BRIEFING
EPRS | European Parliamentary Research Service
Author: Marcin Szczepański
Members' Research Service
PE 637.967 – July 2019
EN
Economic impacts of artificial
intelligence (AI)
SUMMARY
Artificial intelligence plays an increasingly important role in our lives and economy and is already
having an impact on our world in many different ways. Worldwide competition to reap its benefits
is fierce, and global leaders – the US and Asia – have emerged on the scene.
AI is seen by many as an engine of productivity and economic growth. It can increase the efficiency
with which things are done and vastly improve the decision-making process by analysing large
amounts of data. It can also spawn the creation of new products and services, markets and
industries, thereby boosting consumer demand and generating new revenue streams.
However, AI may also have a highly disruptive effect on the economy and society. Some warn that
it could lead to the creation of super firms – hubs of wealth and knowledge – that could have
detrimental effects on the wider economy. It may also widen the gap between developed and
developing countries, and boost the need for workers with certain skills while rendering others
redundant; this latter trend could have far-reaching consequences for the labour market. Experts
also warn of its potential to increase inequality, push down wages and shrink the tax base.
While these concerns remain valid, there is no consensus on whether and to what extent the related
risks will materialise. They are not a given, and carefully designed policy would be able to foster the
development of AI while keeping the negative effects in check. The EU has a potential to improve
its standing in global competition and direct AI onto a path that benefits its economy and citizens.
In order to achieve this, it first needs to agree a common strategy that would utilise its strengths and
enable the pooling of Member States' resources in the most effective way.
In this Briefing
Context
Economic potential of AI
Impact of AI on manufacturing
Effects of AI on firms, industries and
countries
AI impacts on labour markets and
redistributive effects of AI
Selected policy implications of AI
EPRS | European Parliamentary Research Service
2
Context
Artificial intelligence (AI) is a term used to describe machines performing human-like cognitive
processes such as learning, understanding, reasoning and interacting. It can take many forms,
including technical infrastructure (i.e. algorithms), a part of a (production) process, or an end-user
product. AI looks increasingly likely to deeply transform the way in which modern societies live and
work. Already today, smartphone smart assistants, such as Siri, perform a variety of tasks for users;
furthermore, all Tesla cars are connected and things that any one of them learns are shared across
the entire fleet. AI also matches prices and cars when one orders an Uber ride, and curates what
social media offer to a user based on their past behaviour. With the rise of AI come the important
questions of how much it will affect businesses, consumers and the economy in more general terms.
Employees are increasingly interested in knowing what AI means for their job and income, while
businesses are also keen to find ways in which they can capitalise on the opportunities presented
by this powerful phenomenon. There is a global accord that AI technologies have the potential to
revolutionise production and contribute to addressing major global challenges, a view shared by
organisations such as the
OECD and the European Commission.
Rapidly increasing computing power and connectedness have made it possible to compile and
share large volumes of valuable data, which is now more accessible than ever before. This has
created momentum for AI technologies. Importantly, AI patents have been on the rise worldwide
(see Figure 1), with a 6 % average yearly growth rate between 2010 and 2015, which is higher than
the annual growth rate observed for other patents.
The countries at the forefront of research during this period were Japan, South Korea and the United
States, which together accounted for almost two-thirds of AI-related patent applications. South
Korea, China and Chinese Taipei have recorded a remarkable increase in the number of AI patents
compared to their past results. EU Member States contributed 12 % of the total AI-related inventions
over 2010-2015, a decrease from the 19 % recorded in the previous decade.
A 2019 report on AI
by the World Intellectual Property Organization (WIPO) shows that there has
been a boom in the number of scientific papers in the field since the start of the century, followed
by an upsurge in patent applications between 2013 and 2016. This could indicate a switch from
theoretical research to the practical application of AI technologies in commercial products and
Figure 1 – AI patents worldwide, 2000-2015
Source: OECD, Science, Technology and Industry Scoreboard, 2017.
0
2
4
6
8
10
12
14
16
18
20
2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015
Thousands
Artificial intelligence (AI) patents
-5
5
15
25
%
Annual growth
0 10 20 30 40
I ND
GB R
CAN
FRA
DEU
T WN
CHN
EU28
USA
KOR
JPN
%
Top economies' shares
in AI-related patents
Economic impacts of artificial intelligence
3
services. The WIPO reckons that the large number of patents in machine learning shows that this is
currently the main application field of AI, while deep learning (used, for example, in speech
recognition) and neural networks are the fastest-growing technologies. The OECD also attributes
recent progress in AI to the development of deep learning using artificial neural networks.
The WIPO report reveals that the largest number of AI-related patents is in areas such as
telecommunications, transport, life- and medical sciences, and personal devices that compute
human–computer interaction. Smart cities, agriculture, e-government, banking and finance are the
most dynamically growing areas of application. The WIPO report also highlights the dynamic growth
in the number of AI patents registered by China, pointing out that since 2014, it has recorded the
highest number of first-patent filings. According to the WIPO, China, the US and Japan together
account for 78 % of total AI-related-filings, while between 2000 and 2015 almost one in five AI patent
families featured a European country.
1
Some argue that in the AI race, the EU has a structural disadvantage: a lack of scale manifested by a
lack of a huge homogenous pool of data
, which is an essential precondition for a thriving AI
ecosystem. In the EU, the level of AI uptake by companies is low, and AI-related investment and
patent numbers are lagging behind the US and Asia. However, the EU has the potential to leverage
its high value-added manufacturing and industry base and use its well-qualified workforce to
improve its global position. It can also use its
regulatory prowess and clout to become a global
leader in AI governance, and use tools, such as standards, to its advantage. Some see developed EU
countries, particularly northern European ones, as the inevitable winners in the global AI revolution.
Taking into account the fierce global competition in AI, the European Commission maintains that a
solid coordinated framework is necessary to advance European efforts in this undoubtedly
promising sector, an urgency
recognised by many EU Member States. It also considers AI one of the
most strategic technologies of the 21st century.
Economic potential of AI
The majority of studies emphasise that AI will have a significant economic impact. Research
launched by consulting company Accenture
covering 12 developed economies, which together
generate more than 0.5 % of the world's economic output, forecasts that by 2035, AI could double
annual global economic growth rates. AI will drive this growth in three important ways. First, it will
lead to a strong increase in labour productivity (by up to 40 %) due to innovative technologies
enabling more efficient workforce-related time management. Secondly, AI will create a new virtual
workforce – described as 'intelligent automation' in the report – capable of solving problems and
self-learning. Third, the economy will also benefit from the diffusion of innovation, which will affect
different sectors and create new revenue streams.
A study by PricewaterhouseCoopers (PwC) estimates that global GDP may increase by up to 14 %
(the equivalent of US$15.7 trillion) by 2030 as a result of the accelerating development and take-up
of AI. The report anticipates the next wave of digital revolution to be unleashed with the help of the
data generated from the Internet of Things (IoT), which is likely to be many times greater than the
data generated by the current ‘Internet of People’. It will boost standardisation and consequently
automation, as well as enhancing the personalisation of products and services. PwC sees two main
channels through which AI will impact on the global economy. The first involves AI leading to
productivity gains in the near term, based on automation of routine tasks, which is likely to affect
capital-intensive sectors such as manufacturing and transport. This will include extended use of
technologies such as robots and autonomous vehicles. Productivity will also improve due to
businesses complementing and assisting their existing workforce with AI technologies. It will
require investing in software, systems and machines based on assisted, autonomous and
augmented intelligence; this would not only enable the workforce to perform its tasks better and
more efficiently but would also free up time allowing it to focus on more stimulating and higher
value-added activities. Automation would partially remove the need for labour input, leading to
productivity gains overall.
EPRS | European Parliamentary Research Service
4
Eventually, the second channel – the availability of personalised and higher-quality AI-enhanced
products and services – will become even more important, as this availability is likely to boost
consumer demand that would, in turn, generate more data. Or, as
PwC puts it: 'in turn, increased
consumption creates a virtuous cycle of more data touchpoints and hence more data, better
insights, better products and hence more consumption'. Although the benefits will be felt globally,
North America and China are expected to gain the most from AI technology (see Figure 2). The
former will likely introduce many productive technologies relatively soon, and the gains will be
accelerated by advanced readiness for AI (of both businesses and consumers), rapid accumulation
of data and increased customer insight.
It is likely to take more time for China to feel the full effect of AI, but this effect will eventually occur
in the country's huge manufacturing sector and then move up the value chain into more
sophisticated and high-tech-driven manufacturing and commerce. Europe will also experience
significant economic gains from AI, while developing countries are likely to record more modest
increases due to lower rates of adoption of AI technologies.
2
The McKinsey Global Institute
expects that around 70 % of companies would adopt at least one type
of AI technology by 2030, while less than half of large companies would deploy the full range.
McKinsey estimates that AI may deliver an additional economic output of around US$13 trillion by
2030, increasing global GDP by about 1.2 % annually. This will mainly come from substitution of
labour by automation and increased innovation in products and services. On the other hand, AI is
likely to create a shock in labour markets and associated costs needed to manage labour-market
transitions; this shock would be incurred as an effect of negative externalities such as loss of
domestic consumption due to unemployment.
A 2016 study by Analysis Group (funded by Facebook), considers that AI will have both direct and
indirect positive effects on jobs, productivity and GDP. Direct effects will be generated by increased
revenues and employment in firms and sectors that develop or manufacture AI technologies, which
may also create entirely new economic activities. Indirect ones will come from a broader increase of
productivity in sectors using AI to optimise business processes and decision-making, as well as
increase their knowledge and access to information. Altogether they envisage much more modest
gains (US$1.49-2.95 trillion) over the next decade.
Figure 2 – Expected gains from AI in the different regions of the world by 2030
Source: The macroeconomic impact of artificial intelligence, PwC, 2018.
Economic impacts of artificial intelligence
5
Other sources argue that AI will have limited impact on growth, as exemplified by sectors enjoying
the highest productivity growth rates, yet witnessing a decline in their overall share in the economy.
Despite progress brought by AI, some areas of the economy would remain essential yet hard to
improve, retaining human labour that would be well remunerated. Ultimately, this would constrain
new technologies from having an impact on the overall economy. AI may even partly discourage
future innovation by accelerating imitation, which would limit the return on innovation.
AI and the future of productivity
According to a well-known productivity paradox, we are experiencing low productivity in an age of
accelerating technological progress. One possible explanation for this is that the diffusion of those
capabilities of AI that can spur productivity remains limited. Even with their broad uptake, their full effect
may only materialise with ensuing waves of complementary innovations. On the contrary, some experts
say that the ICT revolution has reached maturity and that research productivity is declining sharply,
having diminishing impacts on the economy. Taking into account the low rate of increase in physical and
human capital, which can have a stronger effect on overall productivity compared with innovation, they
foresee only a gradual evolution of productivity due to AI. According to opposing views, AI will
significantly improve human capital by offering novel ways of teaching and training the workforce.
Some consider that in reality, technological progress has a much greater impact on productivity than
shown by many estimates, as a result of mis-measurement. The OECD expects that through detection of
patterns in enormous volumes of data, AI will significantly improve decision-making, cut costs and
optimise the use of production factors and consumption of resources in every sector of the economy.
Overall, it seems likely that, while AI has significant potential to boost productivity, the final effects will
depend on the rate of AI diffusion across the economy and on investment in new technologies and
relevant skills in the workforce.
Impact on manufacturing
AI is one of the cornerstones of the growing digitalisation of industry ('Industry 4.0'). Technologies
underpinning this process – such as IoT, 5G, cloud computing, big data analytics, smart sensors,
augmented reality, 3D printing and robotics – are likely to transform manufacturing into a single
cyber-physical system in which digital technology, internet and production are merged in one. In
the smart factories of the future, production processes would be connected and
AI solutions would
be fundamental in linking the machines, interfaces, and components (using, for example, visual
recognition). Large amounts of data would be collected and fed into AI appliances, which would in
turn optimise the manufacturing process. The
OECD reckons this use of AI can be 'applied to most
industrial activities from optimising multi-machine systems to enhancing industrial research'.
Deployment of AI in production is likely to increase over time, due to the development of automated
learning processes. Fundamentally, it is likely to boost the competitiveness of the manufacturing
sector through efficiency and productivity gains enabled by data analysis, and supply chains would
be based on these gains. AI would also boost automation, ensure stronger quality control of
products and processes, and preventive diagnostics of machinery status, while also ensuring timely
maintenance, near-zero downtime, fewer errors and defective products. Manufacturers would be
able to access new markets, since their products would be more customised, varied and of higher
quality. Although the building blocks already exist, Industry 4.0 may not be realised before the
middle of the next decade, as it demands a combination of various technologies, which, according
to some, will take 20-30 years to
mainstream. The OECD forecasts that in the long-term, AI may lead
to scientific breakthroughs that could even create entirely new, unforeseen industries.
Effects on firms, industries and countries
McKinsey argues that AI and automation may on one hand facilitate the rise of massively scaled
organisations, and on the other will enable small players and even individuals to undertake project
work that is now mostly performed by bigger companies. This could spawn the emergence of very
small and very large firms, the end result being a barbell-shaped economy in which mid-sized
EPRS | European Parliamentary Research Service
6
companies lose out. Other likely effects are increased competition, firms entering new areas outside
their previous core business, and a deepening divide between technological leaders and laggards
in every sector. 'Early adopters
', that is, companies that fully absorb AI tools over the next five to
seven years, will most probably benefit disproportionately. At the other end of the spectrum would
be the slow adopters or non-adopters, which are likely to experience some economic decline. The
market share is likely to shift from the laggards to the front-runners, which would be able to
gradually attract more and more of the profit pool of their industry. This would lead to a
'winner-
takes all' phenomenon, similar to what is currently observed on tech markets. Advances in AI and
technology could enable front-runners to make a decisive break from the pack and become
'
superstars' enjoying the highest productivity levels. This can have significant consequences. For
example, the OECD has raised the question as to why apparently non-rival technologies are not
diffused to all firms. It may well be that the widening productivity gap between firms can be
attributed to the highly uneven process of technological diffusion, which favours global frontier
firms over laggards. This may occur because global frontier firms can better protect their
advantages; this could eventually even contribute to a slowdown in aggregate productivity growth
in the economy. These widening productivity and innovation gaps are surely going to attract a lively
policy debate on the unequal distribution of the benefits of AI.
In this context, it is useful to look at the industries that are moving to the forefront of AI deployment.
McKinsey
sees AI as already having a significant impact and great commercial potential in sectors
such as marketing and sales, supply chain management, logistics and manufacturing. A 2018 survey
by the
Boston Consulting Group points to the transport, logistics, automotive and technology
sectors as already being at the forefront of AI adoption. It also reveals that process industries (such
as chemicals) are lagging behind.
PwC expects that thanks to AI all sectors of the economy will see
a gain of at least 10 % by 2030. The report says that the services industry is to gain the most (21 %),
with retail and wholesale trade as well as accommodation and food services also expected to see a
large boost (15 %).
Current AI adoption levels across the world vary, making it possible that the gap between advanced
and lagging countries will widen. AI front-runners, located mostly in developed countries, are likely
to increase their lead over their counterparts in developing countries. This potential effect is likely
to be compounded by the fact that high wages in developed economies create a stronger incentive
to substitute labour with AI than in lower-wage economies. Moreover, AI may make it economical
for some manufacturers to bring back production from poorer countries.
3
AI impact on labour markets and redistributive effects of AI
If indeed technologies, such as AI, robotics and automation, are widely deployed across the
economy, there will be job creation (as a result of demand in sectors that arise or flourish due to this
deployment), as well as job destruction (replacement of humans by technology). As a 2018
meta-
study of results shows, there is no consensus among experts, with predictions ranging 'from
optimistic to devastating, differing by tens of millions of jobs even when comparing similar time
frames'.
4
A forecast by think-tank Bruegel warns that as many as 54 % of jobs in the EU face the
probability or risk of computerisation within 20 years. The effect is likely to be more nuanced, and
there seems to be a consensus among researchers that there will be significant workforce shifts
across sectors of the economy, accompanied by changes in the nature and content of
jobs, which
would require reskilling.
5
Furthermore, job polarisation is probable: lower-paid jobs that typically
require routine manual and cognitive skills stand the highest risk of being replaced by AI and
automation, while well-paid skilled jobs that typically require non-routine cognitive skills will be in
higher demand. Studying the
patterns of previous industrial revolutions indicates that job
destruction will be stronger in the short and possibly medium term, while job creation will prevail
in the longer term. Nonetheless, labour relations may alter, with more frequent job changes and a
rise in precarious work, self-employment and contract work, which would possibly weaken workers'
rights as well as the role of trade unions.
Economic impacts of artificial intelligence
7
The disruptive effects of AI may also influence wages, income distribution and economic inequality.
Rising demand for high-skilled workers capable of using AI could push their wages up, while many
others may face a wage squeeze or unemployment. This could affect even mid-skilled workers
,
whose wages may be pushed down by the fact that high-skill workers are not only more productive
than them thanks to the use of AI, but are also able to complete more tasks. The changes in demand
for labour could therefore worsen overall income distribution by affecting overall wages. Much will
depend on the pace, with faster change likely to create more undesirable effects due to
market
imperfections. Theoretically, the more AI solutions replace routine labour, the more productivity
and overall income growth will rise and the more sharply inequality will increase. This may lead to a
'paradox of plenty': society would be far richer
overall, but for many individuals, communities and
regions, technological change would only reinforce
inequalities. There are indeed fears that the current
trends of shifting the distribution of national income
away from labour, which leads to deeper inequality
and the concentration of wealth in 'superstar'
companies and sectors, will indeed only be
exacerbated by AI.
On the other hand, many economists are positive,
saying that it will be hardest for AI to replace the
'sensor-motor skills' required in non-standard and
non-routine jobs, such as that of security staff,
cleaners, gardeners and chefs. Others add that
automation always has an
ambiguous impact on
inequality: low-skill automation always increases
wage inequality, and high-skill automation always
reduces it. In conclusion, it is therefore uncertain that
at least over the short to medium term, the rise in
inequality due to AI automation will be significant.
Selected policy implications
AI has significant potential to boost economic
growth and productivity, but at the same time it
creates equally serious risks of job market
polarisation, rising inequality, structural
unemployment and emergence of new undesirable
industrial structures.
EU policy needs to create the conditions necessary for nurturing the potential of AI, while
considering carefully how to address the risks it involves. A recent economic paper shows that if
labour income
does not benefit from the economic gains generated by AI, consumption may
stagnate and restrict growth, thereby having an adverse effect on the economy. Questions about
distributing the gains from AI are therefore fundamental in managing its outcomes. Tax policies
could help to rebalance the shift from labour to capital, and shelter vulnerable groups from socio-
economic exclusion.
The European Political Strategy Centre describes the internal and external challenges the EU is
facing. The former include low investment and a slow uptake of AI technologies by companies and
the public sector, and the necessity to establish a regulatory framework that does not stifle
technological progress, while at the same time adhering to key fundamental EU principles. The latter
include fierce global competition, with other jurisdictions benefitting from structural advantages.
The centre suggests that the EU should address these by developing an investment-conducive
framework and becoming a leader in setting global AI quality standards. A precondition to
Taxing robots
Bill Gates is one of many who argue that robots that
take somebody's job should pay taxes, so as to prevent
new technologies from diminishing the public money
that supports society. In 2017, the
European
Parliament rejected the idea of imposing a robot tax on
owners to fund support for retraining of workers put
out of their jobs by robots. However, if automation
leads to significant falls in income tax receipts and
increases the pressure on government finances (e.g.
through increased welfare and retraining
expenditure), such a tax may be unavoidable in the
future. In 2018,
South Korea, the most robotised
country in the world, lowered the tax deduction on
business investments in automation, a move that
seems to acknowledge some
experts' concerns about
excessive incentivising of automation. The debate on
this topic is picking up, but if a robot tax were to be
introduced, some fundamental questions regarding a
clear and agreed
definition and the possible forms of
taxation need to be answered. One possibility is to
come up with an international solution that would
allow such a tax to be effective in the global economy.
This solution might lie along an uneasy path of
imposing taxes on the digital economy – an issue that
is hotly debated both
internationally and at EU level.
Source: OECD, 2017.
EPRS | European Parliamentary Research Service
8
successfully harness the potential of AI is to develop relevant skills in education and work as well as
funding research and pooling resources to deliver true EU added value. Importantly, the EU has the
necessary tools, such as a powerful competition policy, to address market distortions and power
asymmetries. Issues
, such as responsibility and liability, security and safety of AI-driven decision-
making, raise many questions that need to be addressed in the near future. While public authorities
are starting to focus on AI and national AI strategies are being developed, the need for a common
EU-level path becomes more urgent than ever.
MAIN REFERENCES
PricewaterhouseCoopers, The macroeconomic impacts of artificial intelligence, February 2018.
European Political Strategy Centre, The age of artificial intelligence, EPSC Strategic Notes, March 2018.
Gries T. and Naudé W., Artificial Intelligence, Jobs, Inequality and Productivity: Does Aggregate Demand
Matter?, Institute of Labor Economics, Discussion paper No 12005, November 2018.
OECD, Digital economy outlook 2017, October 2017.
McKinsey Global Institute, Notes from the AI frontier – Modeling the impact of AI on the world economy,
discussion paper, September 2018.
ENDNOTES
1
The report elaborates further: 'The European Patent route is mainly used by European applicants to seek protection in
several countries directly from first patent filing, but also by U.S. patent applicants, whereas the PCT route is used mainly
by applicants in the U.S., Japan and China (...) 15.1 percent of all the AI patent families identified in this report include a
European application.' EU countries also file for PCT patents.
2
The PwC paper groups the following states as 'northern Europe': Austria, Belgium, the Czech Republic, Denmark,
Estonia, Finland, France, Germany, Ireland, Latvia, Lithuania, Luxembourg, the Netherlands, Poland, Sweden, the United
Kingdom, Switzerland and Norway. 'Southern Europe' includes Cyprus, Greece, Hungary, Italy, Malta, Portugal, Slovakia,
Slovenia, Spain, Bulgaria, Croatia, Romania, Albania, Belarus, Ukraine, the rest of the EFTA countries and the rest of
eastern Europe.
3
McKinsey estimates that leading AI countries could capture an additional 20-25 % in net economic benefits compared
with today, while developing countries could capture only about 5-15 %. China is an important exception.
4
There are numerous factors at play that render the making of forecasts of the final effect a challenging task. For
example, AI diffusion may be slow, which will limit its impact on employment. On the other hand, AI can result in
product innovations that foster growth in demand, thereby creating new jobs.
5
In 31 OECD countries, 14 % of jobs are at high risk of automation, while a further 32 % will change significantly.
DISCLAIMER AND COPYR
IGHT
This document is prepared for, and addressed to, the Members and staff of the European Parliament as
background material to assist them in their parliamentary work. The content of the document is the sole
responsibility of its author(s) and any opinions expressed herein should not be taken to represent an official
position of the Parliament.
Reproduction and translation for non
-
commercial purposes are authorised, provided the source is
acknowledged and the European Parliament is given prior notice and sent a copy.
© European Union,
2019.
Photo credits: ©
Alexander Limbach / Fotolia.
[email protected]ropa.eu (contact)
www.eprs.ep.parl.union.eu
(intranet)
www.europarl.europa.eu/thinktank
(internet)
http://epthinktank.eu
(blog)