From the U.S. Government Accountability Office, www.gao.gov

Transcript for: Technology Assessment: Artificial Intelligence

Description: New Artificial Intelligence technologies may fill you with
hope or dread. But AI has arrived, and GAO takes a look at how it may
affect some industry sectors.  

Related GAO Work: GAO-18-142SP: Technology Assessment: Artificial
Intelligence: Emerging Opportunities, Challenges, and Implications.

Released: March 2018


[ Background Music ]

[ Tim Persons: ] It's important not to be afraid of the issues. It is
more appropriate to be concerned but not overly concerned.

[ Matt Oldham: ] Welcome to GAO's Watchdog Report, your source for news
and information from the U.S. Government Accountability Office. I'm Matt
Oldham. Artificial intelligence is a concept that was once found only in
science fiction from The Jetsons maid Rosie to Hal in 2001: A Space
Odyssey, it seems many in popular culture have believed AI is an
inevitable evolution in humanity's journey. Granted, there's no telling
if we'll ever face annihilation threat from Skynet or benefit from the
services of a Lieutenant Commander Data but one thing is looking almost
certain, AI technologies are coming and they're bringing changes. I sat
down with Tim Persons, GAO's Chief Scientist, to talk about a
comptroller general forum that explored AI, its potential benefits,
challenges, and implications. So Tim, did the forum focus on any
particular industry sectors?

[ Tim Persons: ] Yes. So we focused on four different things that
conveyed a different context of risk in each case. So let me say what I
mean. The four were cyber security risks. So that's the idea of
information risks. Automated vehicles. So that's safety of life risk,
right? If you get in a car, you want to be safe. Criminal justice risk.
So it's social and constitutional risk management. And then financial
services. So financial risks. And what we see is that, although AI in
essence even across all these sectors is essentially the same
technology, it expresses itself differently and will need to be
regulated differently or understood differently in its various
expressions.

[ Matt Oldham: ] I'd like to take a look at each of those. Let's start
off with cyber security. It seems like this is a natural fit for AI
technology.

[ Tim Persons: ] That cyber is a key pain point across, again, all
sectors. We all remember we're at risk for losing our credit card
information and we're concerned about our social security numbers and
rightly so. And the problem is so massive and complex, it's increasingly
apparent that we need artificial intelligence like machines to help us
just police what's going on and try to identify fraud and things like
that to help mitigate it. So a lot of opportunities there. And yet, of
course, some policy challenges there particularly in terms of preserving
civil liberties and not having a surveillance state, for example.

[ Matt Oldham: ] We recently did a podcast on emerging technologies in
the financial services sector and one of the things that was the focus
was something like robo investing. Is this the sort of AI application in
financial services?

[ Tim Persons: ] Yes and financial services has, at least, a couple
things we looked at in the forum. One was on this just how financial
services are already being changed and there's a lot of powerful work
going on in terms of automated investing and things like that which
takes out some of the human emotional risk in the marketplace. So that's
an interesting dynamic. On the regulatory side, of course, our SEC is
our important regulator of the financial and it's powerful to them
because their job is to help make sure the playing field is even and to
be able to look across the markets and try and sniff out anything that
looks amiss or fraudulent or out of place.

[ Matt Oldham: ] Moving on to automated vehicles, this is certainly a
topic that's been in the news quite a bit for some time. What did the
forum look at in the automated vehicle sector?

[ Tim Persons: ] Right. So one of the key lessons learned out of our
Department of Transportation, what they've been doing is they've
established something called a regulatory sandbox where they're allowed
to say, 'well, let's see how the car operates so that we can inform and
make better regulations.' We know with vehicles, you're going to -- we
have regulation now on cars. You're going to need more but the fact that
you have a driverless system changes the dynamic significantly. So it
was important to construct that regulatory sandbox to be able to have
the science and innovation go on and yet be informed, symbiotically is
what we would say, you know together it's working with the regulatory
system to say well what are virtuous rules that should be written that
don't, at the same time, kill our innovation and competitiveness in the
U.S. So a lot of excitement about the technology but there's just risks
to manage obviously since you're taking the human out from behind the
wheel.

[ Matt Oldham: ] Rounding out the four sectors is criminal justice. Now
this is one that seems a little more abstract to me. How does artificial
intelligence play in the criminal justice sector?

[ Tim Persons: ] Every police municipality, state or local municipality
has the issue of finite resources. There's only so many things, crime of
course is to a greater or less extent challenges in all these areas. So
how you distribute the resources and yet, how you, at the same time,
preserve constitutional liberties and avoid and mitigate biases and so
on, that's the key opportunity here for them. It's a challenge and
opportunity for criminal justice. So the idea about making a more blind
system of justice, more effective system is what's in play and yet
again, not taking the humans fully out of the loop. That's never
intended on these things on the higher end decision making where wisdom
and knowledge is required beyond what a machine can or will have.

[ Background Music ]

[ Matt Oldham: ] I'll be the first to admit maybe it's too easy to
relate AI to characters in books and movies without acknowledging AI is
gaining admission into our everyday lives. And maybe that's because of
how difficult it is to predict how much change AI will bring. Tim, what
came out of this forum on artificial intelligence technologies?

[ Tim Persons: ] The economy is and will continue to be affected in
different ways. Thinking of this in a foresight context is useful to try
and just anticipate or mitigate. Doesn't mean we're going to be able to
control things. In fact, I think it'll be just setting the opportunity
for things to proliferate but lots of sectors are being and will be
affected. I think there's going to be a lot of regulatory reform that's
needed on the government side and yet, I think there's going to be
continued innovation on -- in terms of the private sector doing their
work and that they should be allowed to continue to go on but to think
through, I think perhaps more holistically in terms of ethical, legal,
and social issues as well. It's the idea of you can do something with
technology but should you? Or how can it be done is going to be an
important area? So on the government side, I think just understanding
its affects, how we do jobs and training, how we come up with what's
called explainable AI so that we understand what the algorithm is likely
if not guaranteed to do. And being able to just have that ongoing
symbiosis between private side and then the public sector side.

[ Matt Oldham: ] If there's one concept that our listeners can take away
from this forum, what would it be?

[ Tim Persons: ] I think the main thing is be not afraid. You think
about great movies and stuff, I enjoy the sci-fi movies myself. Those
are fun but that's not the reality. And so it's important not to be
afraid of the issues. It is, I think, more appropriate to be concerned
but not overly concerned. It's a how might we be able to utilize this
for greater good and yet mitigate risks that we do want to avoid those
unintended consequences. We know this is going to have effects on jobs
and yet there's opportunities. It'll be different kinds of jobs that are
coming out and I think we need to think about in that way where we're
cautiously optimistic about what this can and should do but again,
keeping an eye carefully on each of these areas. These are just several
of many, many others and what it means moving forward. So thinking
prospectively I think will be important for this technology for sure.

[ Background Music ]

[ Matt Oldham: ] Thanks for listening to the Watchdog Report. To hear
more podcasts, subscribe to us on Apple Podcasts. 

[ Background Music ]

[ Matt Oldham: ] For more from the congressional watchdog, the U.S.
Government Accountability Office, visit us at gao.gov.