Presentation by Herman Kienhuis (Curiosity VC) on Investing in AI for ABS Alumni.pdf

Presentation to ABS Alumni
Amsterdam, 13 June 2024
1
CONFIDENTIAL - © 2024 Curiosity Venture Capital B.V.
Investing in AI

2
Introduction
The (Gen)AI revolution
The VC investment landscape
Our investment approach & portfolio
Q&A
Agenda

4
Introduction:
Herman Kienhuis
Chemical Engineering
McKinsey & Company
INSEAD MBA
ilse media / SanomaDigital (10+ ventures)
SanomaVentures (23 investments)
KPN Ventures (18 investments)
River Venture Partners (20 investments)
Curiosity VC (13 investments)
equality - education - sustainability

5
Introduction:
Curiosity VC
✧ €32 mln early stage venture capital fund
✧ Backed by 95 private investors and the Dutch
government (RVO)
✧ Investing in emerging B2B software companies
(in Benelux, Nordics & Baltics), applying
data/ML/AI technologies to create value for
businesses and society
✧ Team of 7, HQ in Amsterdam, supported by a
co-owners community of advisors and founders
✧ Current portfolio 13 companies, including a.o.:
Strise 󰐘, Dreamdata, BeCause 󰎴, Freesi 󰎿,
QA.tech 󰐴, Altura, Deeploy, Neople 󰐗

From AI evolution
to GenAI revolution
6

7
We’ve entered the decade of ‘intelligent computing’
2020’s
mainframe
computers
1970’s
desktop PC’s
1980’s
portable
computers
1990’s
internet
2000’s
mobile &
cloud
computing
2010’s
intelligent
computing
time
ubiquity
and
value
of
computing
● Big Data
○ 5G/IOT
○ Database
technology
● Processing power
○ GPU’s
○ HPC clusters
● Artificial intelligence
○ Computer vision
○ Natural language
processing
○ Machine learning
○ Neural networks
○ Generative AI 🚀

8
AI technologies have been in development for
decades, but has now entered mainstream
1950: Turing test
1956: Term AI was coined
1997: IBM’s Deep Blue defeats
Gary Kasparov
1998: MIT develops an
emotionally intelligent robot
Kismet
1999: Sony launched pet
robot dog AIBO
1970-1990:
AI winter
1961: First industrial robot
replaced humans at assembly
line
1964: MIT develops pioneering
chatbot ELIZA
1966: A general purpose mobile
robot developed at Stanford
2002: iRobot launched
autonomous vacuum
cleaner robot
2009: Google built ﬁrst self
driving car
2011: IBM Watson wins Jeopardy!; Apple
launches Siri
2014: chatbot EUGENE passes Turing test;
Amazon launches Alexa
2017: Google introduces transformer model
BERT
2018: OpenAI introduces transformer model
GPT
2020: OpenAI launches GPT-3
2021: Google (Deepmind) launches
AlphaFold
2022: OpenAI launches ChatGPT;
Midjourney launched
2023: Google launches Bard, Anthropic
launches Claude, Mistral launches 7B
50s 60s 90s 00s 10s 20s

9
Advances in computing power have allowed for training
large language models (LLMs) on all www data
Source: Stanford AI Index Report 2024

10
These LLM models caused a breakthrough in
language understanding and performance
Source: Stanford AI Index Report 2024
Image classification (ImageNet Top-5)
Visual commonsense reasoning (VCR)
Natural language inference (aNLI)
Medium-level reading comprehension (SQuAD 2.0)
Multitask language understanding (MMLU)
Visual reasoning (VQA)
English language understanding (SuperGLUE)
Basic-level reading comprehension (SQuAD 1.1)
Competition-level mathematics (MATH)

11
And OpenAI was the first to bring this technology to
the public via ChatGPT…
Source: Demand Newsletter

12
…with Big Tech and other challengers following
● Launched Bard in 2022, Gemini in 2023
● Started integration with Google
products in 2024
● Invested $2B in Anthropic
● Launched Apple Intelligence in 2024,
based on integration of OpenAI models
● Launched first model (Llama, open
weights) in 2023
● Launched Llama 3 in 2024
● Launched 1st model (Claude) in 2023
and Claude 3 in 2024, outperforming
GPT3.5
● Raised $7.6B
● Launched first model (Mistral7B, open
weights) in 2023
● Raised $1.1B
● Invested $13B in OpenAI (since 2019)
● Integrating LLMs in all MS products
● Launched Co-pilot in 2023
● Launched 1st model in 2022 in 100
languages and focused on enterprises
● Raised $885m
● Started by Elon Musk
● Launched first model (Grok) in 2023
● Raised $6B
● Invested $4B in Anthropic
● AWS launched Bedrock AI offering with
multiple hosted models & tools
● Launched ist model (Pi) in 2023
● Team got acquired by Microsoft
● Raised $1.5B
Mistral 󰏃
Cohere 󰎟
Anthropic
X.ai
Inflection

13
GenAI-driven automation is expected to create massive
productivity increases and trillions in economic value
“AI could increase corporate proﬁts by $4.4 trillion per
year” (The economic potential of Generative AI: The
next productivity frontier, June 2023)
“Generative AI could raise global GDP by 7%” (April
2023)
“Generative AI to Become a $1.3 Trillion Market by
2032; GenAI to add about $280 billion of new software
revenue” (June 2023)
“Generative AI has the potential to raise Dutch GDP
with 9%, e.g. €80-85B in 10 years” (March 2024)

14
And companies are already seeing substantial impact
GitHub Copilot helps
developers code up to 55% faster

15
With many jobs exposed to genAI automation, but only a
small fraction at risk of displacement
Source: Implement Consulting research (Mar 2024)

16
This has created a boom in startups and investments,
building on top of this new GenAI infrastructure
Investments ($B)
Source: Dealroom.co State of AI Investing (May 2024)

17
New applications are emerging, aiming to disrupt
existing services markets
Description: Disrupting:
Launched autonomous software developer Devin;
raised $196m
Commercial copywriting assistant; raised $131m
Legal research and writing assistant; raised $106m
Text-to-speech voice generation services; raised $101m
AI-powered photo creation and editing app;
150m downloads, raised $62m
Video translation/localization app; 2m users
OpenAI powered video creation app
Software
development
Marketing services
Legal services
Voice acting
Commercial
photography
Translation services
Film production
Example

18
Potentially creating inflated expectations for the
short term
Source: Gartner hype cycle for AI (Jul 2023)

19
Also there are clear worries about the risks and
potential negative impact on society
Source: McKinsey State of AI in early 2024

20
And it's apparent that AI technology requires guardrails
and regulations to mitigate key risks
Key risks of Generative AI
1. Hallucinations: GenAI systems can generate incorrect
information
2. Harmful Content: GenAI systems can be used to create
harmful content, e.g. porn deep fakes
3. Algorithmic Bias: GenAI systems can generate biased
information based on its unrepresentative training set
4. Mis-information: GenAI systems can be used to create
fake information at scale for the purpose of
misinformation and influencing opinion
5. Intellectual Property: GenAI systems might breach
copyrighted material used in its training set
6. Privacy: GenAI systems might leak personal identifiable
information provided in prompts or training data
7. Cybersecurity: GenAI systems might be vulnerable to
hacks and can be used to create malware & phishing
content at scale
Model fine-tuning, retrieval augmented
generation (RAG) and education
Filters, fingerprinting and legislation
Improve training data and
bias/representation compensation (e.g.
Google Gemini overdid this)
Filters, fingerprinting, and legislation
Licensing, opt-out system and legislation
Anonymized or synthetic training data; PI
filters for prompts, legislation
Update security scans, education, legislation
Mitigation measures

21
The EU is taking the lead in AI regulation, and recently
approved the EU AI Act
Unacceptable
risk
High risk
Limited risk
Minimal risk
Against EU values, infringing fundamental rights
● Cognitive behavioural manipulation of people
● Social scoring; classifying people
● Biometric identification and categorization
● Remote ‘real-time’ biometric identification
● Products falling under other safety rules
● Critical infrastructure
● Education and employment
● Access to essential public and private services (e.g.
healthcare, financial services)
● Law enforcement, immigration, democratic processes
● Chatbots
● Emotion recognition systems
● Generating or manipulating content
All other AI systems
Prohibited
Requirements and controls throughout life cycle
● Implementation of risk management system
● Technical documentation is formalized and kept up
to date
● AIS compliance and declaration to the authority
Minimum transparency requirements
● inform people that they are interacting with an AI
● When relevant, specificies that content has been
artificially generated or manipulated
● Informs users of its general operation
No obligation towards the AI Act

The Venture Capital investment landscape
22

23
What is Venture Capital?
● Young, innovative,
small companies
● Capital to finance
growth &
development
● Investing in equity,
minority
shareholdings
● Long-term horizon
● Backed by private
investors, family
offices, institutional
investors,
corporates and
public funds

KSF’s for Venture Capital Firms
1. Raise funding
● Track record
● LP network
● Differentiating strategy
2. Source deals
● Proprietary data/network
● Data-driven analysis
(outbound)
● Brand presence (inbound)
4. Win deals
● Reputation & personality
● Speed of decision making
● Terms (founder friendly)
5. Support portfolio company
growth
● Expertise & coaching
(board, advisor)
● Support platform
● Follow-on funding (network)
6. Distribute returns
● Manage towards timely exits
● Terms (investor friendly)
● Manage portfolio risk
3. Assess deals
● Assessment processes
● Expert network
● Quality of decision making
Raise
funding
Source
deals
Assess
deals
Win deals
Support
portco
growth
Distribute
returns
8-10 year
fund cycle
24

Venture Capital investments globally are up again
25
Source: CBInsights - State of Venture (Q1 2024)

26
With AI being one of the leading segments
Source: Dealroom.co State of AI Investing (May 2024)

27
Distribution of VC investments per region

Venture Capital investments in The Netherlands
28

29
In Q1 2024, The Netherlands grew to the #4
spot as VC & startup ecosystem in Europe
Source: Dealroom.co Netherlands Tech Update (Apr 2024)

Our investment approach & portfolio
30

Our Investment Process
Deal sourcing Deal
screening
Deal
Assessment
Term sheet
negotiations
Due
Diligence
• Personal
network
• Investor, advisor
& founder
network
• Startup
databases
• Matchmaking
platforms
• Events
• Pitchdeck
• Product demo
• Key Metrics
• Financials &
forecast
• Captable
• NDA
• Team assessment
• Market &
competition
analysis
• Product USPs &
roadmap
• Customer & sales
pipeline data
• Customer
references
• Unit economics
• Go-to-market
• Hiring plan
• Financial model
• Valuation analysis
• Term sheet • Legal DD
• Commercial DD
• Financial DD
• Tech DD
• Team DD
Transaction
documentation
• Investment
agreement /
share purchase
agreement
• Shareholder
agreement
• Articles of
Association
��
31
500 250 50 10 6 5

32
Criteria for assessment of AI startups
Regular
● Complementary, experienced
team
● Unique solution to a valuable
problem
● Traction with happy customers
● Large, internationally scalable
potential
● Realistic, consistent growth plan
● Fair terms
AI-specific
● Data science & AI knowledge in team
● Access to proprietary training data or
algorithms
● Reinforcing data feedback loop
● Vertical workflow & system integrations
● Customer maturity wrt data/AI
● Compliance with upcoming AI regulation
● AI ethics policy
● Future-proof vs expansion of generic AI
from BigTech

33
How we support portfolio companies
Strategic and financial management
● Help set a clear strategic focus and objectives
● Strengthen financial planning and KPI/metrics management
Commercial development
● Provide customer-centric perspective on proposition and roadmap
● Strengthen data-driven marketing & sales processes
● Introduce to prospects and partners
Organizational & leadership development
● Assist with recruiting of (leadership) talent, introduce to candidates
● Coach founders on personal development and team dynamics/issues
Follow-on fundraising
● Help create fundraising strategy and documentation
● Introduce to investors and support negotiations
General support
● Provide access to experts, best-practices (resources) and p2p and advisor network
● Provide premium access to startup support service providers
pre-seed seed venture growth later-stage
🛠
🎯
󰠼
🛒
💰
Co-owners:
2 General Partners
4 Investment team
16 Expert Advisors
30 Portfolio Founders

35
Strise: KYC/KYB automation for banks

36
Because: Sustainability data platform for hotels

37
Altura: AI bid writing for public tenders & RFPs

38
Orq.ai: LLM model orchestration and prompt
management platform

39
Deeploy: Explainable AI deployment software

40
QA.tech: Fully automated software testing

42
Contact
Amsterdam office:
TNW West
Kon. Wilhelminaplein 1
1062 HG Amsterdam
Herman Kienhuis
herman@curiosityvc.com

Presentation by Herman Kienhuis (Curiosity VC) on Investing in AI for ABS Alumni.pdf

More Related Content

Similar to Presentation by Herman Kienhuis (Curiosity VC) on Investing in AI for ABS Alumni.pdf

More from Herman Kienhuis

Recently uploaded

Presentation by Herman Kienhuis (Curiosity VC) on Investing in AI for ABS Alumni.pdf