Ai - Artificial Intelligence predictions-2018-report - PWC

1
2018 AI predictions
pwc.com/us/AI2018
2018 AI
predictions
8 insights to shape
business strategy
pwc.com/us/AI2018

2
2018 AI predictions
pwc.com/us/AI2018
Here’s some actionable advice on artificial intelligence (AI), that you can
use today: If someone says they know exactly what AI will look like and
do in 10 years, smile politely, then change the subject or walk away.
AI is remarkably complex and advancing quickly. It’s doing far more in
some areas, and far less in others, than anyone would have guessed a
decade ago. It’s impossible for anyone today to give a precise vision of
how the next ten—much less five—years will unfold. That’s not to say
that it’s impossible to make broad predictions about AI’s impact in the
coming years and decades. We’ve done that elsewhere.
Our aim here is different: to make specific predictions about AI trends
for the next 12 months, then draw out key implications for business,
government, and society as a whole. We’re confident in making near-
team forecasts because these nascent trends are already underway,
though they aren’t yet attracting the attention they deserve.
We’ve made eight such predictions. They’re based not just on insights
from AI visionaries and computer scientists. They’re also informed by
what our leaders in assurance, consulting, and tax see on the ground
with clients around the world who are grappling with how to put AI to
work in their organizations and prepare their employees for a world in
which AI is everywhere.
We hope you’ll consider how these predictions relate to your own
organization.
1. AI will impact employers before it impacts employment
2. AI will come down to earth—and get to work
3. AI will help answer the big question about data
4. Functional specialists, not techies, will decide the AI talent race
5. Cyberattacks will be more powerful because of AI—but so
will cyberdefense
6. Opening AI’s black box will become a priority
7. Nations will spar over AI
8. Pressure for responsible AI won’t be on tech companies alone
PwC AI
predictions
for 2018
2018 AI predictions
pwc.com/us/AI2018 2

3
2018 AI predictions
pwc.com/us/AI2018
AI will impact employers before it
impacts employment
Everyone has seen the headlines: Robots and AI will destroy jobs. But we
don’t see it that way. We see a more complex picture coming into focus,
with AI encouraging a gradual evolution in the job market that—with the
right preparation—will be positive. New jobs will offset those lost. People
will still work, but they’ll work more efficiently with the help of AI.
Most people have heard that AI beat the world’s greatest grandmaster
in chess. But not everyone knows what can usually beat an AI chess
master: a “centaur,” or human and AI playing chess as a team. The human
receives advice from an AI partner but is also free to override it, and it’s the
established process between the two that is the real key to success.
This unparalleled combination will become the new normal in the
workforce of the future. Consider how AI is enhancing the product design
process: A human engineer defines a part’s materials, desired features,
and various constraints, and inputs it into an AI system, which generates a
number of simulations. Engineers then either choose one of the options, or
refine their inputs and ask the AI to try again.
This paradigm is one reason why AI will strengthen the economy. At
the same time, however, there’s no denying that in some industries,
economies, and roles—especially those that involve repetitive tasks—jobs
will change or be eliminated. Yet in the next two years, the impact will be
relatively modest: PwC’s forthcoming international jobs automation study,
due in February 2018, estimates that across 29 countries analyzed, the
share of jobs at potential high risk of automation is only 1 percent by 2020.
Why organizations will succeed or fail
The upshot? In 2018, organizations will start realizing they need to change
how they work. As they do so, they’ll need to be especially mindful of what
has come before: failed tech transformations. There are several reasons
why this happens, but two in particular are relevant to the way so many
organizations are approaching AI. They’re pigeon-holing AI talent. And
they’re thinking and working in silos.
1

4
2018 AI predictions
pwc.com/us/AI2018
AI-savvy employees won’t just need to know how to choose the right
algorithm and feed data into an AI model. They’ll also have to know how to
interpret the results. They’ll need to know when to let the algorithm decide,
and when to step in themselves.
At the same time, effective use of AI will demand collaboration among
different teams. Consider an AI system that helps hospital staff decide
which medical procedures to authorize. It will need input not just from
medical and AI specialists, but also from legal, HR, financial, cybersecurity,
and compliance teams.
Most organizations like to set boundaries by putting specific teams in
charge of certain domains or projects and assigning a budget accordingly.
But AI requires multidisciplinary teams to come together to solve a
problem. Afterward, team members then move on to other challenges but
continue to monitor and perfect the first.
With AI, as with many other digital technologies, organizations and
educational institutions will have to think less about job titles, and
more about tasks, skills, and mindset. That means embracing new ways
of working.
of executives say AI will help humans and
machines work together to be stronger using
both artificial and human intelligence
67%
Source: PwC Consumer Intelligence Series: Bot.Me, 2017
Base: 500 business executives
#1. AI will impact employers before it impacts employment

5
2018 AI predictions
pwc.com/us/AI2018
Implications
Popular acceptance of AI may occur quickly
As signs grow this year that the great AI jobs disruption will be a false
alarm, people are likely to more readily accept AI in the workplace and
society. We may hear less about robots taking our jobs, and more about
robots making our jobs (and lives) easier. That in turn may lead to a faster
uptake of AI than some organizations are expecting.
The organizational retooling will commence
It will be a lengthy process, but some forward-thinking organizations
are already breaking down the silos that separate data into cartels
and employees into isolated units. Some will also start on the massive
workforce upskilling that AI and other digital technologies require. This
upskilling won’t just teach new skills. It will teach a new mindset that
emphasizes collaboration with co-workers—and with AI.
Would work with
AI manager if
meant more
balanced workload
Would free
employees from
menial tasks
Would offer
employees new
work
opportunities
Would follow AI
system if predicted
most efficient way to
manage project
50%64%78% 65%
Base: 500 business executives; percent agreeing with statement
How workers think about human-machine AI centaurs
Base: 500 business executives; percent agreeing with statement
#1. AI will impact employers before it impacts employment

6
2018 AI predictions
pwc.com/us/AI2018
AI will come down to earth—and
get to work
There are plenty of publications promising an AI-powered future that will
look like magic: fleets of autonomous cars that never crash or encounter
traffic jams, robot doctors that diagnose illness in milliseconds, and smart
infrastructure that optimizes flows of people and goods and maintains itself
before repairs are ever needed. All that may come—but not in 2018.
Executives think that AI will be crucial for their success: 72% believe it will
be the business advantage of the future. The question is: What can it do for
me today? And the answer is here.
Augmenting human productivity
If AI sounds far-fetched, what about a tool to perform repetitive white-collar
tasks, so managers can spend their time on analysis? How about one that
detects fraud and increases supply chain resilience?
This is the value of AI in 2018: it lies not in creating entire new industries
(that’s for the next decade), but rather in empowering current employees
to add more value to existing enterprises. That empowerment is coming in
three main ways:
• Automating processes too complex for older technologies
• Identifying trends in historical data to create business value
• Providing forward-looking intelligence to strengthen human decisions
Value from tedious tasks
Consider how most companies’ finance functions spend a large portion
of their time: wading through data coming from ERP, payment processing,
business intelligence, and other systems. Many staff members also spend
hours each day poring through legal contracts and emails, or performing
mundane transactional tasks.
The result is that value-adding analysis is what many finance professionals
only do when they have time left over from their other, routine tasks.
2

7
2018 AI predictions
pwc.com/us/AI2018
Ranking Industry High-potential use cases
Healthcare
• Supporting diagnosis by detecting variations in patient data
• Early identification of potential pandemics
• Imaging diagnostics
Automotive
• Autonomous fleets for ride sharing
• Semi-autonomous features such as driver assist
• Engine monitoring and predictive, autonomous maintenance
Financial services
• Personalized financial planning
• Fraud detection and anti-money laundering
• Automation of customer operations
Transportation
and logistics
• Autonomous trucking and delivery
• Traffic control and reduced congestion
• Enhanced security
Technology,
media, and
telecommunications
• Media archiving, search, and recommendations
• Customized content creation
• Personalized marketing and advertising
Retail and consumer
• Personalized design and production
• Anticipating customer demand
• Inventory and delivery management
Energy
• Smart metering
• More efficient grid operation and storage
• Predictive infrastructure maintenance
Manufacturing
• Enhanced monitoring and auto-correction of processes
• Supply chain and production optimization
• On-demand production
Now imagine an AI system scanning all the function’s data, identifying
trends and anomalies, performing many transactions automatically, and
flagging relevant issues for further attention. Imagine AI also identifying
and explaining likely risks and offering data-driven forecasts to support
managers’ analysis and decisions.
It may not be as sexy as a smart city, but this kind of practical AI is ready
right now. And it’s often “sneaking in through the backdoor.” Enterprise
application suites from Salesforce, SAP, Workday, and others are
increasingly incorporating AI.
Source: PwC Global AI Impact Index, 2017
3
1
1
4
5
6
7
8
Where industries will put practical AI to work
Ranking of AI impact by its potential to free up time, enhance quality, and enhance personalization
#2. AI will come down to earth—and get to work

8
2018 AI predictions
pwc.com/us/AI2018
Implications
Business problems will open the door to AI
Leaders don’t need to adopt AI for AI’s sake. Instead, when they look
for the best solution to a business need, AI will increasingly play a role.
Does the organization want to automate billing, general accounting and
budgeting, and many compliance functions? How about automating parts
of procurement, logistics, and customer care? AI will likely be a part of
the solution, whether or not users even perceive it.
New kinds of ROI measures are needed
Sometimes the best approach to gauge AI’s value is to use the same
measures you’d apply to any other business investment: metrics such
as increased revenue or reduced costs. But AI’s most powerful benefits
are often indirect, so organizations will want to explore other measures
of ROI. Automated full-time equivalents can capture how AI is freeing
human workers from mundane tasks. Other metrics can show how AI is
improving human decision-making and forecasts.
of business executives say AI solutions
implemented in their businesses have
already increased productivity
54%
#2. AI will come down to earth—and get to work

9
2018 AI predictions
pwc.com/us/AI2018
AI will help answer the big question
about data
Many companies haven’t seen the payoff from their big data
investments. There was a disconnect. Business and tech executives
thought they could do a lot more with their data, but the learning
curve was steep, tools were immature, and they faced considerable
organizational challenges.
Now, some are rethinking their data strategy as the landscape matures
and AI itself becomes more real and practical. They’re starting to ask the
right questions, like: How can we make our processes more efficient? and
What do we need to do to automate data extraction?
Public sector
Power & utilities
Automotive
Industrial products
Energy & mining
Tech, media, & communications
Healthcare
Financial services
Consumer markets
Hospitality & leisure
43%
37%
33%
33%
30%
27%
26%
24%
13%
30%
Source: PwC 2017 Global Digital IQ Survey
Q: To what extent do you agree with the following statement (strongly agree): We effectively
utilize all the data we capture to drive business value
Bases: 72, 217, 135, 322, 237, 75, 375, 131, 156, 433
Few businesses get value from their data—but AI
could change that
3
Source: PwC 2017 Global Digital IQ Survey
Q: To what extent do you agree with the following statement (strongly agree): We effectively
utilize all the data we capture to drive business value
Bases: 72, 217, 135, 322, 237, 75, 375, 131, 156, 433

10
2018 AI predictions
pwc.com/us/AI2018
At the same time, organizations are now able to take advantage of new
tools and technical advancements, including:
• Easier methods for mining less-structured data, including natural
language processing for text indexing and classification
• Enterprise application suites that incorporate more AI
• Emerging data lake-as-a-service platforms
• Public clouds that can take advantage of different kinds of data
• Automated machine learning and data management
Feeding the AI beast
Despite these advances, many organizations still face a challenge. Many
kinds of AI, such as supervised machine learning and deep learning,
need an enormous amount of data that is standardized, labeled, and
“cleansed” of bias and anomalies. Otherwise, following the ancient rule—
garbage in, garbage out—incomplete or biased data sets will lead to
flawed results. The data must also be specific enough to be useful, yet
protect individuals’ privacy.
Consider a typical bank. Its various divisions (such as retail, credit card,
and brokerage) have their own sets of client data. In each division,
the different departments (such as marketing, account creation, and
customer service) also have their own data in their own formats. An AI
system could, for example, identify the bank’s most profitable clients and
offer suggestions on how to find and win more clients like them. But to do
that, the system needs access to the various divisions’ and departments’
data in standardized, bias-free form.
The right approach to data
It’s rarely a good idea to start with a decision to clean up data. It’s almost
always better to start with a business case and then evaluate options for
how to achieve success in that specific case.
A healthcare provider, for example, might aim to improve patient
outcomes. Before beginning to develop the system, the provider would
quantify the benefits that AI can bring. The provider would next look at
what data was needed—electronic medical records, relevant journal
#3. AI will help answer the big question about data

11
2018 AI predictions
pwc.com/us/AI2018
articles, and clinical trials data, among others—and the costs of acquiring
and cleansing this data.
Only if the benefits—including measures of indirect benefits and how
future applications can use this data—exceed the costs should this
provider move forward.
That’s how many organizations will ultimately reform data architecture and
governance: with AI and other technologies offering value propositions
that require it.
Implications
Success will lead to success
Those enterprises that have already addressed data governance for one
application will have a head start on the next initiative. They’ll be on their
way to developing best practices for effectively leveraging their data
resources and working across organizational boundaries.
Third-party data providers will prosper
There’s no substitute for organizations getting their internal data ready to
support AI and other innovations, but there is a supplement: Vendors are
increasingly taking public sources of data, organizing it into data lakes,
and preparing it for AI to use.
More synthetics are coming
As data becomes more valuable, advances in synthetic data and other
“lean” and “augmented” data learning techniques will accelerate. We
may not need, for example, a whole fleet of autonomous cars on the road
to generate data about how they’ll act. A few cars, plus sophisticated
mathematics, will be sufficient.
of executives say big data at their company
would be improved through the use of AI59%
2018 AI predictions
pwc.com/us/AI2018
11
#3. AI will help answer the big question about data

12
2018 AI predictions
pwc.com/us/AI2018
Functional specialists, not techies,
will decide the AI talent race
As AI spreads into more specific areas, it will require knowledge and skill
sets that data scientists and AI specialists usually lack.
Consider a team of computer scientists creating an AI application to
support asset management decisions. The AI specialists probably aren’t
experts on the markets. They’ll need economists, analysts, and traders
working at their side to identify where the AI can best support the human
asset manager, help design and train the AI to provide that support, and
be willing and able to use the AI effectively.
And since the financial world is in constant flux, once the AI is up and
running, it will need continual customizing and tweaking. For that too,
functional specialists—not programmers—will have to lead the way.
The same is true not just throughout financial services, but in healthcare,
retail, manufacturing, and every sector that AI touches.
Citizen data scientists wanted
AI is becoming more user friendly. Users no longer need to know how
to write code in order to work with some AI applications. But most still
demand far more technical knowledge than a spreadsheet or word
processing program does.
For example, many AI tools require users to formulate their needs into
machine learning problem sets. They also require an understanding of
which algorithms will work best for a particular problem and a particular
data set.
The exact level of knowledge required will vary, but we can broadly divide
AI’s demands on human knowledge into three categories. First, most
members of an AI-enabled enterprise will need some basic knowledge
of AI’s value as well as what it can and can’t do with data. Second, even
the most mature AI program will always need a small team of computer
scientists. The third group, which many organizations aren’t yet paying
attention to, are AI-savvy functional specialists.
4

13
2018 AI predictions
pwc.com/us/AI2018
They won’t have to be programmers. They will have to understand the
basics of data science and data visualization and the basics of of how AI
“thinks.” They’ll have to be citizen data scientists.
Retail analysts, engineers, accountants, and many other domain experts
who know how to prepare and contextualize data so AI can make optimal
use of it will be crucial to enterprise success. As AI leaves the computer
lab and enters everyday work processes, these functional specialists will
be even more important than computer scientists.
Implications
Faster upskilling means faster AI deployment
Enterprises that intend to take full advantage of AI shouldn’t just bid
for the most brilliant computer scientists. If they want to get AI up and
running quickly, they should move to provide functional specialists with AI
literacy. Larger organizations should prioritize by determining where AI is
likely to disrupt operations first and start upskilling there.
Upskilling will lead to new approaches to learning
Organizations will have to upskill many of their employees to learn the
basics of data science and how to think like an AI application. Given the
enormity of this task, companies must find ways to assess the skills of
high-potential learners and put them on individual learning paths, to get
them up to speed quickly.
of jobs requiring data science and analytics
skills are in fields other than AI67% Source: PwC and Business Higher Education Forum, Investing in America’s data
science and analytics talent, 2017
#4. Functional specialists, not techies, will decide the AI talent race

14
2018 AI predictions
pwc.com/us/AI2018
Cyberattacks will be more
powerful because of AI—but so
will cyberdefense
What’s one job where AI has already shown superiority over human
beings? Hacking. Machine learning, for example, can easily enable a
malicious actor to follow your behavior on social media, then customize
phishing tweets or emails—just for you. A human hacker can’t do the job
nearly as well or as quickly.
The more AI advances, the more its potential for cyberattacks grows too.
Techniques like advanced machine learning, deep learning, and neural
networks enable computers to find and interpret patterns. They can also
find and exploit vulnerabilities.
Intelligent malware and ransomware that learns as it spreads, machine
intelligence coordinating global cyberattacks, advanced data analytics to
customize attacks—unfortunately, it’s all on its way to your organization
soon. And AI itself, if not well-protected, gives rise to new vulnerabilities.
Malicious actors could, for example, inject biased data into algorithms’
training sets.
AI to the rescue
Just as we expect AI to be a growing cyberthreat this year, we’re also
confident it will be part of the solution. Already, scalable machine learning
techniques combined with cloud technology are analyzing enormous
amount of data and powering real-time threat detection and analysis. AI
capabilities can also quickly identify “hot spots” where cyberattacks are
surging and provide cybersecurity intelligence reports.
The winner of the US Defense Department’s DARPA Cyber Grand
Challenge, a cybersecurity competition, used AI deep learning—and the
Pentagon has purchased the technology.
Yet even in cybersecurity, some things only people can do. Humans are
better at absorbing context and thinking imaginatively. Cyberwars won’t
simply be two sets of computers battling it out. But AI will become an
important part of every major organization’s cybersecurity toolkit.
5

15
2018 AI predictions
pwc.com/us/AI2018
Implications
Don’t bring a knife to a gunfight
In other parts of the enterprise, many organizations may choose to go slow
on AI, but in cybersecurity there’s no holding back: Attackers will use AI,
so defenders will have to use it too. If an organization’s IT department or
cybersecurity provider isn’t already using AI, it has to start thinking immediately
about AI’s short- and long-term security applications. Sample use cases
include distributed denial of service (DDOS) pattern recognition, prioritization of
log alerts for escalation and investigation, and risk-based authentication.
Cybersecurity may speed up AI’s acceptance
Since even AI-wary organizations will have to use AI for cybersecurity,
cyberdefense will be many enterprises’ first experience with AI. We see this
fostering familiarity with AI and willingness to use it elsewhere. A further
spur to AI acceptance will come from its hunger for data: The greater AI’s
presence and access to data throughout an organization, the better it can
defend against cyberthreats. Some organizations are already building out
on-premise and cloud-based “threat lakes,” that will enable AI capabilities.
An AI hack may increase public fears
Many people are already nervous about AI. Even more are concerned
about cybersecurity. It’s possible that when AI makes headlines this year, it
won’t be for helping humanity—but for having enabled a major hack. Better
cybersecurity can mitigate this risk. Besides developing feature sets for AI
capabilities, this heightened security will require companies to augment the
data and compute platforms that support advanced analytics with privileged
access monitoring, object-level change management, source code reviews,
and expanded cybersecurity controls, among other precautions.
of executives say their organization plans to
invest this year in cybersecurity safeguards
that use AI and machine learning
27%
Source: PwC 2018 Global State of Information Security® Survey
Base: 9.500 business and technology executives
#5. Cyberattacks will be more powerful because of AI—but so will cyberdefense

16
2018 AI predictions
pwc.com/us/AI2018
Opening AI’s black box will become
a priority
Might AI-powered autonomous weapons become serial killers? Could an AI
system told to reduce air pollution decide that the most logical way to do
so is to eliminate the human race? Such fears may make for good thrillers,
but the danger is manageable.
Here’s the secret about AI that many of its proponents don’t like to
mention: It’s not that smart—at least not yet. AI is getting better at pattern
and image recognition, automating complex tasks, and helping humans
make decisions. All that offers opportunities for enterprises that could be
worth trillions of dollars.
In the past, for example, to teach an AI program chess or another game,
scientists had to feed it data from as many past games as they could find.
Now they simply provide the AI with the game’s rules. In a few hours it
figures out on its own how to beat the world’s greatest grandmasters.
That’s extraordinary progress, with immense potential to support human
decision making. Instead of playing chess, an AI program with the right
rules can “play” at corporate strategy, consumer retention, or designing a
new product.
But it’s still just following rules that humans have devised. With appropriate
attention paid to responsible AI, we can safely harness its power.
A real risk
If AI should always be controllable, it’s not always understandable. Many
AI algorithms are beyond human comprehension. And some AI vendors
will not reveal how their programs work to protect intellectual property. In
both cases, when AI produces a decision, its end users won’t know how it
arrived there. Its functioning is a “black box.” We can’t see inside it.
That’s not always a problem. If an ecommerce website uses mysterious
algorithms to suggest a new shirt to a consumer, the risks involved are low.
6

17
2018 AI predictions
pwc.com/us/AI2018
But what happens when AI-powered software turns down a mortgage
application for reasons that the bank can’t explain? What if AI flags
a certain category of individual at airport security with no apparent
justification? How about when an AI trader, for mysterious reasons,
makes a leveraged bet on the stock market?
Users may not trust AI if they can’t understand how it works. Leaders may
not invest in AI if they can’t see evidence of how it made its decisions. So
AI running on black boxes may meet a wave of distrust that limits its use.
Understanding of
AI model decision
making
Understanding reasoning
behind each decision
Provability
Mathematical
certainty behind
decisions
Explainability
Transparency
Source: PwC
What it means to look inside AI’s black box
Source: PwC
#6. Opening AI’s black box will become a priority

18
2018 AI predictions
pwc.com/us/AI2018
Implications
Many black boxes will open
We expect organizations to face growing pressure from end users and
regulators to deploy AI that is explainable, transparent, and provable.
That may require vendors to share some secrets. It may also require users
of deep learning and other advanced AI to deploy new techniques that
can explain previously incomprehensible AI.
Organizations face tradeoffs
Most AI can be made explainable—but at a cost. As with any other
process, if every step must be documented and explained, the process
becomes slower and may be more expensive. But opening black boxes
will reduce certain risks and help establish stakeholder trust.
Enterprises need a framework for AI explainability decisions
Explainability, transparency, and provability aren’t absolutes; they exist on
a scale. A framework to assess business, performance, regulatory, and
reputational concerns can enable optimal decisions about where each AI
use case should fall on that scale. A healthcare firm using AI to help make
life-or-death decisions has different needs than a private equity fund
using AI to identify potential targets for further research.
#6. Opening AI’s black box will become a priority

19
2018 AI predictions
pwc.com/us/AI2018
Nations will spar over AI
AI is going to be big: $15.7 trillion big by 2030, according to our research.
The AI pie is so big that besides individual companies, countries are
working on strategies to claim the biggest possible slice.
The US started off strong, with a trio of reports in 2016. They outlined
a plan to make the US an AI powerhouse and thereby boost both the
economy and national security.
Recommendations included increased federal funding, regulatory
changes, the creation of shared public data sets and environments, the
definition of standards and benchmarks, workforce development, and
ways for AI to bolster cybersecurity and the military.
But since a new administration entered at the start of 2017, the
government has abandoned this plan. It’s cutting AI research funds.
Its recently passed tax reform, however, could boost AI in the US. The
lower corporate tax rate, provisions for repatriating cash from overseas,
and permission to expense 100 percent of capital investments is
likely to spur investment in AI and other technologies. And the current
administration’s emphasis on deregulation could help AI in certain
sectors, such as drones and autonomous vehicles.
The new AI leaders
While the US government moves on from its AI development plan, other
countries are taking action.
The UK last year launched a plan to improve access to data, AI skills, and
AI research and uptake. Its latest budget added funding for a Centre for
Data Ethics and Innovation to drive responsible AI, for exploratory work
on data trusts, and for new AI fellowships and researchers.
Canada—already an AI leader—is also working to make AI a lynchpin of
its future economy. The federal government last year launched its Pan-
Canadian Artificial Intelligence Strategy. The program includes funding
for AI research centers in collaboration with private companies and
universities. It also aims to attract and retain top AI talent.
7

20
2018 AI predictions
pwc.com/us/AI2018
Japan meanwhile released an AI technology strategy with a three-
phase plan to achieve a true AI ecosystem. Building on successes in
robotics, Japan’s government envisions joining AI with other advanced
technologies, such as the internet of things, autonomous vehicles, and
the blending of cyber and physical space.
Other countries with recently released national AI plans include Germany
with ethical guidelines for automated driving and its Industrie 4.0 initiative,
and the UAE with a strategy to use AI to boost government performance
and various economic sectors.
China stands apart
China’s next-generation AI plan, released in 2017, declared AI as
a strategic national priority for the country and showcased the top
leadership’s vision for a new economic model driven by AI.
Unlike the US, the Chinese government is putting this plan into practice.
For example, it recently commissioned Baidu to create a national “deep
learning laboratory” together with leading universities—and it’s investing
an undisclosed sum in the effort.
The country is already strong in AI. Baidu, Alibaba, and Tencent are
among the global AI leaders. Chinese programmers recently won the
ImageNet AI competition. And its leading ecommerce companies
are using highly sophisticated AI in their warehouses and across
the business.
Other countries also have innovative engineers, universities, and
companies. But China stands apart in the extent to which its government
is prioritizing AI. Our research indicates that China will reap the most
benefit from AI over the next decade: some $7 trillion in GDP gains by
2030, thanks to an uptick in productivity and consumption.
#7. Nations will spar over AI

21
2018 AI predictions
pwc.com/us/AI2018
Implications
China’s investment may awaken the West
If China starts to produce leading AI developments, the West may
respond. Whether it’s a “Sputnik moment” or a more gradual realization
that they’re losing their lead, policymakers may feel pressure to change
regulations and provide funding for AI.
More national and regional strategies will come
More countries should issue AI strategies, with implications for
companies. It wouldn’t surprise us to see Europe, which is already
moving to protect individuals’ data through its General Data Protection
Regulation (GDPR), issue policies to foster AI in the region.
Collaboration may come too
National competition for AI will never cease—there’s too much money at
stake. But we do expect growing opportunities, facilitated by the UN, the
World Economic Forum, and other multilateral organizations, for countries
to collaborate on AI research in areas of international concern.
26.1% China
14.5% North America
10.4% Developed Asia
11.5% Southern Europe
9.9% Northern Europe
5.6% Africa, Oceania, & other Asian markets
5.4% Latin America
$15.7
trillion
potential GDP gain
Source: PwC Global Artificial Intelligence Study, 2017
Where AI gains will be realized
AI's impact on GDP by 2030
Source: PwC Global Artificial Intelligence Study, 2017
#7. Nations will spar over AI

22
2018 AI predictions
pwc.com/us/AI2018
Pressure for responsible AI won’t be on
tech companies alone
New technologies often bring new fears, justified or not, and not just among
conspiracy theorists. Seventy-seven percent of CEOs in a 2017 PwC survey
said AI and automation will increase vulnerability and disruption to the way
they do business. Odds are good that if we asked government officials, the
response would be similar.
Leaders will soon have to answer tough questions about AI. It may be
community groups and voters worried about bias. It may be clients fearful
about reliability. Or it may be boards of directors concerned about risk
management, ROI, and the brand.
In all cases, stakeholders will want to know that organizations are using AI
responsibly, so that it strengthens the business and society as a whole.
The result, we believe, will be pressure to adopt principles for responsible AI.
77%
76%
73%
71%
67%
64%
Increased vulnerability and disruption to business
Potential for biases and lack of transparency
Ensuring governance and rules to control AI
Risk to stakeholders’ trust and moral dilemmas
Potential to disrupt society
Lack of adequate regulation
Source: PwC CEO Pulse Survey, 2017
Q: Which of the following issues surrounding AI adoption concern you the most?
Base: 239
What’s holding AI back in the enterprise?
8
Source: PwC CEO Pulse Survey, 2017
Q: Which of the following issues surrounding AI adoption concern you the most?
Base: 239

23
2018 AI predictions
pwc.com/us/AI2018
A global movement begins
We’re not alone in this belief. The World Economic Forum’s Center for
the Fourth Industrial Revolution, the IEEE, AI Now, The Partnership on AI,
Future of Life, AI for Good, and DeepMind, among other groups, have all
released sets of principles that look at the big picture: how to maximize
AI’s benefits for humanity and limit its risks.
Some areas of relative consensus (which we fully support) among these
institutions include:
• designing AI with an eye to societal impact
• testing AI extensively before release
• using AI transparently
• monitoring AI rigorously after release
• fostering workforce training and retraining
• protecting data privacy
• defining standards for the provenance, use, and securing of data sets
• establishing tools and standards for auditing algorithms
With any new technology (and many old ones too), the golden rule we
follow is to do more than compliance requirements demand. Regulators
and laws often lag innovation. Organizations that don’t wait for
policymakers to issue orders, but instead voluntarily use new technology
responsibly, will reduce risks, improve ROI, and strengthen their brands.
Implications
New structures for responsible AI
As organizations face pressure to design, build, and deploy AI systems
that deserve trust and inspire it, many will establish teams and processes
to look for bias in data and models and closely monitor ways malicious
actors could “trick” algorithms. Governance boards for AI may also be
appropriate for many enterprises.
#8. Pressure for responsible AI won’t be on tech companies alone

24
2018 AI predictions
pwc.com/us/AI2018
Public-private partnerships and public-citizen partnerships
One of the best ways to use AI responsibly is for public and private
sector institutions to collaborate, especially when it comes to AI’s societal
impact. Likewise, as more governments explore the use of AI to distribute
services efficiently, they’re engaging citizens in the process. In the UK,
for example, the RSA (Royal Society for the encouragement of Arts,
Manufactures and Commerce) is conducting a series of citizen juries on
the use of AI and ethics in criminal justice and democratic debate.
Self-regulatory organizations to facilitate responsible
innovation
Since regulators may scramble to keep up, and self-regulation has
its limits, self-regulatory organizations (SROs) may take the lead with
responsible AI. An SRO would bring users of AI together around certain
principles, then oversee and regulate compliance, levy fines as needed,
and refer violations to regulators. It’s a model that has worked in other
industries. It may well do the same for AI and other technologies.
#8. Pressure for responsible AI won’t be on tech companies alone

© 2018 PwC. All rights reserved. PwC refers to the US member firm or one of its subsidiaries or affiliates,
and may sometimes refer to the PwC network. Each member firm is a separate legal entity.
Please see www.pwc.com/structure for further details.
This content is for general information purposes only, and should not be used as a substitute for
consultation with professional advisors. 407855-2018
pwc.com/us/AI2018
Authors
Mike Baccala
US Assurance Innovation
Leader
Chris Curran
US New Ventures Chief
Technology Officer
Dan Garrett
US Technology Consulting
Leader
Scott Likens
New Services and
Emerging Technologies
Leader, US, China and
Japan
Anand Rao
Global and US Artificial
Intelligence Leader
Andy Ruggles
US Tax Reporting and
Strategy Leader
Michael Shehab
US Tax Technology and
Process Leader
Editor
Chrisie Wendin
US Editorial Director,
Technology
Marketer
Sarah Weiss
US Tech and Emerging
Technology Marketing
Contributors
Cristina Ampil
US Integrated Content,
Managing Director
Maria Axente
UK AI Programme Driver
Bjarne Berg
US Tax Analytics and
Innovation Leader
Richard Berriman
UK Strategy & Data Science
Innovation Team
Euan Cameron
UK AI Leader
Chris Castelli
US Editorial Director,
Risk and Regulatory
Mike Flynn
US Assurance Principal
Ilana Golbin
US AI Accelerator Manager
John Hawksworth
UK Chief Economist
Mir Kashifuddin
US Cybersecurity and Privacy
Principal
Art Kleiner
Editor in Chief, PwC Global
Matt Lieberman
US Advisory Marketing Leader
Rob McCargow
UK AI Programme Leader
Alan Morrison
US Integrated Content,
Technology
Diego Muro
US Tax Transfer Pricing
Principal
Horacio Pena
US Tax Transfer Pricing Leader
Tom Puthiyamadam
Global Digital Services Leader
Pia Ramchandani
US AI Accelerator Director
David Resseguie
US Emerging Tech Tax Leader,
Tax Technology
David Sapin
US Advisory Risk and
Regulatory Leader
Jas Sidhu
UK Disruptive Innovation Lead
John Sviokla
US Marketing Leader
Gerard Verweij
Global Data & Analytics
Leader

Ai - Artificial Intelligence predictions-2018-report - PWC

More Related Content

What's hot

Similar to Ai - Artificial Intelligence predictions-2018-report - PWC

More from Rick Bouter

Recently uploaded

Ai - Artificial Intelligence predictions-2018-report - PWC