4th ANNUAL AI 100

100 AI Startups And

Trends To Watch In
2020

1
For our clients, CB Insights accelerates their digital
strategy & transformation efforts with data, not opinion.

We offer the smartest way to

discover and manage your response
to emerging tech and startups.
cbinsights.com
CB Insights helps us compress our time-
to-decision when gathering and
analyzing data and getting an external
view on what's happening in the market
so we can quickly take action.

Meraj Mohammad
Vice President, Ventures Group, ADP
The CBI Community — Membership Has Perks
DISCOUNTED TICKETS TO
WHAT IS CB INSIGHTS?

CB Insights accelerates our clients' digital

strategy and transformation efforts with
data, not opinion.

We offer the smartest way to discover and

manage your response to emerging tech
and startups.

CLICK HERE TO SIGN UP FOR FREE

ABOUT THE ANALYST

Deepashri Varadharajan

Lead Intelligence Analyst & Writer, AI

@deepalearning | dvaradharajan@cbinsights.com

Deepashri Varadharajan is a lead analyst at CB Insights, researching

the intersection of AI and different industries. She is passionate about
studying new business models and opportunities that AI is helping
shape, and how traditional corporations and governments are adapting
to a rapidly changing landscape.
Deepashri received her undergraduate degree in electronics and
communications engineering from VIT University in India and an MS
from the Columbia University Graduate School of Journalism. Before
CB Insights, she worked for media houses including Al Jazeera
America.
Contents
What we’ll cover

8 14

How we pick the 100 most Trends to watch in 2020

promising AI startups

6
2020 AI 100 cohort
• Represent 13 countries, including Chile, China,
Sweden, South Africa, and Japan
• Operate across 15 core industries and several
cross-industry solutions
• Mix of startups are different stages of R&D and
commercialization
• Over $7.4B in funding across 300+ deals from
600+ unique investors.

7
How we picked the AI 100 finalists (1/2) Methodology
The AI 100 selections are a
rigorous, data-informed process,
combining CB Insights datasets
with the research team’s industry
expertise.

The CB Insights datasets that were

leveraged to pick finalists include:
• Patent Analytics to assess R&D
strength and direction
• Company Mosaic Scores to
evaluate startup health and
growth potential, based on our
National Science Foundation-
backed algorithm.
• Business Relationships to
identify partnerships and client-
vendor relationships
• Market Sizing Tools to assess
growth opportunities and
demand for the tech
• Competitor Data to assess tech
novelty, market strength, and
competitive landscape
• News Trends for sentiment
analysis
How we picked the AI 100 finalists (2/2)

Business model feasibility Tech feasibility

Investor quality Team strength

Recent R&D or product launches

Data submitted by applicants
2020 AI 100 cohort
• Represent 13 countries, including Chile, China,
Sweden, South Africa, and Japan
• Operate across 15 core industries and several
cross-industry solutions
• Mix of startups are different stages of R&D and
commercialization
• Over $7.4B in funding across 300+ deals from
600+ unique investors.

10
Startups at different stages of funding & R&D AI 100 Unicorns
The list includes well-funded
companies, with 10 startups valued
at $1B+, working on a diverse range
of AI solutions. These include:

• Faire: Wholesale marketplace

that uses machine learning to
match local retailers with goods
that are predicted to sell well in
their specific locations.

• Butterfly Network: Portable

ultrasound device with AI-
assisted diagnostics

• Snyk: Security solutions for

DevOps

• Graphcore: Processors for AI

workloads

11
Startups at different stages of funding & R&D AI 100: Promising
early-stage startups
The list also includes early-stage
startups forming partnerships with
key industry players, and
companies with a strong R&D
focus.

• Israel-based Razor Labs works

with mining companies to
provide AI-enabled solutions.

• Zesty.ai, an insurance tech

startup focused on catastrophe
risk, recently partnered with
MetLife to use AI to help predict
wildfire risk in California.

• InstaDeep is focused on PCB

design as well as core AI R&D,
including reinforcement learning
research.

• Dorabot is a startup focused on

reinforcement learning for
robotic manipulation and
warehouse logistics.

12
AI 100

Trends to watch
in 2020

13
2020 trends to watch

How Alphabet will

help city
governments
optimize resources

14
Alphabet’s org structure: Key AI initiatives

15
Alphabet’s org structure: Key AI initiatives

16
Waste recycling
AMP Robotics is an AI and robotics
companies building solutions for
waste recycling. Its product focus
includes:

• Identifying and sorting

electronic waste

• Sorting municipal solid waste

• Handing materials at
construction and demolition
sites

Sidewalk and AMP have partnered

to launch a pilot in Toronto.

17
Urban planning
A fundamental part of smart city planning is
understanding how people commute in an area,
including where and when they travel, and what
influences their behavior and choices, such as using
public transport.

Sidewalk Labs spinout Replica is using machine

learning and location data from mobile phones to
answer these questions. Replica claims it does this
without collecting any personally identifiable
information on commuters.

The Portland government will pay Replica over $450K

for a one-year initial service period, and Illinois has
signed a $3.6M deal with the startup for a 3-year
period.

18
2020 trends to watch

Federated learning will

bring in a new data
partnership ecosystem

19
How can you partner with other organizations for
building AI capabilities, without sharing your raw data?

20
The federated learning approach Training AI with siloed data
Federated learning shows promise in training AI in
industries with sensitive and siloed data. In 2020, it
will enable a new data partnership model without
requiring users to actually share the raw data.

Google initially debuted the tech for its Android

keyboard, Gboard, to predict what a user will type
next. The approach is now finding applications in
industries including financial services and
healthcare.

Implications
• Global model, local data: With federated
learning, users can train AI on data stored locally,
and only share AI model updates with a cloud
server. The “global model” will then benefit all
partners in the network to improve their local AI
applications.

• Increase data diversity: Federated learning will

enable more cross-institutional or cross-country
partnerships, eventually allowing the global AI
model to benefit from diverse local datasets.

Image source: Google AI blog; Read more about federated learning in the 2020 AI trends report. 21
 Federated learning in
healthcare
Google Ventures-backed AI startup
OWKIN, is using the approach to
enable different cancer treatment
centers to collaborate without
patients’ data ever leaving the
premises.

Nvidia partnerships

Nvidia and OWKIN recently

partnered to expand federated
learning applications in healthcare.

Nvidia’s AI-powered hardware and

software framework for healthcare,
called Clara, now supports
federated learning. Initial users of
the tech include American College
of Radiology, MGH and BWH Center
for Clinical Data Science, and UCLA
Health.

Image source: OWKIN 22

2020 trends to watch

Commercial deepfakes
will automate brand
marketing

23
Commercial deepfakes are here Deepfakes are hyper-
realistic AI-generated
images and videos. Media
companies are ready to
monetize the “benign” side
of the controversial tech.
• Startup Superpersonal
swaps out users’ faces in
short video clips for virtual
try-ons.

• Samsung published a paper

on using neural nets to
create realistic “talking
heads.”

Image source: Superpersonal (top), Samsung AI research (bottom) 24

How AI was used to make
David Beckham ‘speak’ 9
language
Startup Synthesia used deepfake
tech to make David Becham speak
in 9 different languages in a
campaign video for the NGO
Malaria Must Die.

The startup has since raised $3M

from LDV Capital, Mark Cuban, and
others.

25
Hyper-realistic synthetic voice
generation is another emerging
application of AI in media &
entertainment.

Wellsaid Labs, a spin out from the

Allen Institute of Artificial
Intelligence, is focused on
generating synthetic voice overs for
news, media, education, and other
industries.

26
2020 trends to watch

Next-gen hacking: AI
trojans, voice spoofing,
and smart evasion will
evolve

27
Cyber AI startups will face new attack vectors New-age hacks will evolve on
two fronts: fooling AI systems
Funding to startups developing AI-based cybersecurity solutions and leveraging AI to launch
sophisticated attacks.
As cyber AI startups raise
more funds to protect
companies and consumers
with the tech, a new crop of
hackers and malware that
target weaknesses unique to
AI will emerge.

“...if you could truly

understand how a certain
model works, and the type of
features it uses to reach a
decision, you would have the
potential to fool it
consistently, creating a
universal bypass.” –
Skylight Cyber
28
With open-source tools
available, the barrier for entry
into AI is low for hackers.
• Media reports of AI-generated
voice spoofing first emerged in
Europe in Q3’19.

• IBM developed a proof-of-

concept for a deep learning-
powered malware, dubbed
DeepLocker, back in 2018.
Hackers can make hyper-
targeted malware by concealing
it within AI code. DeepLocker
only unlocks the attack when it
encounters a very specific
trigger, such as a particular
person’s face or voice.
Otherwise, the code sits
undetected in benign everyday
applications.

29
AI 100: Cyber AI startups to watch in 2020

Blue Hexagon SnetinelOne

Total funding: $37M Total funding: $430M
Select investors: Benchmark, Select investors: Insight Partners, Accel,
Altimeter Capital Redpoint Ventures, Vista Equity Partners,
Samsung Ventures, Qualcomm Ventures

Obsidian Security Abnormal Security

Total funding: $29.5M Total funding: $24M
Select investors: Google Ventures, Greylock Select investors: Greylock Partners
Partners, and Wing Venture Capital

30
Track all 100 startups using the CB Insights AI 100 (2020) Collection
Or sign up for a free 30-day trial here
dvaradharajan@cbinsights.com
@deepalearning
@cbinsights
Thanks for joining