Zach Warren, Legaltech News
Panelists
at the Stanford CodeX FutureLaw Conference separated fact from fiction,
explaining how AI truly functions in today’s legal world
For everything you need to know about artificial
intelligence in the legal profession, it’s worthwhile to turn to the greatest
of legal minds, as channeled by ROSS Intelligence’s Andrew Arruda: Mugatu from
Zoolander. “AI, so hot right now!”
Yes, everybody is talking in grandiose terms
about the future of AI in law. However, at the Stanford CodeX FutureLaw
Conference at Stanford Law School on May 20, a panel of AI experts cut through
the hype surrounding artificial intelligence to find the true future and applications
for AI in law.
The “Hot or Not: Watson and Beyond” panel,
moderated by professor Dan Katz of IIT Chicago-Kent College of Law, featured
panelists who work with AI on a daily basis, including Khalid Al-Kofahi,
Thomson Reuters; Andrew Arruda, ROSS Intelligence; Charles Horowitz, The MITRE
Corporation; Himabindu Lakkaraju, Stanford University; and Noah Waisberg, Kira
Systems.
Between IBM, Microsoft Azure, AWS, Google, and
others, the sheer number of machine learning offerings is increasing, the panelists
noted. However, before diving into machine learning’s role in the legal
profession, Waisberg explained, it’s important to discern between what AI does
and does not do.
“We need to think about whether the tasks that
they are optimized for are tasks that lawyers do,” Waisberg explained. “There
is a common misconception that you throw machine learning at a problem, and
it’s solved.”
For example, for current legal research
questions, or contract disputes, machine learning can be a boon. For other
types of questions, such as those of nuanced law or positioning a firm in a
marketplace, it may not.
“What we need to do before we breathlessly
exclaim that any machine learning is right for a problem is think about what
the problem is that we’re trying to solve,” he added.
Still, there are enough applications of machine
learning in law that it is right to be excited, Al-Kofahi explained. After all,
for the tasks that AI is conditioned for, “you want to cut down a tree not with
an axe but with a chainsaw.”
The success of the AI implementation, though,
may not fall with the tool itself. As Al-Kofahi put it, “It’s not about the
tool; it’s about the mechanic who is using the tool.”
Lakkaraju, who has been working on the
intersection of AI and law in Stanford’s laboratories, agreed. She said that
for current AI technology, effectiveness often boils down to one point: How do
you ask the question?
“Formulation lies in the hands of people trying
to use these machine learning tools, to some extent,” she explained. She added
that although these tools were originally built to be all-encompassing, their
current application in the field has been “diversified to an extent where it is
for every specific task.”
Because of the necessity of having people
guiding the tools, it’s important to make sure that these are the right people,
Waisberg added. “The best lawyers to work with machine learning tools are
actually good lawyers. … If you have a system that is learning from people,
it’s really garbage in, garbage out.”
Horowitz noted that this does not necessarily
mean people with a background in legal technology or artificial intelligence;
without those barriers, AI researchers and users will have a larger base from
which to pool people to teach an AI system. To that end, Waisberg said, “It’s
more attitudinal than expertise – you need to embrace the idea of working with
a machine.”
And many people have embraced it. Shortly before
FutureLaw, ROSS
Intelligence announced at a conference at Vandy Law School that BakerHostetler would be
adapting its AI systems for use in the firm. At FutureLaw, Arruda announced
that two
more firms were jumping on board: Latham & Watkins and von Breisen &
Roeper.
“Sometimes, when you talk about machine learning or AI, we already hit the singularity 3 years ago for some reason,” Arruda said. “It’s Day 1, but on Day 1, we’ve already found a use for it.”
Ultimately, the panelists indeed agreed that AI
is “hot” right now, but perhaps not at the earth-shattering level that many
prognosticators outside the industry believe. Horowitz said that maturity is
coming, in the same way that it took software development many years to mature.
“It depends entirely on expertise and data, and
neither of those are necessarily true of software development. … But if you
give us 20 years, we’ll arrive at the same point of maturity.”
No comments:
Post a Comment