Artificial intelligence and policy: quo vadis?
| Pages | 238-263 |
| Date | 13 May 2019 |
| DOI | https://doi.org/10.1108/DPRG-09-2018-0054 |
| Published date | 13 May 2019 |
| Author | Anastassia Lauterbach |
| Subject Matter | Information & knowledge management,Information management & governance,Information policy |
Artificial intelligence and policy: quo vadis?
Anastassia Lauterbach
Abstract
Purpose –This paperaims to inform policymakers about keyartificial intelligence (AI) technologies,risks
and trends in national AI strategies.It suggests a framework of social governance to ensureemergence
of safe and beneficialAI.
Design/methodology/approach –The paper is based on approximately 100 interviews with
researchers, executivesof traditional companies and startups and policymakersin seven countries. The
interviewswere carried out in January-August 2017.
Findings –Policymakers still needto develop an informed, scientifically grounded and forward-looking
view on what societiesand businesses might expect from AI. There is lack of transparencyon what key AI
risks areand what might be regulatory approaches to handlethem. There is no collaborative frameworkin
place involving all important actors to decide on AI technology design principles and governance.
Today’s technology decisions will have long-term consequences on lives of billions of people and
competitivenessof millions of businesses.
Research limitations/implications –The research did not include a lot ofinsights from the emerging
markets.
Practical implications –Policymakers will understandthe scope of most important AI concepts, risks
and nationalstrategies.
Social implications –AI is progressing at a very fast rate, changing industries, businesses and
approaches how companieslearn, generate business insights, designproducts and communicate with
their employees and customers.It has a big societal impact, as –if not designed with care –it can scale
human bias, increase cybersecurity risk and lead to negative shifts in employment. Like no other
invention, it can tightencontrol by the few over the many, spread false information and propaganda and
therewithshape the perception of people, communitiesand enterprises.
Originality/value –This paper is a compendium on themost important concepts of AI, bringing clarity
into discussions around AI risks and the ways to mitigate them. The breadth of topics is valuable to
policymakers,students, practitioners,general executives and board directorsalike.
Keywords Artificial intelligence, Machine learning
Paper type Viewpoint
From the demonstration of an artificial intuition with DeepMind AlphGo Zero to
reviewing legal documents and screening through heaps of medical images to
detect diabetic retinopathy more efficiently than humans can, artificial intelligence
(AI) is progressing at a very fast rate. If implemented consequently by companies, AI
touches every aspect of organizational culture while empowering employees to
independently make decisions based on real-time data. It amplifies weaknesses as well. If
many conflicting goals compete within corporate silos, AI is used just for the efficiency
game or for cost cutting. We cannot neglect that a one-sided focus on AI in businesses can
have a big societal impact, e.g. loss of jobs through automation or failure to add new
building blocks to education systems.
Like every other technology, AI can be used to benefit individuals and businesses. At the
same time, like no other invention, it can tighten control by the few over the many, spread
false information and propaganda and therewith shape the perception of people,
communities and enterprises. Fake news is just one negative example of how technology
can be misused to inflate democracy. Adversarial AI is a fast-growing risk, capable to
Anastassia Lauterbach is
based at the Lauterbach
Consulting and Venturing
GmbH, Friedrichallee,
Bonn, Germany.
Received 6 September 2018
Revised 16 March 2019
Accepted 16 March 2019
PAGE 238 jDIGITAL POLICY, REGULATION AND GOVERNANCE jVOL. 21 NO. 3 2019,pp. 238-263, ©EmeraldPublishing Limited, ISSN 2398-5038 DOI 10.1108/DPRG-09-2018-0054
change real-time data and algorithms, be it in an autonomous driving scenario to confuse
sensors’ objects recognition or be it changing a company’s financials during data transfer.
Facial recognition in a city, smart speakers as witnesses in a courtroom or a credit denial
decided by algorithms –there are many cases where a policymaker might want to get
involved (Table I).
Regulation has a critical underlying role to play in the adoption and uptake of new AI
technologies, while learning to cope with the never-ending tensions between technology
giants and public, bad actors trying to exploit opportunities to pursue their corrupt goals
and (parts of) society, traditional businesses and digital disruptors of any size and even
different national governments capable to either limit adoption or boost progress of new
technologies. Regulators can create bureaucratic rules that stifle innovation or make it too
costly. On the other hand, they can sponsor collaborative research and communication
between different groups in society and technology companies, seek transparency over
conflicting targets andfind common denominators to support trust.
After the Second World War, countries have seen how transportation, communication and
social technologies advances expand welfare. Adding AI into the equation magnifies all the
sharp angles on what has not been sufficiently considered and properly architected. AI
scales everything with the brute force of mathematics. Understanding this scaling
mechanisms and design of these is crucial for policymakers. Technology illiteracy of the
parties involved in regulation contain not just inefficiencies, it dilutes the understanding of
the consequences in technology adoption on the scale of the whole society, slows down the
creation of retaliation mechanisms to address reputational harm of businesses and
damages sustainability of wholeindustries and companies.
This paper is written from the point of view of an AI practitioner, corporate boards’ non-
executive director and venture partner. It is meant for people structuring questions around
AI technologies for decision-makers in regulations and policy. It follows on the research
published in “The Artificial Intelligence Imperative. A Practical Roadmap for Business”
(Lauterbach and Bonime-Blanc,2018).
The paper provides practical definitions of AI, discusses AI risks and ethics and documents
applications in different industrial verticals and policy work, including in several selected
national AI strategies.
What is AI? So far, there is no common definition of the term. In practice, it coversa group of
technologies and scientific fields that focus on automation, acceleration and extreme
Table I Examples of AI advances and pitfalls in 2018
Advances Pitfalls
Uber has cracked two classic ‘80s video games, Montezuma’s
Revenge and Pitfall, by giving an AI algorithm a new type of
memory. This is a serious achievement in Reinforcement
Learning, focused on introducing it despite of few reliable
reward signals in the games
Revelations about Facebook’s and Cambridge Analytica’s role in
Brexit and the US election 2016 made it clear, that technology
companies did not have the necessary policies and practices in
place to shield democratic elections from fake news and attacks by
foreign adversaries
Alibaba’s Voice assistant was capable to deal with interruption,
non-linear conversation and implicit intent –conditions usually
difficult in Natural Language Understanding
Google Project Maven to enhance drone strikes with imaging
technology met with serious opposition from the company’s
employees, ultimately leading to the cancelation of the contract
between Google and Pentagon and forcing a discussion about AI in
military and defense
Nvidia achieved big quality leap in images of people who did not
exist, further improving Generative Adversarial Networks
(GANs). GANs use two dueling neural networks to train a
computer to learn the nature of a data set well enough to
generate convincing copies
VOL. 21 NO. 3 2019 jDIGITAL POLICY, REGULATION AND GOVERNANCE jPAGE 239
To continue reading
Request your trial