Edinburgh Research Explorer
Review of Preparing for the Future of Artificial Intelligence"
Citation for published version:
Bundy, A 2017, 'Review of Preparing for the Future of Artificial Intelligence"', AI and Society, vol. 32, no. 2,
pp. 285-287. https://doi.org/10.1007/s00146-016-0685-0
Digital Object Identifier (DOI):
10.1007/s00146-016-0685-0
Link:
Link to publication record in Edinburgh Research Explorer
Document Version:
Peer reviewed version
Published In:
AI and Society
General rights
Copyright for the publications made accessible via the Edinburgh Research Explorer is retained by the author(s)
and / or other copyright owners and it is a condition of accessing these publications that users recognise and
abide by the legal requirements associated with these rights.
Take down policy
The University of Edinburgh has made every reasonable effort to ensure that Edinburgh Research Explorer
content complies with UK legislation. If you believe that the public display of this file breaches copyright please
contact openaccess@ed.ac.uk providing details, and we will remove access to the work immediately and
investigate your claim.
Download date: 10. Aug. 2022
Review of “Preparing for the Future of Artificial
Intelligence”
Alan Bundy
November 16, 2016
In October 2016, the USA’s National Science and Technology Council pub-
lished a report on Artificial Intelligence (AI) [4] that summarised evidence from
a wide variety of sources on how they expect AI to develop, what impact it
would have and what actions it recommended the USA’s Government to take.
It built on several previous USA Government reports, e.g., [5, 3] and consulted
widely among AI experts in the USA, e.g., [1] and five workshops. A com-
panion document has also been published: “The National Artificial Intelligence
Research and Development Strategic Plan”, which lays out a strategic plan for
Federally-funded research and development in AI.
Overall, this is a comprehensive report, which I recommend to anyone want-
ing a balanced review of the state of the AI art, its potential impact and what
ethical, economic and societal issues it presents. It does, of course, duck some
of the more difficult issues — or rather recommend that someone else consider
them in detail.
This concern about AI is, of course, triggered by the phenomenal recent suc-
cesses of mainly statistical machine learning in games (Chess, Go and Jeopardy),
self-driving cars, automated assistants (Apple Siri, Amazon Alexa, Google Now
and Microsoft Cortana). The other main driver is the worry that Artificial Gen-
eral Intelligence (AGI) will exceed human intelligence and supplant us as the
dominant species on the Earth — the, so called, Singularity.
The report wisely ignores concerns about the Singularity, claiming that, if
it occurs at all, it’s a long way in the future, and that the immediate actions
should be the same whether or not it occurs. It points out that the AI successes
have been in, what is calls, narrow AI, i.e., often super-human performance in a
very narrow task, e.g., playing Go. I call such systems idiot savants. In contrast,
progress in AGI has been disappointing, with no sign that this will improve in
the foreseeable future.
So the report focuses on the prospects for narrow AI. It claims that it has
already brought “major benefits to the public in fields as diverse as health care,
transportation, the environment, criminal justice, and economic inclusion”. The
report then explores its potential in autonomous vehicles, governance, education,
cyber-security and weapons. This exploration uncovers a number of important
1
issues and concludes with 23 recommendations. These issues and recommenda-
tions include the following:
Advantages: AI can increase productivity, lower costs, make products and
services more widely available, and provide more accuracy and precision.
Exploitation: Private and public institutions need to consider how they can
take advantage of this potential of AI and hire staff to enable them to do
so.
Research: There needs to be increased investment in AI research, especially
in basic research, where it is not in the immediate interests of industry to
invest, so Government has to be involved.
Education: The report anticipates that “AI-enhanced education” can assist
human teachers to make high-quality education more widely available.
The NSTC Committee on Science, Technology and Mathematics Educa-
tion (CoSTEM) includes this potential within its scope.
Ethics: The need not only for ethics to be an integral part of AI education, but
also that it be “augmented with technical tools and methods for putting
good intentions into practice”. Security, privacy and safety, for instance,
need to integrated into design and assured.
Accountability: The report recognises that it is hard to extract explanations
from statistical machine learning programs. Such explanations are es-
sential for many applications. For instance, medics have to take profes-
sional responsibility for their decisions. They cannot uncritically accept a
machine-generated diagnosis or therapy without some explanation of the
reasoning process which generated it. How this crucial drawback should
be addressed is not discussed.
Privacy: The report recommends that the Government assembles rich collec-
tions of data to inform policy making, but mentions only as an aside that
this collection must be “consistent with consumer privacy”. It has nothing
to say about the massive collection of personal data by the private sector,
e.g., Facebook, Google. Privacy has, however, been previously addressed
in [2].
Regulation: Incremental change of regulations is preferred if and when the
inadequacies of existing regulations are revealed. Regulatory agencies need
to call on AI experts to help them anticipate the changes that will be
required.
Collaboration: The USA should collaborate with other countries over research
and regulation. Government should collaborate with industry.
Employment: The report asserts that AI has the potential to eliminate or
drive down the wages of low-skilled jobs, so measures are needed to main-
tain equality and spread the economic benefits broadly. How this is to
2
be done is hived off to a follow-on report. What is not highlighted is the
mismatch between the pace of AI-led disruption and the inertia of soci-
etal readjustment. Also, that the number of jobs lost is likely to be much
more than the number of new jobs created, at least initially. So, even if
the effect is not long-term, we may need to go through a disruptive period
of mass unemployment. Additionally, some people have asserted that it
will not just be low-skilled jobs that will be affected, but also professional,
knowledge-based ones [6].
Autonomous Weapons: The Government should develop a policy consistent
with humanitarian law. Again, this very difficult issue is not discussed in
detail, but hived off to a follow-on report.
Public Debate: The report recommends the Government initiate a public de-
bate about these issues. One crucial ingredient of such a debate is public
understand of the difference between narrow and general AI. Prior to AI,
the space of intelligent systems consisted solely of general intelligence, al-
beit of widely different capacities across the animal kingdom. Given this
background, it is natural to assume that an AI that is world-class in one
area, say playing Go, would at least be able to cope in other areas, say
driving a car. This is not the case for narrow AI systems, but a sensible
debate, e.g., about the dangers of the ‘Singularity’ cannot be conducted
without an understanding of this distinction. The report has nothing to
say about this issue.
My criticisms above are intended to be constructive. The report is a very
useful contribution to the current debate about the impact of AI. It injects
a note of sanity into what is sometimes an over-excited and under-informed
controversy. It would be too much to expect it also to answer all the open
questions. I have tried to point to some of the unanswered open questions that
must be addressed as the debate continues.
References
[1] B. Horvitz, E. Selman. AAAI presidential panel on long-term
AI futures: 2008-2009 study. Technical report, The Associa-
tion for the Advancement of Artificial Intelligence, February 2009.
http://www.aaai.org/Organization/presidential-panel.php.
[2] The Presidents Council of Advisors on Science and Technology. Report to
the president: Big data and privacy: A technological perspective. Technical
report, Executive Office of the President, May 2014.
[3] United States. Executive Office of the President. Big data: A report on
algorithmic systems, opportunity, and civil rights. Technical report, The
White House, May 2016.
3
[4] United States. Executive Office of the President and M. Holden, J.P. Smith.
Preparing for the future of artificial intelligence. Technical report, National
Science and Technology Council, Washington D.C. 20502, October 2016.
[5] United States. Executive Office of the President and John Podesta. Big
data: Seizing opportunities, preserving values. Technical report, The White
House, 2014.
[6] Daniel Susskind Richard Susskind. The Future of the Professions: How
Technology Will Transform the Work of Human Experts. OUP Oxford, 2015.
4