Morningstar Magazine - April/May 2018 - 49

EXHIBIT 2

Opportunities in unsupervised and reinforcement
learning present some exciting scenarios at
the frontier of the vertical. While these techniques
still face a fair number of obstacles, the
possibilities could offer solutions to a large array
of problems. One of the most significant early
achievements in reinforcement learning was
AlphaZero, originating from Google's DeepMind.
This algorithm mastered the games of chess,
Go, and Shogi purely via simulated games against
itself. Reinforcement learning also has the
potential to improve on traditional solutions for
resource allocation problems, personalized user
interfaces (such as healthcare and media),
and applications in robotics and autonomous
vehicles. While there are hurdles to bringing these
techniques into commercial products, proof of
concept at the bleeding edge of AI/ML can attract
significant investment. This has been
demonstrated by the amount of capital flowing
into the autonomous vehicle space. In the end,
the most adaptable VC-backed companies
%"!
and investors will succeed in the AI-centric world.
%!!
That said, many of these applications will take
more time to come to fruition because of the
$"!
complexity of the problems and the computation
power that they require. In response to these
$!!
issues, advances are being pursued in supporting
hardware-such as quantum/high performance
#"!
and hybrid computing-as well as decomposition
techniques to break down complex problems
#!!
into manageable segments.
VC Trends

"!

Venture investment in the AI/ML space tells!a clear
story about its advancement and the increasing $!!&
interest in the vertical over the past decade. Last
year, $6 billion was invested across 643 VC deals in
AI/ML. Paralleling technological breakthroughs,
deal counts have swelled to levels 12 times above
what we saw in 2008 ( E XH I BI T 1 ).
Deal counts across all stages have pushed
higher, but the vast majority of growth has come
through angel and seed and early stage deals
( E X H I B IT 2 ). While mainstream media exposure
of AI/ML has ratcheted up recently, it's clear
that the vertical is still in the early innings.
Even so, there is a segmentation in the market.
Some larger companies are already executing

VC Activity by Stage Angel/seed and early stage deals are driving growth.
Angel/Seed

Early Stage VC

Later Stage VC
300

286
266

250
200
150
91

100
50
0

2006

2007

2008

2009

2010

2011

2012

2013

2014

2015

2016

2017

Source: PitchBook.

EXHIBIT 3

Exit Activity A rise in exit counts indicates that artificial intelligence/machine
learning is entering a new period in the VC cycle.
Exit Value

Exit Count
41.0

$3.0
Bil
2.54

2.5

45.0
37.5

2.0

30.0

1.5

1.44

1.0

22.5
15.0

0.5
0.0

0.01

0.00

0.10

0.00

0.06

2006 2007 2008 2009 2010

0.24
0.00

2011

0.14

0.27

7.5
0.0

0.37

2012 2013 2014 2015 2016 2017

Source: PitchBook.

$!!'

$!!(

$!!)

$!#!

+,-./01..2
on commercial products-such
as online lender
Avant and used-goods marketplace Letgo-
while many others are working to refine tools
as well as expand their applications and product
offerings. Theoretical and practical research
have made huge steps in proving functionality,
but we are still in the initial period of
determining feasibility and use cases for many
businesses. Furthermore, due to the uniqueness
of each data set, execution of AI/ML
techniques in a business usually requires a high
level of customization that hinders early
widespread adoption.

$!##

$!#$

$!#%

$!#*

$!#"

$!#&

345/67
184-.7 9: have exhibited
;48.57rapid
184-.7valuation
9:
AI companies
growth similar to that of the broader VC market.
The exception is the late stage, where medians
have flattened out after a pop in 2013 and
2014, unlike the exceptional growth in 2016 and
2017 experienced by late-stage VC companies
in aggregate. One reason for this anomaly is that
a host of large AI/ML companies raised funds
in 2013 and 2014, including multiple rounds raised
by Palantir, and have yet to return for sizable
follow-on rounds. Additionally, the vertical is still
dominated by earlier-stage companies, which
tempers the median numbers. This is a bit

global.morningstar.com/Morningstarmagazine

49
http://global.morningstar.com/Morningstarmagazine
Table of Contents for the Digital Edition of Morningstar Magazine - April/May 2018
