this episode is Through the Looking Glass: The Future
of AI. And I've seen a few odd things that AI has produced in
my time. Is there anything that comes to mind for you where AI
produces results that are distinctly odd?
Tara Waters:
Well, I think we have to always keep in mind that a machine is not
a human it's unlikely to be trained at least at this point in
time to be exactly like a human. And so it can't just look at a
set of documents or look at a reel of footage and pick out that
thing that really only a human's going to be able to do. I
mean, I'm curious to know what are some of your findings?
Tae Royle:
I think there's a couple of areas that are really exciting at
the moment. There's an organization known as OpenAI, which was
co-founded by Elon Musk and they've created a natural language
processing tool. They have effectively scooped up the entire
internet and cut it down into bite sized chunks. And then the
Machine Learning tool set reassembles those chunks in patterns that
it considers would make sense to a human. And it's fascinating
to read the output of GPT-3, there's thousands of examples on
the internet. If you simply go Google GPT-3, but in effect you have
these very well-written flowing sentences that are gathered up into
somewhat sensible paragraphs. And then you get three quarters of
the way through the paragraph. And a disconnect with reality
becomes starkly evident, and you start descending into this form of
madness. And it's very much like reading the writings of Lewis
Carroll, Ian Matt, the machine that's producing these language
outputs has no conception of the human world and how it all fits
together, but it can write these perfectly lucid literate learned
sentences.
Tae Royle:
And furthermore, you can give it a particular style. So it can
write a sonnet in the style of Shakespeare or John Dunn. It just so
happens that as you read it, you quick really realize that the
sonnet does not necessarily make any sense whatsoever.
Interestingly, a lot of lawyers might reject these sorts of tools
out of hand, because they're not the sort of thing that can
give you legal answers that you can rely on. In some ways the
answers you might get back are often a little bit ridiculous, but
what they are, and this is where I'm going with a new tool that
I've just learned about in the last couple of days, brought out
by IBM. They can write very cogent, very beautifully written script
and text. And it's possible to use that text in order to form
your own arguments. Lawyers normally like to think of themselves as
being very black letter, following prescriptive rules.
Tae Royle:
But the reality is that they are also very persuasive writers. And
the thing that I'm also super excited about is a tool set
coming out of IBM research, which is called Project Debater. And
this is an Artificial Intelligence tool that debates with humans on
specific topics. And I'm super interested to see whether or not
this could be applied in a legal context. By way of wrap-up, what
do we need to do as lawyers, as legal professionals, as humans to
ensure that we get the most out of Artificial Intelligence and
Machine Learning, both to benefit our clients and possibly humanity
as a whole, what should we be doing?
Tara Waters:
I would really love to see the legal industry get together with the
technology industry and to tackle those bigger problems to
understand how do we get access to bigger levels of data in order
to train algorithms to do the more complex and structured thinking
that lawyers do and how can we work across the industry and within
the industry. And let's not look at each other as competitors
in this context, let's look at us all benefiting from the
improvements in AI and Machine Learning used in this industry,
because we're hearing it from our clients on a daily basis.
They are interested, they want us to keep pushing the envelope and
do more. They want us to be safe and secure and give them all the
assurances that they need, of course, but ultimately this is
benefit for the greater good.
Tae Royle:
Yeah, that's right. And I really enjoyed your discussion around
how we can bring our stakeholders with us and engage them because I
think that's really one of the most important parts of any
project. Well, it's crucial in order for the project to get off
the ground is to ensure that we have the right buy-in. And the way
to get that buy-in is to ensure that we have clear deliverables and
hard metrics against which we can be measured so that we can
demonstrate the value of what we're achieving, because
that's how you achieve sustaining innovation is by being able
to demonstrate the value that you've already delivered, and
then to be able to enumerate the value that you're going to
deliver in the future. I love the idea of starting small, have a
very focused project, have a small number of people who are working
on that project will nurture and grow that project and be able to
demonstrate that value. I think that's a great message.
Tara Waters:
Watch this space. As the series has shown, there's so much
happening in AI. So much promise here. We're all really excited
about it. Don't lose faith, don't lose hope. I think that
the future's very bright for this industry.
Tae Royle:
Thank you for listening. To hear more Ashurst podcasts, including
our dedicated channel on all things ESG, please visit
ashurst.com/podcast. To ensure you don't miss future episodes
subscribe now on Apple podcasts, Spotify, or your favorite podcast
platform. If you heard something today or in the previous episodes
that interested you and you want to learn more, we encourage you to
reach out to us on LinkedIn and keep the conversation going. Thanks
again for listening and goodbye for now.
The content of this article is intended to provide a general guide to the subject matter. Specialist advice should be sought about your specific circumstances.
