THE RISKS AND BENEFITS OF ARTIFICIAL INTELLIGENCE AND ROBOTICS - An event in collaboration with and hosted by

Page created by Monica Patterson
 
CONTINUE READING
THE RISKS AND BENEFITS OF ARTIFICIAL INTELLIGENCE AND ROBOTICS - An event in collaboration with and hosted by
THE RISKS AND BENEFITS OF
ARTIFICIAL INTELLIGENCE AND ROBOTICS

An event in collaboration with
and hosted by
THE RISKS AND BENEFITS OF ARTIFICIAL INTELLIGENCE AND ROBOTICS - An event in collaboration with and hosted by
Acknowledgements

This report has been prepared by UNICRI.

UNICRI would like to express its appreciation to the high-level experts and participants of the
workshop held in Cambridge, United Kingdom, in February 2017. Special thanks also go to all
those that supported the workshop at both UNICRI and the Centre for Risk Studies at the
University of Cambridge Judge Business School, in particular: Ms. Marina Mazzini (UNICRI),
Mr. Irakli Beridze (UNICRI) and Dr. Michelle Tuveson (Cambridge Centre for Risk Studies) for
the initiative and overall management of the workshop; Mr. Fabrizio De Rosa (UNICRI) for its
organization and his multimedia services; Ms. Sona Krajciova (Cambridge Centre for Risk
Studies) for logistics; Mr. Odhran McCarthy (UNICRI) for organizational support and the
preparation of this report; and Mr. Beniamino Garrone (UNICRI) for its design.
THE RISKS AND BENEFITS OF ARTIFICIAL INTELLIGENCE AND ROBOTICS - An event in collaboration with and hosted by
Table of Contents

Introduction ............................................................................................................................3
Opening of the workshop ......................................................................................................3
Artificial intelligence and robotics 101: what is it and where are we now ..........................4
Ethics and artificial intelligence ............................................................................................7
The cyber security overlap ....................................................................................................9
From fear to accountability – the state of artificial intelligence journalism..................... 10
Emerging technologies: quantum computing .................................................................... 12
Economic and social implications of robotics and artificial intelligence ......................... 13
Long-term issues of artificial intelligence and the future of humanity ............................. 14
Robotics and artificial intelligence at the United Nations ................................................. 16
Concluding remarks ............................................................................................................. 16

Annex 1: Agenda
Annex 2: Speakers biographies
Annex 3: Event photos

                                                               Page 2 of 18
THE RISKS AND BENEFITS OF ARTIFICIAL INTELLIGENCE AND ROBOTICS - An event in collaboration with and hosted by
1. Introduction
The potential risks and benefits associated with advancements being made in the fields of
artificial intelligence (AI) and robotics were analysed and discussed during a two-day
workshop organized by the United Nations Interregional Crime and Justice Research Institute
(UNICRI) in collaboration with, and hosted by, the Cambridge Centre for Risk Studies. The
event took place at the University of Cambridge Judge Business School (United Kingdom)
from 6 to 7 February 2017.

As part of UNICRI's Artificial Intelligence and Robotics Programme and its Public Information
Programme on New Threats, journalists, representatives from the academia, international
organizations and the private sector from 20 countries met with leading AI and Robotics
experts to deepen their understanding of advancements in AI and robotics, with a special
focus on their potential global security implications.

This report summarizes the issues presented and discussed during the workshop.

       2. Opening of the workshop
The workshop was opened by Mr. Irakli Beridze, Senior Strategy and Policy Advisor at the
United Nations Interregional Crime and Justice Research Institute (UNICRI), and Dr. Michelle
Tuveson, Founder & Executive Director, Cambridge Centre for Risk Studies at the University
of Cambridge Judge Business School, who officially welcomed participants to Cambridge. Mr.
Beridze mentioned that this was the second edition of the workshop. The previous edition
took place at the Netherlands Institute of International Relations Clingendael in The Hague, in
March 2016.

Following the initial welcoming remarks, Dr. Ing. Konstantinos Karachalios, Managing Director
of The Institute of Electrical and Electronics Engineers (IEEE) Standards Association and
member of the Management Council of IEEE, delivered a keynote address which appealed to
the audiences’ logic and emotion when considering the very messy, yet fundamentally
important theme of AI and robotics.

Dr. Karachalios noted that, in September 2015, world leaders at the United Nations adopted
the Sustainable Development Goals, also known as “Transforming our world: the 2030
Agenda for Sustainable Development.” Therein, the United Nations made reference to
technology as one of the major pillars of implementation of the 17 goals of the SDGs. While it
is indeed the case that technology can help us achieve the development goals and have a

                                              Page 3 of 18
THE RISKS AND BENEFITS OF ARTIFICIAL INTELLIGENCE AND ROBOTICS - An event in collaboration with and hosted by
massive impact on our world and our lives - for example by doubling life expectancy – he
noted technology also paradoxically threatens our existence.

Approximately 45 years ago, Buckminster Fuller, the American author, inventor, and architect,
observed that technology had reached a point at which we had the ability to utilise it to provide
the necessary protection and nurturing that our society requires to fulfil our needs and ensure
growth. We were crossing a singularity, he felt, making things such as war obsolete. In this
technical era, he questioned which political system can or should be the structure and
backbone of our society. Or, for that matter, if one was even required at all.

In contrast, the German Philosopher Martin Heidegger held a more pessimistic view of
technology. While many feel that technology is something under our control, this was not the
case for Heidegger. For him, once set on its course, the development of and advancements in
technology were something beyond our control.

Dr. Karachalios felt that the reality is probably somewhere in between and it is up to us to
decide what we feel or believe. For him, this decision was the heart of the two-day workshop.
To help us come to our own conclusions, Dr. Karachalios suggested we look for impact. For
instance, has technology made the world more sustainable? Has technology made the world
more democratic, fair or safe? In his opinion we should seek to measure if and how
technology has delivered under each of these topics before concluding optimistically or
pessimistically.

At the same time, looking more broadly at the agenda of the workshop, Dr. Karachalios
emphasised that in this technical era it is important that we properly educate and inform
ourselves about the potential dangers, risks and dilemmas in our path. With technology, he
noted, we tend to lack a clear understanding of the direction in which we are moving, and
instead follow the impetus of the various drivers or “war machines” that push forward
developments and advancements. Such drivers, Dr. Karachalios clarified, include: militarism,
geopolitics, the religion of “do-ability” within the techno-scientific community and even our own
fear of death. It is important for us to better understand these drivers or “war machines” to be
able to choose our own direction. He added the caveat however, that going against these
drivers will be costly. There will be no reward along the way for the individuals engaged
against the "war machines", but it is his hope that the end result will be worth it for the
collective.

He commended the workshop as a forum to hold dialogue on these drivers or “war machines”
behind technology and encouraged all those present to continue building bridges amongst
throughout the concerned communities to foster engagement.

                                             Page 4 of 18
THE RISKS AND BENEFITS OF ARTIFICIAL INTELLIGENCE AND ROBOTICS - An event in collaboration with and hosted by
3. Artificial intelligence and robotics 101: what is it and
           where are we now
Prof. Noel Sharkey, University of Sheffield, UK, Co-Founder of the Foundation for Responsible
Robotics (FRR) and Chairman of the International Committee for Robot Arms Control
(ICRAC), opened the first session of the workshop. Before commencing however, Prof.
Sharkey stressed that it is important to remember that, while we do not know what the future
holds for us, we do know the present, and, in this reality, there are many problems facing us.
In this regard, he felt it prudent to avoid unnecessarily talking about the future for fear of
distracting ourselves from the real risks and problems that we face today.

To get started, Prof Sharkey provided participants with a brief introduction to AI and robotics.
He noted the origins of the term “AI” – originally coined in 1955 by the American computer
scientist John McCarthy – and observed the great deal of confusion that surrounds the term,
particularly with respect to the word “intelligence”. The resulting ambiguities were lamented by
many, including even John McCarthy himself, who is noted as having wished he never called it
AI at all. Moving along, Prof Sharkey noted that massive claims and false predications of
capabilities is something often associated with AI. More than 60 years since our first foray into
the field of AI and the promises of machines cleaning our homes have not yet materialised.

Notwithstanding this, there have been many significant milestones, notably including Google
DeepMind's 'AlphaGo' competing against, and ultimately prevailing over, world renowned
professional Go player, Lee Sedol, in 2016. Prof. Sharkey noted however, that the systems
making these milestones have been very singular in their function, for example, in playing Go,
Chess or Jeopardy. We have yet to meet a universally intelligent system. In this regard, he
speculated that these milestones resulted not from what the founding fathers of artificial
intelligence were trying to achieve with programming, but rather from a combination of the
massive computational power of big data and sophisticated pattern recognition techniques.
These games are uniquely suited to modern artificial intelligence.

Pushing aside the curtain, Prof. Sharkey explained that what is called “machine learning” is
the computer science behind the field of artificial intelligence. Machine learning or “statistical
parameter estimation”, he added, is in essence a network or several matrices of numbers,
wherein an input is converted into numbers (binary) and multiplied across the matrices. This
results in an output classification of the input. In the event of a mistake, an error correction
formula modifies the weights and the input is inserted again and the whole process is iterated
until the proper classification results.

                                             Page 5 of 18
THE RISKS AND BENEFITS OF ARTIFICIAL INTELLIGENCE AND ROBOTICS - An event in collaboration with and hosted by
Turning to robotics, Prof. Sharkey suggested that robots are conceptually more
straightforward. Simplifying his description, he explained that robots can be understood as
computers with sensors and motors connected to wheels or legs. The computer runs a
process in response to a sensor detecting a stimulus, such as heat, and the computer sends a
signal to the motors to move the robot accordingly. There are two main types of robots –
industrial and service. As with AI, there have been a number of significant developments in
this field. Most notably, there has been an explosion in the quantity of robots, with the World
Federation for Robotics (WFR) predicting that there will be 38 million robots worldwide by
2018 for a wide range of activities, from farming to cleaning to pumping gas to assisting the
elderly.

Given this prediction and the increasing prevalence of robotics in our daily lives, Prof. Sharkey
next turned his attention to shining the spotlight on some important societal issues that are
often overlooked.

For instance, he noted immediate concerns with the use of robotics in both child care and the
care of the elderly. While there are of course many advantages, such as assisting the elderly
and keeping them independent, humans require human contact. Depriving the young or the
old of this can have serious psychological consequences and may, in the case of children,
lead to disorders such as detachment. Noting the significant investments of countries such as
Japan, Korea and the United Kingdom into so-called companion robots for the young and old,
he stressed that we must not be blinded by potential economic benefits. We must remain
sensitive to the very human needs of people.

We have also seen the emergence of robotics and AI on our roads, with the increasing
popularity of autonomous vehicles. While he acknowledged the potential for this technology to
save lives, he also stressed caution. In essence, robots are machines and, as machines, there
is the ever-present possibility of a malfunction. Already we have seen fatalities on the road
with autonomous vehicles, one in the US, one in China and one in the Netherlands. While we
push ahead with this technology, we need to take care that proper systems are in place. At
present, this does not appear to be the case. He pointed out that self-driving cars could save
lives if we get them right and with caution we can engender public trust.

One of the largest points of concern for Prof. Sharkey is when robotics and AI come into
conflict with our right to life. Lethal autonomous weapons systems (LAWS), or so-called “killer
robots”, are not the humanoid robots portrayed in Hollywood movies. Rather they are drones
and submarines, operating either individually or collectively in swarms. While we have yet to
see a fully autonomous weapon that makes targeting and attack decisions without human
intervention, there are now a number of weapon systems with varying degrees of autonomy.

                                            Page 6 of 18
THE RISKS AND BENEFITS OF ARTIFICIAL INTELLIGENCE AND ROBOTICS - An event in collaboration with and hosted by
Uralvagonzavod’s T-14 Armata tank and the Northrop Grumman’s unmanned combat air
vehicle, the X-47B, are examples of very advanced weapons systems edging toward full
autonomy. The problem, Prof. Sharkey explained, is that while these systems are the pinnacle
of advancements in AI, the technology nonetheless encounters difficulties in distinguishing
civilians and combatants, which presents a challenge for complying with International
Humanitarian Law, also known as ‘the laws of war’. At the United Nations level, there is an
ongoing debate on LAWS within the context of the Convention on Prohibitions or Restrictions
on the Use of Certain Conventional Weapons Which May Be Deemed to Be Excessively
Injurious or to Have Indiscriminate Effects (CCW). During the 2014 CCW Meeting of High
Contracting Parties, Member States agreed on a new mandate on LAWS and, since then have
discussed the matter on a yearly basis. While progress is slow, Prof. Sharkey noted that in
2016 the United Nations established a Group of Governmental Experts (GGE) to specifically
discuss, over a three-week period, the legal, ethical and societal concerns of LAWS in the
context of the objectives and purposes of the Convention.

Prof. Sharkey also noted that developments in this field were beginning to spill over into law
enforcement. In this regard, he noted the Skunk Riot Control Copter that is armed with pepper
spray for crowd control and Chaotic Moon’s "stun copter". He noted the positive side of the
technology, which keeps police out of harm’s way, but stressed the importance of not letting
these developments get out of hand by dehumanising violence and changing the very nature
of policing.

Echoing Dr. Karachalios’ earlier comments, Prof. Sharkey concluded noting the importance of
education as we move forward. Specifically, he acknowledged that, while professional ethics is
taught to engineers in universities, this is largely ethics with respect to the client and not the
ethics of social responsibility. We must be more aware of the various rights and societal issues
at stake here.

        4. Ethics and artificial intelligence
Building on the foundation laid by Prof. Sharkey, Ms. Kay Firth-Butterfield, Barrister-at-Law,
Distinguished Scholar at the Robert S. Strauss Center for International Security and Law,
University of Texas, Austin and Co-Founder of the Consortium for Law and Ethics of Artificial
Intelligence and Robotics, led the second session on Ethics and AI. At the outset of her talk,
Ms. Firth-Butterfield, noted that advancements in this field are often compared to the industrial
revolution of the 18th and 19th centuries, but according to her, it is simply not the same thing.
The speed, rate of change, and impact of this new industrial revolution are unparalleled. In this
regard, she expressed that it is absolutely paramount that we already start discussing the

                                             Page 7 of 18
THE RISKS AND BENEFITS OF ARTIFICIAL INTELLIGENCE AND ROBOTICS - An event in collaboration with and hosted by
many ethical concerns raised by advancements in these technologies. We must do so, she
noted, because, in the end, AI is just software, bugs and all. Society cannot afford to “beta-
test” AI, addressing problems with software updates as they arise.

While AI is currently trending upwards, coming more and more into the public eye in recent
years, she felt it necessary not to overplay the capabilities of new technology, for fear of
sparking panic and harkening the next AI winter. If we allow fear to take control of us, we will
only end up with unnecessary, and perhaps ineffectual, regulation. She aptly grounded the
discussion in reality, making reference to influential deep learning specialist Yann LeCun, who
once noted that, at present, we cannot even build a machine as intelligent as a mouse.

Returning to the issue of ethics and AI, she explained that discussions really took off with the
sale of Deepmind to Google in January 2014, and the announcement that as part of the deal
Google was setting up an ethics advisory board to ensure that its AI technology is not abused.
The discussion bloomed thereafter following remarks from Professors Stephen Hawking and
Max Tegmark on the positive and negative effects of AI. Around this time, other commercial AI
entities, created their own ethics boards, for instance Lucid AI set up its Ethics Advisory Panel
in October 2014 which she led and included Professors Max Tegmark, Murray Shanahan and
Derek Jinks as members. This attention contributed a lot of ongoing research into AI ethics
and commendable work such the Future of Life Institute’s 23 high-level principles on AI ethics
— the Asilomar Principles.

In connection with this, Ms. Firth-Butterfield noted that AI and robotics has also recently
received a lot of governmental attention in terms of reports and regulations. Europe, for
instance, is very advanced and even produced a report in 2016 on Robotics and AI, proposing
legal personhood and addressing important issues such as intellectual property. On the other
side of the Atlantic, she said, the White House also produced two noteworthy reports in 2016
outlining a strategy for promoting AI research and development. Within the US, the State of
Nevada has even adopted legislation and regulations which allow for the use of autonomous
trucks on its roads.

In 2012, the Institute of Electrical and Electronics Engineers (IEEE) launched its Global
Initiative for Ethical Considerations in Artificial Intelligence and Autonomous System, an open,
collaborative and consensus based approach to Ethics and AI. Ms. Firth-Butterfield explained
that the IEEE’s goal is to educate, train and empower technologists to prioritize ethical
considerations in the design and development of autonomous and intelligent systems.
Through this initiative, she added, the IEEE issued a report, Ethically Aligned Design, in
December 2016 and is actively seeking contributions from the community to review and
improve the report. She also indicated that work on two new standards projects on

                                            Page 8 of 18
THE RISKS AND BENEFITS OF ARTIFICIAL INTELLIGENCE AND ROBOTICS - An event in collaboration with and hosted by
Transparency of Autonomous Systems and Data Privacy Process was started and, as part of
the IEEE’s peer driven and consensus-based approach, she invited participants to join the
working groups.

Concluding, Ms. Firth-Butterfield reflected on earlier comments by Prof. Sharkey, stating that
we have to question AI when it begins to spill over into our lives. While innovations such as the
use of AI in predictive policing may save lives and prevent harm, they can have a significant
impact on our civil liberties and our freedoms. The question remains: where do we draw the
line with innovation? She further speculated whether, if innovation continues to curb our civil
liberties, will we reach a point whereby it becomes necessary to challenge the very purpose of
innovation. Or, as Salesforce’s CEO, Marc Benioff said at Davos 2017, to even slow down the
rate of innovation through regulation. This may be an option in the future, but it is not
something being considered by Governments at present. Currently, national policies are
focused on maximising technology for economical purposes.

       5. The cyber-security overlap
In the next session, the cyber security overlap was introduced by the team from the
Cambridge Centre for Risk Studies, which consisted of Professor Daniel Ralph, Academic
Director and Professor of Operations Research at University of Cambridge Judge Business
School; Mr. Simon Ruffle, Director of Technology Research & Innovation; and Ms. Jennifer
Copic, Research Assistant.

Kicking off the discussion, Professor Ralph took the floor to explain the work of the Cambridge
Centre for Risk Studies to address systematic risks in business, the economy and society.
Prof. Ralph observed that we have moved beyond the information technology revolution into
the data revolution, and in this new era there are novel risks for us to deal with, like cyber risk.
He defined cyber risk as any risk of financial loss, disruption or damage to the reputation of an
organisation from some sort of failure of its information technology (or operational technology)
systems.

Our environments have also changed, he explained, noting that critical infrastructure has
become an increasingly complicated mix of public and private sector actors, so much so that
we are not fully clear who owns the risk and who is ultimately responsible for it. Governments
and regulators? Private sector critical infrastructure companies? Or society more broadly, as
represented by corporations and private consumers? The result is what Prof. Ralph refers to
as ‘the triangle of pain’. The effects of the triangle are further amplified by growing
interdependency between sectors. The energy sector, for instance, plays a central role for all

                                              Page 9 of 18
critical infrastructure, feeding various other sectors, and, in this regard, cyber-attacks, such as
that which occurred in Ukraine in December 2015 leaving more than 200,000 people without
power, can be extremely harmful.

Building on this, Mr. Ruffle explained that part of the Centre’s ambition is find ways to measure
the economic impact of a catastrophic event and advance the scientific understanding of how
systems can be made more resilient. This is what they refer to as “catastronomics” or the
economics of catastrophes. Specifically, he explained, the Centre employs the use of stress
test scenarios and insurance loss models for cyber-attacks (for example DDOS, financial theft,
malware ransomware). To further the discussion, Mr. Ruffle described the ‘Erebos’ Trojan
attack, a hypothetical cyber-attack from a malware affected laptop targeting the North Eastern
electrical grid in the US from a distance. The attack involves overloading the air-conditioning
system in a facility, causing 50 generators to catch fire, prompting the shutdown of the
electrical grid, leaving 93 million people in 15 states without power — an area responsible for
30% of the country’s economy. Assessing the GDP risk for the next 5 years, they estimate
losses of between 243 and 1,024 billion USD, with the insurance industry losing 21.4 and 71.1
billion USD.

In the final part of the session, Ms. Copic discussed the insurance models for cyber risks,
including affirmative standalone cyber coverage, affirmative cyber endorsements, and silent
cyber exposure polices with gaps in explicit cyber exclusions or without cyber exclusions. She
noted that there is a wide variation in the coverage language in insurance policies, with no two
policies really being the same. To try to measure what these policies really mean in economic
terms, the Centre developed a series of cyber-attack scenarios for insurance accumulation
management. The scenarios can be used to stress test both individual syndicates and the
market as a whole.

       6. From fear to accountability - the state of artificial
          intelligence journalism
Mr. John C. Havens, Executive Director of The IEEE Global Initiative for Ethical
Considerations in Artificial Intelligence and Autonomous Systems and contributing writer for
Mashable and the Guardian, took the floor next on the topic of the media’s representation of
AI. He noted that AI is represented in the media in a highly polarised manner. It is either
dystopian or utopian in nature. Very rarely does the media present balance or even consider
solutions. Fear, he added, is an important tool in journalism, and some editors often rely on
fear to sell their publications. As a result, while the AI community regularly laments the use of

                                             Page 10 of 18
the Terminator in presenting AI to the public, the Terminator will always be there. The
challenge, Mr. Havens suggested, is to get beyond the fear and to really look at what
developments like AI mean for individuals and for our society. Now, more than ever, this is
important as we are at a really important time. Technology is already beginning to change us
he explained, and although the Singularity may still be far away, we have already started to
cross a threshold — to merge with the very technology we are developing.

In light of this, Mr. Havens expressed the importance of ethics in AI, adding that it is not
something we can afford to have only as an afterthought. Thorough examinations and
assessments should be done before a technology is released to make sure it is ethically
aligned with end users, and methodologies applied to provide robust due diligence to identify
and prioritize human values. The problem, he clarified, is this: How will machines know what
we value if we do not know ourselves?

He observed, that as part of his H(app)athon Project, an initiative to create digital tools to drive
global contentment, a survey was conducted to get a sense of how people view their well-
being in the digital age. Psychologists have determined that while values may differ amongst
people, there are twelve values that people in different cultures all prioritize. These values can
help identify what general ethical principles we can honour to build into emerging
technologies. While it is not an easy challenge, it is something the IEEE’s work in creating
their paper, Ethically Aligned Design and new Standards Working Groups, will help.

At the same time, Mr. Havens stressed that we should not lose sight of the importance of data
privacy as this digital boom continues. Data is a valuable commodity. This will increasingly
become apparent with the rise of augmented and virtual reality, where users will be looking
through the lenses of private companies for every part of their waking life.

In his conclusions, Mr. Havens noted that regardless of how the media represents AI,
technology is neither good nor evil. Nevertheless, he warned that it is certainly not inert. He
observed that the World Health Organization estimates that by 2030 depression is expected to
be the largest contributor to disease burden. Noting that unemployment is a major contributor
to depression, he urged that we must recognise the reality that, while autonomous cars will
save lives on our roads, their impact will be profound on our individual and collective psyche.
Not only will the trucker profession become obsolete almost overnight, the entire community
surrounding the profession will also feel the impact, as, for instance, the importance of gas
stations, roadside diners and everyone supporting the industry diminishes. In this regard, Mr.
Havens stressed that when we develop AI, we need to think it in terms of what helps us
increase human well-being versus only prioritizing GDP and exponential growth, which is the
status quo of today.

                                             Page 11 of 18
7. Emerging technologies: quantum computing
Next, Dr. Natalie Mullin, researcher at 1QBit, a Vancouver-based quantum computing software
company, introduced the audience to quantum computing, an emerging technology poised to
have a significant impact on AI and on our society as a whole. She explained that quantum
computing attempts to harness the strange nature of quantum mechanics, and in particular the
phenomena known as superposition and entanglement. Operating on the basis of quantum
bits (or qubits), instead of bits as in the case of classical computers, quantum computers offer
considerable computational advantages over classical computers. Dr. Mullin explained that
there are different types of quantum computers, with the original ideal being a universal
quantum computer. The more limited, yet commercially available, quantum computers are
known as quantum annealing computers.

Dr. Mullin explained that quantum computing is a very relevant emerging technology because
cybersecurity relies on the difficulty of mathematical problems that make up public-key
encryption schemes. For instance, in 2009, researchers solved an encryption problem based
on a 232-digit number (RSA-768). Using multiple computers, this effort took over two years.
On one standard desktop processor, the researchers estimated the task would take
approximately 1,500 years. The current standard of 617-digit numbers (RSA 2048) is 4.3
billion times more difficult than a 232-digit number. Future universal quantum computers could
significantly reduce the time required to solve these complex mathematical problems from
hundreds of years to only a matter of weeks.

Fortunately for cybersecurity’s sake, Dr. Mullin noted that we have not yet reached that point
because to run the algorithm to crack RSA 2048, we would need a computer than operates
with 4096 qubits and millions of qubits for error correction. Given that current universal
quantum computers control between 5 and 20 qubits, this is massively beyond our current
capabilities.

Nevertheless, given that quantum annealing computers are scaling up and doubling in power
every 12 to 24 months, we have seen increasing interest in this field. For example, in 2015,
the National Security Agency in the United States announced that it is getting ready to move
towards quantum resistant algorithms. The National Institute of Standards and Technology
(NIST) similarly called for the creation of such algorithms.

Turning to the issue of AI, Dr. Mullin went on to explain how quantum computing can benefit AI
and the 2013 launch of MIT and Google’s drive toward quantum algorithms for machine

                                             Page 12 of 18
learning. As an illustration, she noted that quantum computing might boost facial recognition
capabilities by reducing the number of data points required for accurate facial recognition.

Returning again to the issue of security, Dr. Mullin noted that, the more qubits we have, the
more useful the computers become, but also the more dangerous they become in terms of
cybersecurity.

Concluding, Dr. Mullin noted that it is a very interesting time for quantum computing. For
instance, IBM has said they can scale up their universal quantum computer to 50-100 qubits
within the next decade, while D-wave has recently released a 2,000-qubit sparsely-connected
quantum annealing computer. Interestingly, Google is at the same time advocating for a more
powerful quantum annealing computer than D-Wave’s computer with fewer, better-connected,
qubits. Only last week, she observed, the University of Sussex unveiled its plan for the
construction of a universal quantum computer, as opposed to a quantum annealing computer.
Even though the blueprints indicate the computer will be the size of a football pitch, it is a
significant breakthrough. In this regard, the future is bright for quantum computing and we
should start to think about how we apply this technology to innovatively address problems in
our society.

       8. Economic and social implications of robotics and
          artificial intelligence
Mr. Olly Buston, Founding Director of Future Advocacy, opened the next session, examining
the term “AI” and noting that there are difficulties in defining it, primarily because “intelligence”
is a challenging term to define in itself. To overcome this, the AI community uses a broad
understanding of AI based on problem solving abilities. In spite of its artificial nature however,
AI is something that touches upon the very the essence of being human — intelligence. After
all, he observed, intelligence is in our name: Homo sapiens. Intelligence is how our species
distinguishes itself from animals. In this regard, AI raises questions about the nature of
humanity.

Mr. Buston noted a significant development boom in AI as of late, which primarily results from
advancements in machine learning. Quoting Professor Stephen Hawking, he observed that
the creation of AI will be “either the best, or the worst thing, ever to happen to humanity”. The
difference between this intelligence revolution and past revolutions, he added, is its
extraordinary scale and the scope of change. He noted that, although AI will, for instance,
turbo charge productivity, maximise effective use of resources, create safer roads and even
lead to the creation of entire new categories of employment for the children of the future, there

                                              Page 13 of 18
are serious underlying economic and social concerns that should not be glossed over.
Exemplifying this, he noted that 35% of the routine intellectual jobs in the United Kingdom are
at risk from automation over the next 20 years. This amounts to approximately 15 million jobs.
Similarly, he referred to the State of Nevada’s recent adoption of legislation and regulations
allowing for the use of autonomous trucks on its roads, observing that “truck driver” is the top
job in the majority of US States. Moving beyond the West, he questioned what automation
means for low-income countries, or those countries whose economic strategy is based upon
the provision of low cost labour. Developments in AI and robotics will seriously challenge
these countries to diversify their economic strategies.

Notwithstanding this, there seems to be a certain degree of complacency regarding
automation and a clear lack of public and political focus. He observed that a recent poll in the
UK indicated that British people generally tended not to be worried about jobs being replaced.
In the political context, he noted that there has only been a total of 32 references to AI in the
House of Commons. Notwithstanding this, he noted some positive political developments,
including the recent the House of Commons Science and Technology Committee Report on
the Robotics and Artificial Intelligence in 2016.

Rather than overly focusing on the negative aspects however, Mr. Buston felt it is important to
be propositional and focus on what can be done to ensure the progression of this intelligence
revolution is consistent with these economic and social concerns. Regulating AI however, he
explained, would not be the solution, simply because regulating AI is as unrealistic as
regulating mathematics. If regulation is to be seriously considered, it would need to target
applications of AI to specific functions. Rather, Mr. Buston proposed the following steps: 1) AI
should be at the heart of the UK’s industrial and trade deals in a post-Brexit environment. 2)
Detailed and granular research on AI’s impact on the job market must be performed. 3) Smart
strategies to address automation's impact on the job market must be developed. 4) A universal
online self-employment system should be developed. 5) Radical reform of the educational
system should be undertaken.

In conclusion, Mr. Buston noted that we always hear about the need to avoid “hindering
innovation” and, while this is true, we must also start thinking about the need to avoid
“hindering our wellbeing”. The problem, he suggested, is that general understanding of the
real impact of robotics and AI is weak, and the media’s presentation of a dystopian robotic
future does not help the cause. To really affect change, the various concerned sectors must
organize themselves into a community and tackle the issues head on with an advocacy
strategy firmly focused on our economic and social wellbeing.

                                             Page 14 of 18
9. Long term issues of artificial intelligence and the
           future of humanity
Following Mr. Buston’s talk on some of the immediate issues associated with AI and robotics,
Mr. Kyle Scott of the Future of Humanity Institute (FHI) at the University of Oxford turned the
focus to longer-term issues. Referencing Prof. Nick Bostrom and his seminal works on AI, Mr.
Scott observed that, humans, as a species, exist in a relatively unstable state, somewhere
between extinction and utilising technological maturity to unlock cosmic endowment. These
are the two attractor states and mankind is on a balance beam between these attractors,
trying to avoid being drawn toward extinction.

Mr. Scott continued, noting that, if earthquakes and volcanos have not destroyed mankind
over the past thousands of years, we can assume the likelihood that they will destroy us in the
next 100 years is low. In this regard, if we as a species are to face an extinction-level event it
is more likely to be from a new challenge, with emerging technologies being primary
candidates.

Looking at the concept of artificial intelligence Mr. Scott explained that intelligence is what
separates us from gorillas; intelligence and cooperation enabled our species to take over this
planet, and an AI that far surpasses our own intelligence is likely to have a similar strategic
advantage. He went on to distinguish between the different types of AI: narrow intelligence,
which can exceed human performance in a narrow domain such as chess; general
intelligence, which can match human performance in a variety of domains; and
superintelligence, which greatly exceeds the performance of humans in virtually all domains of
interest. Deep Blue, Mr. Scott elaborated, was great at chess but could not play checkers. The
algorithms used to create AlphaGo on the other hand are already a more broad AI than Deep
Blue as they can achieve superhuman performance in a variety of games. If an AI is capable
of matching human-level performance in designing improved agents, it is plausible that this
could trigger an "intelligence explosion" resulting in an artificial "superintelligence".

Speculation    about    AI   is   however,   Mr.     Scott    explained,   significantly   affected   by
anthropomorphic bias, leading to erroneous analogies with humans. He continued, noting that,
humans tend to rationalise their behaviour and position in the world on the basis of concepts
like “Mind”, “Consciousness” and “Spirit”. Accordingly, some feel that true ‘artificial’ intelligence
can never really be achieved because an AI can never have these special characteristics that
are unique to us alone. Mr. Scott went on to point out that an agent does not need these
special characteristics in order to perform at a superhuman level. A heat seeking missile, for
example, is capable of tracking targets at a superhuman level of performance but does not

                                              Page 15 of 18
have ‘consciousness’. These anthropomorphic mistakes, he noted, can affect our ability to
understand and assess the potential of these technologies.

In terms of timelines, Mr. Scott noted that, while they do not like to give predictions, the Future
of Humanity Institute did conduct a general survey amongst eminent representatives of the AI
community, seeking to ascertain when they thought AI would arrive. Most experts felt that
there was a 50% probability that human-level machine intelligence would arise between 2020
and 2075. The median answer to the survey, he stated, was by 2040-2050. Mr. Scott
concluded, noting that there was also general consensus that, before we reach this stage,
there was a lot of work to be done in addressing the problems of AI safety.

       10. Robotics and artificial intelligence at the United
         Nations
In the final session, Mr. Irakli Beridze, Senior Strategy and Policy Advisor at UNICRI, talked
about the United Nations’ (UN) perspective on advancements in the field of robotics and AI. To
start, Mr. Beridze noted that to a large degree the public considers AI and Robotics as
something futuristic, something confined to the realm of science fiction. As we have already
heard many times today though, he observed, this is very much not the case. AI and robotics
are already all around us and are here to stay. Exemplifying this, he noted the rate of
technological change in both the civil and military settings from 2003 to 2017.

However, how has the world’s foremost international organization reacted to these
technological developments? Explaining the structure and nature of the UN system, Mr.
Beridze noted that there has not been a common unified approach to AI and robotics within
the UN. Notwithstanding this, a number of organizations and agencies within the UN system
have taken note of the advancements in the field of AI and robotics from their respective
positions. Perhaps most notably, as Prof. Sharkey also mentioned earlier, since 2014 there
has been an ongoing discussion of lethal autonomous weapons systems (LAWS), or so-called
“killer robots”, in the context of the United Nations’ Convention on Certain Conventional
Weapons (CCW). The establishment of a Group of Governmental Experts (GGE) in December
2016 to specifically discuss the legal, ethical and societal concerns of LAWS over a three-
week period later this year is a significant advancement in these discussions. The United
Nations Institute for Disarmament Research (UNIDIR), he added, has also supported the
discussion of LAWS at the CCW through its work on the weaponization of increasingly
autonomous technologies in the context of security and disarmament, work that has resulted

                                             Page 16 of 18
in the publication of a series of reports to frame the complex issues that surround LAWS in an
accessible manner.

Moving beyond the realm of conflict, he observed that the International Telecommunication
Union is set to explore the latest developments in AI innovation and their implications at its
upcoming AI for Good Global Summit in Geneva in June 2017, which will be the first of a
series of annual conferences on AI Innovation. The ITU has recognized the important role of
AI in the achievement of the United Nations' Sustainable Development Goals (SDGs) and in
helping to solve humanity's challenges by capitalizing on the unprecedented quantities of data
now being generated on sentiment behaviour, human health, commerce, communications,
migration and more. The United Nations Chief Information Technology Officer, Ms. Atefeh
Riazi, has also noted the world-changing potential of AI, underlining at the same time the
importance considering both the positive and negative moral and ethical implications and the
importance of crafting appropriate policies.

Looking closer to home, Mr. Beridze explained that UNICRI launched its own programme on
AI and robotics in 2015. The programme seeks to support the development of an international
infrastructure to identify and understand in greater detail the risks and benefits of
advancements in AI and robotics; to facilitate stakeholder discussions; and to support the
development of international and national approaches that minimize the risks and maximise
the benefits of AI. He then described some of UNICRI’s contributions in this regard, which
included the organization of a side-event during the 70th session of the UN General Assembly
in New York in October 2015 with the participation of renowned experts, including Prof.
Tegmark and Prof. Bostrom, to brief delegations on the current and likely future capabilities of
artificially intelligent systems. The event was repeated in 2016, with the support of 1QBit, the
FBI, SICPA, DSTL, and eminent ethicist and scholar Prof. Wendell Wallach. Following the
2015 event, the UN Group of Friends on Chemical, Biological, Radiological and Nuclear
(CBRN) Risk Mitigation and Security Governance acknowledged the increasing importance of
AI and robotics for international security and decided to remain seized of the matter. In
November 2015, Mr. Beridze noted, UNICRI also collaborated with the Organization for the
Prohibition of Chemical Weapons (OPCW) and The Hague Security Delta to organize a side-
event during the 20th session of the Conference of States Parties to the Chemical Weapons
Convention (CWC). During the event member states were briefed on the technological trends
and on what AI and robotics innovations might mean for the implementation of the Chemical
Weapons Convention. Subsequently, in March 2016, UNICRI launched the first edition of its AI
and robotics public-awareness and educational programme with a training course for media
and security professionals hosted by the Netherlands Institute of International Relations
Clingendael. Today’s workshop, Mr. Beridze observed, is the second edition of this

                                               Page 17 of 18
programme, which he hopes will continue into the future, becoming a more regular event and
spreading from Europe to Africa, the Americas, and Asia.

All this, Mr. Beridze explained, is building toward to opening of UNICRI’s Centre on AI and
Robotics later this year in the city of The Hague, the Netherlands. The Centre, he continued,
will serve to enhance understanding of the risk-benefit duality of AI and robotics through
improved coordination, knowledge collection and dissemination, awareness-raising and
outreach activities. Recognizing the potential for AI to contribute to the implementation of the
SDGs, Mr. Beridze said that one of the Centre’s immediate ambitions is to find ways to bring
practical AI tools to the developing world to support implementation of SDGs. Overall, it was
his hope that this Centre will support policy-makers in developing balanced and appropriate
policies.

Concluding remarks

Before bringing the event to a close, Mr. Beridze invited the speakers to deliver final remarks
on the most important take-home point from the workshop. In general, there was consensus
amongst the speakers that the current status quo with respect to AI and robotics should not
continue. Change is required and this change should have AI ethics at its core. Although there
is an increasing interest in the field, with more and more discussion of the various issues, what
is evidently still lacking is a collective strategy on how to proceed in addressing the many
economic, social, legal, and ethical concerns associated with AI and robotics. Fostering a
global AI and robotics community and securing the support of global governance are critical
steps in this regard. It was suggested that inspiration might be taken from the struggle of
climate change advocates against the economic approach to energy and how they
approached placing climate change on the international agenda, prompting the formulation of
international legal instruments such as the Kyoto Protocol to The United Nations Framework
Convention on Climate Change.

In conclusion, Mr. Beridze reflected on Dr. Karachalios’ keynote address, in which he
highlighted the importance of dialogue and building bridges amongst concerned communities,
and argued that international dialogue on risk and benefits of AI and robotics should continue.
Relying on the combined expertise in the room, he felt that they could bring techno-scientific
communities and UN Member States together on AI and robotics for the benefit of humanity.

With that, Dr. Tuveson and Mr. Beridze, on behalf of the Cambridge Centre for Risk Studies
and UNICRI, officially brought the event to a close, thanking the distinguished faculty of
speakers for their fascinating contributions and the participants for their active participation in
discussions.

                                             Page 18 of 18
The Risks and Benefits of Artificial Intelligence and
Robotics

A workshop for media and security professionals
Date:        6 - 7 February 2017; 09:00 - 17:00

Location: University of Cambridge Judge Business School
             Trumpington Street, Cambridge, UK CB2 1AG

Meeting Convenors:
   • Mr. Irakli Beridze, Senior Strategy and Policy Advisor, United Nations Interregional Crime and Justice
      Research Institute
   • Dr Michelle Tuveson, Founder & Executive Director, Cambridge Centre for Risk Studies at the University
      of Cambridge Judge Business School

   Workshop Agenda

     Monday 6 February 2017

     08:50 – 09:00 Welcome and Introductions, UNICRI and the Cambridge Centre for Risk Studies

     09.00 – 09.30 Keynote Address: Konstantinos Karachalios, Ph.D, Managing Director of The Institute of
     Electrical and Electronics Engineers (IEEE) Standards Association and Member of the Management Council of
     IEEE

     Session One:

     09:30 – 10:15 Artificial Intelligence and Robotics 101: What Is It and Where Are We Now?, Prof. Noel
     Sharkey, University of Sheffield, UK, Co-Founder of the Foundation for Responsible Robotics (FRR) and
     Chairman of the International Committee for Robot Arms Control (ICRAC) (TBC)

     10:15 – 10:45 Discussion moderated by Prof. Sharkey

     10:45 – 11:15 Coffee & Tea

         Hosted by:
Session Two:

11:15 – 12:00 Ethics and Artificial Intelligence, Ms. Kay Firth-Butterfield, Barrister-at-Law, Distinguished
Scholar, Robert S. Strauss Center for International Security and Law, University of Texas, Austin, Co-Founder,
Consortium for Law and Ethics of Artificial Intelligence and Robotics

12:00 – 12:45 Discussion moderated by Ms. Firth-Butterfield

12:45 – 13:30 Lunch in the common room.

13:30 – 14:15 Demo from Darktrace (in the large lecture theatre)

Session Three:

14:15 – 15:00    The Cyber-Security Overlap:

       The Triangle of Pain: The Role of Policy, Public and Private sectors in mitigating the Cyber Threat,
       Professor Daniel Ralph, Academic Director, Cambridge Centre for Risk Studies & Professor of Operations
       Research, University of Cambridge Judge Business School

       Modeling the Cost of Cyber Catastrophes to the Global Economy - Simon Ruffle, Director of
       Technology Research & Innovation, Cambridge Centre for Risk Studies

       Towards Cyber Insurance: Approaches to Data and Modeling - Jennifer Copic, Research Assistant,
       Cambridge Centre for Risk Studies

15:00 – 15:45    Discussion moderated by Dr. Michelle Tuveson, Executive Director, Cambridge Centre for Risk
Studies

15:45 – 16:15 Coffee & Tea

Session Four:

16:15 – 17:00 From Fear to Accountability - the State of Artificial Intelligence Journalism, Mr. John C.
Havens, Executive Director of the IEEE Global Initiative for Ethical Considerations in the Design of Autonomous
Systems and contributing writer for Mashable and The Guardian.

17:00 – 17:45 Discussion moderated by Mr. Havens

Tuesday 7 February 2017

09:00 – 09:15 Recap of First Day Takeaways, Mr. Irakli Beridze, UNICRI

Session Five:

09:15 – 10:00 Emerging Technologies: Quantum Computing, Dr. Natalie Mullin, 1Qbit

10:00 – 10:30 Discussion moderated by Dr. Mullin

10:30 – 11:00 Coffee & Tea

Session Six:

11:00 – 11:45 Economic and Social Implications of Robotics and Artificial Intelligence, Mr. Olly Buston,
Founding Director, Future Advocacy

11:45 – 12:30 Discussion moderated by Mr. Buston
12:30 – 14:00 Lunch

Session Seven:

14:00 – 14:45 Long term Issues of Artificial Intelligence and the Future of Humanity, Mr. Kyle Scott, the
Future of Humanity Institute (University of Oxford)

14:45 – 15:30 Discussion moderated by Mr. Scott

15:30 – 16:00 Coffee & Tea

Session Eight:

16:00 – 16:45 Robotics and Artificial Intelligence at the United Nations, Mr. Irakli Beridze, UNICRI

16:45 – 17:15    Discussion moderated by Mr. Beridze

Panel Discussion:

17:15 – 18:15 Open panel discussion moderated by Mr. Irakli Beridze, UNICRI, and Dr. Michelle Tuveson, the
Cambridge Centre for Risk Studies.

Panellists include:

    •   Dr. Ing. Konstantinos Karachalios
    •   Mr Olly Buston,
    •   Mr. John C. Havens,
    •   Dr. Natalie Mullin,
    •   Mr. Kyle Scott,
    •   Ms. Firth-Butterfield, and,
    •   Dr Stephen Cave, Executive Director, Leverhulme Centre for the Future of Intelligence, University of
        Cambridge
The Risks and Benefits of Artificial Intelligence and
Robotics
A workshop for media and security professionals
   Speakers’ Biographies

                        Senior Strategy and Policy Advisor at UNICRI, with more than 18 years of experience in
                        leading highly political and complex multilateral negotiations, developing stakeholder
                        engagement programmes and channels of communication with governments, UN agencies,
                        International Organizations, think tanks, civil society, foundations, academia, private industry
                        and other partners on an international level.

                        Prior to joining UNICRI served as a special projects officer at the Organisation for the
                        Prohibition of Chemical Weapons (OPCW) undertaking extensive missions in politically
                        sensitive areas around the globe. Recipient of recognition on the awarding of the Nobel
                        Peace Prize to the OPCW in 2013.

                        Since 2015, Initiated and heading the first UN programme on Artificial Intelligence and
                        Robotics. Heading the creation of the UN Centre on AI and Robotics with the objective to
   Mr. Irakli Berize    enhance understanding of the risk-benefit duality of AI through improved coordination,
                        knowledge collection and dissemination, awareness-raising and global outreach activities.
                        He is a member of various international task forces and working groups advising
                        governments and international organisations on numerous issues related to international
                        security, emerging technologies and global political trends.

                        Michelle Tuveson is a Founder and Executive Director at the Cambridge Centre for Risk
                        Studies hosted at the University of Cambridge Judge Business School. Her responsibilities
                        include the overall executive leadership at the Centre. This includes developing partnership
                        relationships with corporations, governments, and other academic centres. Dr Tuveson
                        leads the Cambridge CRO Council and she chairs the organising committee for the
                        Cambridge Risk Centre's Annual Risk Summits. She is one of the lead organisers of the
                        Aspen Crisis and Risk Forum. She is an advisor to the World Economic Forum's 2015
                        Global Risk Report and a contributor to the Financial Times Special Report on Risk
                        Management. She is also an advisor to a number of corporations and boards as well as a
 Dr. Michelle Tuveson   frequent conference speaker.

   Hosted by:
Dr Tuveson has worked in corporations within the technology sector with her most recent
                        position in the Emerging Markets Group at Lockheed Martin. Prior to that, she held positions
                        with management strategy firm Booz Allen & Hamilton, and US R&D organisation MITRE
                        Corporation. Dr Tuveson's academic research focusses on the application of simulation
                        models to study risk governance structures associated with the role of the Chief Risk Officer.
                        She was awarded by the Career Communications Group, Inc. as a Technology Star for
                        Women in Science, Technology, Engineering and Maths (STEM). She earned her BS in
                        Engineering from the Massachusetts Institute of Technology, MS in Applied Math from
                        Johns Hopkins University, and PhD in Engineering from the University of Cambridge. She is
                        a member of Christ's College Cambridge.

                        A globally recognized leader in standards development and intellectual property, Dr. Ing.
                        Konstantinos Karachalios is managing director of the IEEE Standards Association and a
                        member of the IEEE Management Council.

                        As managing director, he has been enhancing IEEE efforts in global standards development
                        in strategic emerging technology fields, through technical excellence of staff, expansion of
                        global presence and activities and emphasis on inclusiveness and good governance,
                        including reform of the IEEE standards-related patent policy.

                        As member of the IEEE Management Council, he championed expansion of IEEE influence
                        in key techno-political areas, including consideration of social and ethical implications of
                        technology, according to the IEEE mission to advance technology for humanity. Results
                        have been rapid in coming and profound; IEEE is becoming the place to go for debating and
                        building consensus on issues such as a trustworthy and inclusive Internet and ethics in
                        design of autonomous systems.

Dr. Ing. Konstantinos   Before IEEE, Konstantinos played a crucial role in successful French-German cooperation
     Karachalios        in coordinated research and scenario simulation for large-scale nuclear reactor accidents.
                        And with the European Patent Office, his experience included establishing EPO’s patent
                        academy, the department for delivering technical assistance for developing countries and
                        the public policy department, serving as an envoy to multiple U.N. organizations.
                        Konstantinos earned a Ph.D. in energy engineering (nuclear reactor safety) and masters in
                        mechanical engineering from the University of Stuttgart.

                        Noel Sharkey PhD DSc FIET FBCS CITP FRIN FRSA Emeritus Professor of AI and
                        Robotics University of Sheffield, co-director of the Foundation for Responsible Robotics
                        http://responsiblerobotics.org and chair elect of the NGO: International Committee for Robot
                        Arms Control (ICRAC) http://icrac.net. He has moved freely across academic disciplines,
                        lecturing in departments of engineering, philosophy, psychology, cognitive science,
                        linguistics, artificial intelligence, computer science, robotics, ethics, law, art, design and
                        military colleges. He has held research and teaching positions in the US (Yale and Stanford)
                        and the UK (Essex, Exeter and Sheffield).

                        Noel has been working in AI/robotics and related disciplines for more than 3 decades and is
                        known for his early work on neural computing and genetic algorithms. As well as writing
                        academic articles, he writes for national newspapers and magazines. Noel has created
                        thrilling robotics museum exhibitions and mechanical art installations and he frequently
                        appears in the media and works in popular tech TV shows such as head judge of robot
                        wars. His research since 2006 has been on ethical/legal/human rights issues in robot
 Prof. Noel Sharkey
                        applications in areas such as the military, child care, elder care, policing, autonomous
                        transport, robot crime, medicine/surgery, border control, sex and civil surveillance. A major
                        part of his current work is advocacy (mainly at the United Nations) about the ethical, legal
                        and technical aspects of autonomous weapons systems.
You can also read