Artificial intelligence,

real benefits

There are many stereotypes and preconceptions about artificial intelligence, fuelled in no small part
by science fiction books and films. But in reality AI should be considered as a tool, albeit a highly
sophisticated one that is constantly evolving and improving as our human intelligence deepens.
Following the launch of the EU’s coordinated plan on artificial intelligence to accelerate the development
of this key technology for the benefit of all, Roberto Viola takes a look at how AI is in fact already very
much a part of our day-to-day lives.

What makes AI different from any other type

of tool is the ability to learn and act
accordingly. In the same way that human intelligence has
The popular culture view of AI is of intelligent robots
– essentially artificial bodies driven by artificial intelligence –
and it is certainly true that these two technologies can be
allowed us to flourish as a species by turning our collective put together for a myriad of uses. Robots have been used
hand to pretty much anything, it is the ability of artificial for many years to do things that humans cannot or do
intelligence to improve so many different aspects of our lives not want to do, from repetitive tasks on a production line
that is so exciting. to defusing explosives. But adding AI into the mix vastly
expands the scope of what a robot can do. For example, one
The other enduring image of AI is that of a future European consortium has developed an AI-driven robot that
technology (even if many of the popular culture examples can learn everyday tasks such as making pizza or reloading
set their AI-dominated world of the future in what is now the printer, things that are not difficult for humans to do but
our past). But in fact AI is already a day-to-day reality for which have largely been beyond the reach of even the most
many of us, from apps that know what kind of music we like sophisticated robots until now. And if this was not enough to
Updated 18 December 2018

without us asking to ‘personal assistants’ on our smartphones fill the stereotype of the domestic robot of the future, lending
or in our homes that can seemingly answer any question we a hand around the home like any human family member, at
may have in a matter of seconds. Yet these simple examples least two different European research projects are working
are just scratching the surface of what AI can do. Artificial on developing robots that actually dream, processing
intelligence is already making us healthier, giving us cleaner experiences during ‘down time’ and potentially learning how
air and energy, keeping us safe online or on the move and to react to new and different situations.
improving the quality of our work.

1
Keeping us healthy best-known example is probably the MareNostrum
supercomputer in Barcelona, Spain, immortalised by
If the domestic robot is the most common image of novelist Dan Brown in his book Origin, where it is the
AI, the technology is in fact being used to help humans computing power behind his fictional AI character, Winston.
in many other fields, allowing us to push the boundaries
of our own knowledge and skills. Nowhere is this more MareNostrum is very much real, however, and its
evident, perhaps, than in the field of healthcare, where AI ever-increasing computational power is being used in
can help both doctors and patients. a number of ways to improve our understanding of the
human body. For example, it can analyse huge quantities
One European project is tackling the issue of breast of genomic data to broaden our understanding of how
cancer screening, for example. Current screening techniques cancer affects the human body, and to develop protein
lead to 10-20% of patients being wrongly informed that and drug models that could lead to personalised medical
they do not have breast cancer, and tissue-sampling treatments based on each patient’s specific needs.
methods are lengthy and often inaccurate. Researchers are MareNostrum is also being used to develop extremely
now developing a more accurate technique that overlays accurate virtual models of the human heart, helping
MRIs scans taken by AI-driven robots with images from doctors and surgeons learn more about the way this vital
ultrasound and pressure sensors to give a much clearer organ works – a project that is so cutting-edge that Brown
view of potentially diseased areas, in turn making it easier specifically mentions it in Origin.
to target the area for tissue samples. Another European
project is working on a remote surgery robot to assist Meanwhile, a team of scientists from the Jülich
surgeons, using virtual reality and AI to understand the supercomputing centre in Germany are leading a
surgeons’ intentions and perform some of the simple tasks pan-European project to designed to improve artificial
carried out by their human assistants during surgery. intelligence understanding more effectively how the
human brain works. The Human Brain Project, which also
includes supercomputers from across the EU, including
MareNostrum, is not only developing extremely detailed
models of the human brain, but also using complex
simulations to learn how the brain actually works. This
ambitious and ground-breaking project has massive
potential benefits for both human healthcare and
advancing the development of AI in the future.

Keeping us moving and safe

If there is one autonomous, intelligent machine that
has particularly captured the imagination of the public in
recent times, it is the self-driving car. There are more than
1.3m road accident-related deaths each year according to
the World Health Organisation (of which around 25,000
were in the EU), and cutting out the human error that is so
Where machines really can outperform humans is often the cause of these tragedies is the chief goal of the
in their capacity to process and analyse much larger self-driving car.
quantities of data. Even ‘normal’ AI can work more rapidly
than a human brain – although human input is always
needed at the beginning to define the algorithms the AI
uses in its analysis – but in some cases the quantities of
data, and the complexity of the issue, require an entirely
superior level of computational power behind the AI.

The complexities of how the human body works

is a case in point. To effectively model and analyse
this most complicated of ‘machines’ requires a
much bigger computer ‘brain’ in many cases, which
is where the supercomputer comes in. There are an
increasing number of these behemoths appearing
across the world, including several in Europe where the

2
 Human drivers have to process huge quantities of data
simultaneously in order to drive safely. Cars driven using
AI work in much the same way. The human eyes, ears, feet
and hands are replaced by sensors, but the data collected
sends back all the same signals: how close the car is to
other vehicles, what speed is it travelling, the behaviour of
other road users such as cyclists or pedestrians, etc. One
European project is working on developing a system that
will allow cars to switch seamlessly between their human
and automated drivers, based on data from real drivers in
a variety of driving conditions and cases.

What is particularly interesting about this project is its

focus on the insurance industry: concerns about the ability
to guarantee the safety of all road users is one of the main A new supercomputer based in Bologna and due to come
reasons why we do not already have fleets of driverless online in 2020 will focus on improving weather predictions,
or self-driving cars on our roads. Therefore reassuring and it is no coincidence that Italy will be the host country
insurers that autonomous driving is synonymous with for the latest generation of weather supercomputer which
safer driving will be key. has been located in Reading, UK, since the late 1970s.
Italy has suffered in particular from some of the most
AI is also being used to keep road users safe in other devastating environmental disasters in recent times,
ways as well. A European project has looked whether a great human and economic cost, and the hope is that
adaptive cruise control, an AI-driven ‘smart’ system that improving our ability to predict when and where floods,
speeds up or slows down a car as necessary to keep up fires and earthquakes will happen will help save lives.
with the one in front, could help reduce the risk of traffic
jams and accidents. No matter how effective the predictions become,
however, it will never be possible to avoid the disasters
AI will also have a major role to play in the ‘smart cities’ altogether, and AI can also be used to improve our human
of the future in helping to control traffic flow, for example response to crises. For example, the AI-driven SmokeBot
by adapting traffic light rotations in real time to meet on- robot can help firefighters in search and rescue missions
the-ground traffic flow demands, or by analysing available where visibility is extremely low.
parking spaces and updating information systems. In fact, AI
will have many potential applications in cities, from helping
to identify people in crowds of people based on face-
recognition software – helpful in preventing terrorist attacks
or finding a child lost in the crowd – to analysing air quality
conditions and taking appropriate action such as limiting
traffic flows or sending out automatic alerts to citizens (all
applications already tested in various EU-funded projects).

Protecting us
and our environment
Accelerating the shift to less polluting cars and reducing
the risk of peaks in emissions caused by traffic jams is just
one way in which artificial intelligence can help protect us Artificial intelligence is also being used to drive the
and our environment. energy transition that is so vital for our society. The
hills around Barcelona are home to a large number of
Weather and climate is one particular field where we windfarms that help power the city with sustainable
have huge quantities of both historical and real-time data, energy, and MareNostrum’s computing power is being put
and where the artificial intelligence and computational to good use in accurately mapping the best location for
power of supercomputers can be combined to develop new turbines. Working with the Spanish energy company
extraordinarily accurate models. In a world where climate Iberdrola, the Barcelona Supercomputing Center is
and weather systems are changing more rapidly than analysing geographical and weather data to predict the
ever, it is more vital than ever before to be able to maximum gain in energy from each new turbine, vastly
accurately predict the extreme weather conditions that improving efficiency.
have such a devastating effect.

3
 Contributing to improving efficiency and reducing
emissions are also an important part of the use of AI in
the agriculture sector. In an industry where much of the
work is repetitive, intensive and hard, artificial intelligence
can help in a number of areas. For example, smart robots
already exist that can tell whether crops are ripe and
then harvest them, while data from sensors in the fields
can be processed using AI to automatically plant, irrigate
and fertilise crops using precise amounts. The work on
improving weather predictions is also important for the
food and farming sector, where farmers’ crops can be
destroyed by one bad storm or drought.

Getting the information and advice we need also often

means talking to companies, service suppliers or local
authorities, and AI can anticipate and respond to people’s
needs in a variety of different ways. One European project
is working on a platform that can be embedded in any
public administration’s website and improve interaction
with the public through a highly sophisticated question-
and-answer service, meaning people can get the answers
they need from their local authorities quickly, simply and
at a time that suits them.

And as the homes, workplaces, streets and spaces of

Automated systems to reduce the need for human the future become increasingly connected – the so-called
interaction in difficult or dangerous situations are not Internet of Things – our interaction with them will become
limited to farming. From robots that can inspect and clean increasingly virtual and driven by artificial intelligence.
sewers to aerial robotic systems that can both inspect and We will move from the mass-production and delivery of
maintain hard-to-reach machinery – not to mention AI services, such as electricity or water, to smarter, more
simulators of the conditions on the planet Mars – artificial personalised systems, bringing important resource
intelligence is being used to help humans in a very wide savings. Pilot projects across Europe are already working
range of conditions and industries. on addressing the massive infrastructure and coordination
challenges that this brings, paving the way for the smart
society of the future with data and AI at its heart.

Helping us learn and interact

One of the most common areas where AI is used widely
is in learning – after all, most of us come across AI every
day through our smartphone apps and search engines
that help us advance our knowledge and find information.

Artificial intelligence is being used to turn science

fiction into science fact in this area as well. Until recently,
real-time translation piped directly into our ears was a
convenient way of explaining why every species in the
galaxy spoke the same language as Captain Kirk, but
European cooperation
advances in machine translation, driven by AI, are making and support
this a real-world reality, at least for human languages.
So what do all the projects and initiatives mentioned
And in a world where people increasingly get their news here, beyond the obvious link with machine learning or
and information online and from a very wide range of artificial intelligence, have in common? The answer is that
source, a European project is using artificial intelligence as all of them are supported to a greater or lesser extent by
the basis for systems that can separate fact from fiction the European Union, and are part of Europe’s ambitious
automatically, addressing issues of disinformation and aim of building the world’s leading digital society.
‘fake news’ in today’s multilingual online environment.

4
 The EU has been supporting artificial intelligence for only be able to do this if the conditions are right across
many years, in no small part through its research and the whole of our society. The EU’s single market, with a
development programmes, FP7 and Horizon 2020. As the common set of standards, rules and protections, is a rich
technology has advanced, so has the support, and AI-related breeding ground for this to happen, and Europe is uniquely
projects are now funded through a myriad of different placed to provide the joined-up policy-making it will need.
programmes, covering everything from rural development
to education and skills. For the next seven-year EU budget
period, which is due to start in 2021, AI and the wider
digital economy will play an even more central role: a new
funding programme, Digital Europe, has been proposed,
with €9.2 billion potentially available to support the further
development of the EU’s digital single market.

AI and the supercomputing capacity that it is increasingly

linked to and reliant on are set to receive the lion’s share
of the proposed funding – €2.5 billion and €2.7 billion
respectively – underlining the particular importance that
the EU is giving to this key technology with so many
possibilities. But AI will also benefit from other funding
opportunities within the Digital Europe programme,
for example from a share of the €2 billion set aside to
boosting cybersecurity and keeping data sage, from the In our hypothetical robot of the future, AI and
€700 million proposed for developing digital skills or the supercomputers are obviously the brains, fuelled by the
€1.3 billion for ensuring the widespread use of digital data flowing through them; the EU plays the role of the
technologies by public administrations and services. heart, making sure the ‘body’ gets the resources it needs.
The veins and arteries, through which the data flows, are
represented by the EU-wide rules on how data can be
collected, transferred and used; the antibodies that keep
the data safe from harm are the EU’s rules on networks
and cybersecurity, and general data protection rules.
The nerves are high-speed broadband and 5G, sending
information back and forth at ultra-high speeds to keep
the organism moving.

The EU also works to keep our robot safe and healthy.

Humans need food to survive and we have developed
sophisticated supply chains – from farm to fork – to help us;
AI will need the same security of supply to ‘feed’ it with data.
EU-wide pilot projects focused on the Internet of Things are
already working on how to do this; data is of no use if it
Although this is the first time the European Commission cannot be collected, pooled, transferred and accessed simply
has proposed a dedicated funding stream for digital, and easily, and the right infrastructure and framework need
it will not be the only programme supporting artificial to be in place across the EU for this to happen.
intelligence development. The Horizon Europe research
programme will continue to fund much of the top-level Humans have an ethical code, an understanding of
research in this area, while several of the sector-specific what is right and wrong, and a system of laws to support
funding programmes can also be potential sources of it. A similar code and set of rules will be needed for AI as
support. For example, some of the food and farming well. The EU is working on developing guidelines for the
applications could be supported through Common use of AI and updating rules on product liability, trying to
Agricultural Policy funding. find answers to such thorny questions as what the limits
should be on the use of AI or who is responsible if a self-
driving car is hacked and has an accident.
More than just money And it is important to recognise that an AI-centred
Important – not to say vital – as the EU’s continued society of the future will not necessarily be a utopia for
financial support will be to the development of AI, the EU’s everyone, at least not at first. If artificial intelligence can
role goes far beyond just the money. AI has the potential take over many tasks currently performed by humans, it is
to improve our lives in almost everything we do, but it will going to be very important to make sure that the humans

5
have the opportunity to do something else instead. The (despite what science fiction would have us believe), the
goal is not for AI to push people out of work but rather experts behind the guidelines underline the need to ensure
to improve the quality of their work, by allowing them that this does not happen. The concern is not, however,
to work less or in much better conditions, for example. that a self-conscious machine would somehow ‘take over’,
The EU can help in numerous ways, through support but rather that it would effectively be an ‘ethical object’
to help people learn new skills for example, or through and therefore subject to the same values as humans.
competition rules that ensure that a robot workforce
cannot undercut a human one. The guidelines also stress the importance of making
sure the technology behind AI is sufficiently robust.
Algorithms have to be reliable (we need to be sure the
AI will do what we want it do, each time it is used) and
Ethical guidelines strong enough to protect against attacks. There should
also be a ‘fall-back’ plan in case of problems with AI
Humans have an ethical code, an understanding of (for example, AI systems that fail should ask for human
what is right and wrong, and a system of laws to support intervention before continuing).
it. A similar code and set of rules is needed for AI as
well. The EU’s High Level Expert Group on AI, a team Finally, the guidelines also set out how these principles
of experts from academia, industry and government, can be maintained (and the problems avoided) by
is developing a set of ethics guidelines that will ensure following certain procedures such as making sure that
that trustworthy AI adheres to human rights, principles privacy and safety are guaranteed from the outset and
and values. The first version of the guidelines have just that testing and trialling is rigorous. They also stress that
been published and set out how AI should adhere to basic EU rules and regulations will be necessary to underpin the
principles: that AI should not be used to harm humans in ethical guidelines, as well as an appropriate framework
any way; that it should not restrict human freedom (in for accountability, the development of codes of conduct
other words, people should not be subjugated or coerced and awareness raising about the ethical implications of
by AI-driven machines); that it should be used fairly and artificial intelligence.
not to discriminate or stigmatise; that it should operate
transparently (in other words, we need to know for what
purpose AI is being developed and how it will be used);
and that it should only be developed for the wellbeing of Walking the same path
individuals and society as a whole, including in the most
sustainable manner possible. Every EU country is committed to supporting the
development of AI in Europe and several already have
national strategies on how to do so. Making the connection
between these various national strategies is the core of
the EU’s strategy on AI: coordinating and strengthening
the work of the various AI research centres across Europe;
supporting the development of AI applications across
every sector of society; and pooling resources at the EU
level with access for everyone.

For all its ambition, the EU is still lagging behind other

parts of the world when it comes to investing in AI. But this is
changing. The European Commission has already agreed to
The guidelines also address a number of issues that increase EU research funding for AI to €1.5 billion between
need to be considered when developing AI, for example now and 2020; expected investments from Member States
that it should not be used to identify individuals from and from the private sector should bring this figure closer to
their data without consent, that humans must always be €20 billion for the same period. If the Digital Europe funding
able to override decisions taken by machines and that programme is agreed, EU support for AI from 2021-2027
humans must always be aware when they are interacting will be €7 billion (€1 billion per year), and the aim is to
with a machine. The experts also back a recent resolution combine this with national funding and investment by the
by the European Parliament against the development private sector to reach around €20 billion a year.
lethal autonomous weapons systems - learning machines
with cognitive skills that can potentially decide who, when Now the Commission is publishing its coordinated
and where to fight without human intervention. plan on artificial intelligence, its proposals on where to
target these investments and how best to coordinate the
And while they stress that the likelihood of AI-driven development of AI across the EU while keeping it squarely
machines becoming self-conscious is extremely small focused on human needs.

6
 Each EU country will have to put in place a national While the action plan focuses squarely on AI, the EU
AI strategy – and commit funding to support it has also committed in parallel to investing heavily in the
– by mid-2019; the Commission and Member States supercomputing power needed to move AI to the next
will work together with academia and industry to develop level. The EuroHPC (for High Performance Computing)
a common research and innovation agenda for AI, and joint undertaking is designed to develop a pan-European
a new fund to support European AI start-ups will be supercomputing infrastructure and supporting research
developed, with €100 million available by 2020. Networks and innovation in the field. The simple fact of the matter
of European AI research excellence centres will be created is that global supercomputing is dominated by the US and
to improve cooperation between the best research China, and very few of the EU’s installations can currently
teams in Europe and to help develop world-leading AI rival their counterparts in these countries. EuroHPC’s aim
testing facilities. Private and public sector access to AI is to create at least two supercomputers that will be in the
applications and technologies will be improved using top five at the global level by 2020, with an investment
the existing EU-wide network of Digital Innovation of at least €1 billion, and three to four others that would
Hubs, which already act as the link between academia be in the top 25. There are currently seven European
and research and the wider commercial market. Making supercomputers in the global top 25, including all those
sure Europe has the skills and the expertise necessary to in the five host countries of the Partnership for Advanced
support this vision is also a key part of the proposals, with Computing in Europe (PRACE) which represents the best
a focus on developing and retaining more home-grown that Europe has to offer in supercomputing. These include
talent in AI, and on retraining. the supercomputers in Barcelona and Bologna, the JUWELS
supercomputer in Jülich (plus two others in Germany) and
the Tera 1000-2 supercomputer in Paris, as well as the Pix
Daint in Zurich, Switzerland. Together, these five hosting
members and 20 other country members are working
on more than 600 scientific projects using European
high performance computing, including many related to AI.

Putting the human

back in artificial intelligence
Artificial intelligence is a tool – albeit a highly
sophisticated one – that can help improve the quality of
live for everyone, and as such something to be excited
about rather than to fear.

Since data is the lifeblood of AI, there is a particular But like any tool, artificial intelligence still needs
focus on improving the collection of and access to high human interaction to function, and it is this human focus
quality datasets across the EU and in every domain. For on AI that underpins the EU’s approach. Developing
public data, for example, this means identifying which artificial intelligence is not an end in its own right - it has
data sets can be made available (anything from public to serve the greater good of society. That AI can do this is
transport movements to public utility information), undeniable; it is now up to Europe to make sure the right
making sure they are open and accessible to everyone conditions are in place to make it happen. n
and providing the space to store them safely on the cloud.
For health data such as patient information, medical Roberto Viola
records and diagnostic results, meanwhile, the aim is to Director General, DG Communication Networks,
develop a common database of health images. Content and Technology, European Commission

In the longer term, the EU will invest jointly with Member

States and the private sector in the creation of a common
European Data Space that makes data easily available for
re-use to innovators, businesses and public sector. More
All images © Getty / iStock

immediately, an AI-on-demand platform will be created to

pool knowledge and expertise on AI, including algorithms
and tools, to help expand access to the existing body of
work on AI in the EU. The proposals also reaffirm the EU’s
commitment to developing ethical guidelines for AI and
encouraging the take up of AI by the public sector.

7
 EU-supported projects mentioned
3rdhandrobot.eu - robotic third hand VineRobot - remote sensing of grape ripeness
https://cordis.europa.eu/project/rcn/110160_en.html https://cordis.europa.eu/news/rcn/128526_en.html
RoboHow - robots that learn http://www.robohow.eu ERMES - crop management
https://cordis.europa.eu/project/rcn/188830_en.html
Robdream http://robdream.eu & Dream http://www.
robotsthatdream.eu - robots that dream ECHORD++ - sewer inspections
http://echord.eu/pdti/pdti-urban-robotics-sewer-inspection/
MURAB - MRI and ultrasound assisted biopsy
https://www.murabproject.eu Aeroarms - aerial inspection and maintenance
https://aeroarms-project.eu/
SMARTsurg - remote surgery
http://www.smartsurg-project.eu iMars - visualising the Red Planet
https://cordis.europa.eu/project/rcn/188855_en.html
SMuFIN - genome identification http://cg.bsc.es/smufin
FANDANGO - fake news discovery https://fandango-project.eu/
AlyaRed - computational heart
https://www.bsc.es/news/bsc-news/alya-red-computational-heart Simpatico - public administration interface
https://www.simpatico-project.eu/
KConnect - medical database http://www.kconnect.eu
Internet of Things Large Scale Pilots
VI-DAS - human and automated driving http://vi-das.eu/
https://european-iot-pilots.eu/,
Optimum - mobility and traffic flow including MONICA https://www.monica-project.eu/ on
http://www.optimumproject.eu/ security at major events
and Synchronicity http://synchronicity-iot.eu/
TRAMAN21 - traffic management using adaptive
on the cities of the future
cruise control http://www.traman21.tuc.gr/
EuroHPC - supercomputer joint undertaking
SmokeBot - robot for low-visibility sites
https://eurohpc-ju.europa.eu/index.html
http://130.243.105.49/Research/mro/smokebot/
PRACE - supercomputer partnership http://www.prace-ri.eu/
SEDAR - wind farm production capacity simulation
https://www.iberdrola.com/top-stories/iberdrola-shares-with-you/
installation-wind-turbines