Morningstar Magazine - April/May 2018 - 50

Strategies

EXHIBIT 4

surprising given the acceleration of demand for
AI/ML deals in recent years, although valuations at
minimum doubling over 10 years isn't tepid growth.
Exits

After years of negligible exit activity, the past two
years represent a shift to a new period in AI/ML
exits ( E X H I B I T 3 ). The dearth of AI/ML exits until
recently is part of the typical VC cycle. VC
investment usually flows into a vertical as it begins
to grow, and it subsequently takes years for
businesses to mature and for strategic acquirers
to enter. This is key because AI/ML exits are
occurring almost exclusively via strategic
acquisitions. We expect this trend to continue as
tech giants work to bolster their in-house AI
offerings ( E X H I B I T 4 ). Additional acquisitions will
be fueled as corporations operating outside
the software/Internet sphere realize the potential
of integrating AI into their businesses.
The largest recorded exits in both 2016 and 2017
came in the autonomous driving space.
GM GM purchased Cruise Automation for $1 billion,
and Aptiv APTV acquired NuTonomy. The Cruise
acquisition represented nearly 40% of recorded AI/
ML exit value for 2016, a reminder of the effect of
outliers within the AI/ML market.

Most Active Acquirers Leading
technology companies are making
strategic acquisitions to bolster their
AI offerings.
Firm

Acquisitions

Alphabet GOOGL

5

Microsoft MSFT

4

Intel INTC

3

Apple AAPL

3

Yahoo

3

Amazon.com AMZN

2

Facebook FB

2

Samsung

2

Airbnb

2

HubSpot HUBS

2

IBM IBM

2

LexisNexus

2

Nuance NUAN

2

Splunk SPLK

2

Xilinx XLNX

2

Source: PitchBook. Data from 12/31/2006 to 12/31/2017.

Limitations

Within the vertical, the dominant sentiment waxes
positive; however, there are limitations that could
slow the spread of AI/ML across all industries.
AI/ML is often accompanied with grandiose
claims on the effects it will have on people's lives,
with the implicit assumption that failure
is impossible. This mindset can be dangerous.
Implementing AI/ML techniques in enterprise
requires realistic ramp-up timelines and
effective management of failures. Otherwise,
algorithmic mistakes-such as a software bug in
an autonomous car-can damage public
perception of the technology and cause an array
of other problems, including delays in execution.

Another common issue with AI/ML occurs when
humans cannot explain a machine's actions-a
situation known as the black box problem-
which obscures auditing and the culpability for the
actions of the machine.

50

Morningstar April/May 2018

AI/ML should ease the burden of both menial
and difficult tasks, but societal impacts
may accompany broad adoption. One of the most
discussed is the idea that AI is going to take
people's jobs. While this may end up being true for
some occupations, advancements will likely
change the nature of jobs, as opposed to
eliminating them entirely. Much like no one had
the job title of social media manager or
YouTuber before the proliferation of internet
applications, the advent of AI/ML will spawn new
industries and occupations.
Key Opportunity: Healthcare

While AI/ML has the potential to affect almost
every industry, we see a few as ripe for extensive
transformation or disruption. One prime
opportunity for AI/ML applications is healthcare,
which can provide a great deal of value and,
more importantly, save lives. These technologies
will be employed in three basic areas.

First, the automation of menial tasks for doctors
and other healthcare workers-especially
the tedium of manually inputting patient data into
electronic health records. New drug discovery
is also an exciting use of AI/ML technology-from
treating previously incurable diseases to testing
the potential outcomes of gene manipulation
in CRISPR treatments. Finally, the ability to provide
personalized healthcare for each patient
will improve patient experience as well as solve
myriad problems, including drastically decreasing
the thousands of deaths every year caused by
adverse reactions to medication.
Outlook

As a general-purpose technology, AI/ML has use
cases in virtually every industry and the ability
to reshape the way people live and do business.
For this reason, AI has been referred to as "the
new electricity" by deep learning pioneer Andrew
Ng because of its potential to revolutionize
business and our daily lives, similar to the internet
or database technology. Many large markets are
still waiting to be explored.
For now, almost all commercially successful ML
applications use supervised learning, which
addresses a large market but is limited to areas
that have clean, labeled data. Startups face
stout competition from low-cost options available
via the cloud from behemoths like Alphabet
GOOGL, Amazon.com AMZN, Microsoft MSFT,
Baidu BIDU, and Salesforce.com CRM that have
access to data on a massive scale and highly
developed distribution channels. This is especially
important for brute force applications (such
as speech/object/facial/image recognition and
translation) that thrive off large data sets.
AI/ML startups focusing on fringe or specific areas
or data sets may be the best positioned to
compete and thrive. For example, recent advances
in data digitization of medical records and scans
make healthcare especially well positioned to see
further penetration by AI/ML applications. Even
with the resources available today, however, many
potential uses of AI/ML have been hamstrung by a
lack of actionable learning data. K
Cameron Stanfill is an analyst with PitchBook.
http://www.Amazon.com http://www.Amazon.com http://www.Salesforce.com
Table of Contents for the Digital Edition of Morningstar Magazine - April/May 2018
