Qu speaker series:Ethical Use of AI in Financial Markets

Qu Speaker Series
Ethical Use of AI in Financial Markets
2020 Copyright QuantUniversity LLC.
Hosted By:
Sri Krishnamurthy, CFA, CAP
sri@quantuniversity.com
www.qu.academy
11/12/2020
Online
https://quspeakerseries13.splashthat.com/

2
QuantUniversity
• Boston-based Data Science, Quant
Finance and Machine Learning
training and consulting advisory
• Trained more than 1000 students in
Quantitative methods, Data Science
and Big Data Technologies using
MATLAB, Python and R
• Building a platform for AI
and Machine Learning Exploration
and Experimentation

3
For registration information, go to
https://QuFallSchool.splashthat.com

4
https://Quwinterschool.splashthat.com

7
Speakers
Daniel (Dan) is the Founder of Singapore based Lightbulb Capital. The firm was founded in 2014 in Hong Kong to help
realize the potential of innovation and novel technologies to transform financial services. Today the company is a
corporate finance boutique with a focus on FinTech and blockchain.
Dan is appointed affiliate faculty of Singapore Management University for Innovation in Finance, FinTech, Blockchain, and
digital assets. He is also a visiting professor at IE business school in its top-ranked Master in Finance program. He is also a
Review Editor for Frontiers Financial Blockchain academic journal and works on research projects. In addition, he is a
frequent speaker at seminars on Blockchain and Innovation in Financial Services. Before starting the firm, Dan was Chief
Operating Officer (COO) and an Executive Director of HSBC Securities (Singapore) Pte Limited. He was previously also the
IT Head of HSBC’s Investment Bank in Singapore and Japan. Dan has over 19 years of Investment Banking Technology
experience at UBS in both Germany and the UK, Barclays Capital in both Singapore and Tokyo, as well as Close Brothers in
Frankfurt, his hometown.
He is also a Ph.D. Candidate within the Finance department of Rotterdam School of Management, Erasmus University
where he researches Blockchain and AI and it effects on Finance. Previously he graduated with a Master of Science in
Innovation from Singapore Management University and holds a Master in Finance from IE business school in Madrid,
Spain.

Ethical use of AI in Financial Markets
Not for public distribution

AGENDA
© 2020 Lightbulb Capital – Not for public distribution | 2
• Authors
• Introduction (to AI/ML in Financial Markets)
• Who needs to control who?
• Governments & the ethical use of AI
• Principles – applied to Financial Markets:
• Fairness
• Privacy
• Transparency
• Explainability
• Accountability
• Conclusion
• Q&A
• Contact Information

INTRODUCTION
Authors
Dan Liebau
Co-Author
Current
• Founder of Singapore based Lightbulb Capital, Singapore and Hong Kong
• Affiliate Faculty of Singapore Management University, Singapore
• Lecturer at Rotterdam School of Management, Erasmus University, Netherlands
• Review Editor for Frontiers Financial Blockchain, London
Previous Appointments
• Visiting Professor at IE business school, Spain
• Chief Operating Officer (COO) and an Executive Director of HSBC Securities (Singapore) Pte
Limited.
• IT Head of HSBC’s Investment Bank in Singapore and Japan
• Head of IT, Close Brothers
• Equities IT, Barclays Capital
• Equities IT, UBS
Education
• Ph.D. Candidate at Erasmus University, Research focus on: Blockchain and AI and it effects on
Finance, Netherlands
• Master of Science in Innovation from Singapore Management University, Singapore
• Master in Finance from IE business school in Madrid, Spain

INTRODUCTION
Authors
Tiffany Wong
Co-Author
Current
• Senior Associate at PwC London in the Financial Services Risk and Compliance
• area
• She specialises in Technology Risk and Digital
• Has led the PwC "Disruptive Technology Working Group"
Education
• First Class from BSc (Hons) Actuarial Science at CASS Business School
• CFA Charter-holder
• Chartered Accountant
• Has contributed to In the CFA Institute Consultation of "Artificial Intelligence and its potential
impact on the CFA Institute Code of Ethics and Standards of Professional Conduct”

ETHICAL USE OF AI IN FINANCIAL MARKETS
Definition: Machine Learning
Machine learning is an application of
artificial intelligence (AI) that provides systems the
ability to automatically learn and improve from
experience without being explicitly programmed. It
focuses on the development of computer programs
that can access data and use it to learn for
themselves. (Source: Google)
In the Financial Markets context it is NOT:
- Algorithmic trading (usually expert systems)
- High frequency trading (fast but often also static)

Definition: Natural Language Processing
Natural language processing (NLP) is (…) concerned
with the interactions between computers and human
(natural) languages, in particular how to program
computers to process and analyse large amounts
of natural language data. (Wikipedia)

Definition: Data Science
Data science is a multi-
disciplinary field that uses
scientific methods, processes,
algorithms and systems to extract
knowledge and insights from
structured and unstructured data.
(Wikipedia)

Brief history of AI/ML in markets
5
milestones
in the use of AI in Financial Markets

Brief history of AI/ML in Markets
1. Renaissance Technologies started in the in New
York in 1982, they were the first to use Machine
Learning techniques in Financial Markets

2. Alternative Data is now big business
- Satellite imagines can be used to predict
circulating Oil supply
- FinTech Start-ups like BlueFireAI scrape Chinese
blogs to create of risk signals in the equity market
- Companies like Neudata, Refinitiv and others offer
alternative data for Investment Managers to digest
- Researchers commence to analyse the impact of
alternative data on markets

3. Marcos Lopez de Prado “happened”
He publishes a paper on:
- “A Data Science Solution to the Multiple-
Testing Crisis in Financial Research”
(where he proposes to address selection
bias under multiple testing (SBuMT) by
using the deflated Sharpe ratio (DSR))
- And an entire book on “Advances in
Financial Machine Learning”

4. In 2018 Gu, Kelly and Xiu
publish the paper “Empirical
Asset Pricing via Machine
Learning” on the SSRN platform
Key finding: Machine Learning
can provide an improved
description of expected return
behaviour compared to
traditional forecasting methods

5. In 2019 CAIA launches the Financial
Data Professional Certification (FDP):
- Students have to pass several
Python classes before they can
register for their exam
- Self-study program with latest
academic research papers and
publications

An Oxymoron?

A chapter in a Wiley publication
Part 7: Trust, Transparency & Ethics
Finance may well be a test for the trust consumers place in
Artificial Intelligence.
Data is essential for the AI industry to flourish but what
steps do financial services need to take to ensure
customers trust the sector to innovate through AI?
And what is the trade-off between consumer benefits and
privacy? E-commerce, PSD2 and the rapid increase of
smartphone ownership mean consumers demand efficient
services alongside strong security and privacy protections.
This chapter will focus on the key issues of transparency
and trust, and will outline what digital ethics mean for the
financial sector. As ethics is becoming mainstream in the
debate around AI, contributors will outline what are the
ethical challenges facing fintech players and institutions
alike, from creditworthiness to algorithmic bias and
fairness, and use of personal data and transparency.

Quote No. 1: Who (needs to) control who?
“Humans have found it impossible to encode ethical standards in finance,
so we generally get principles-based guidance like ‘do the right thing’. It will
take AI a long time to learn the difference between right and wrong on
matters where human experts often disagree.”
Matthew Cannon, Managing Partner, Hedge Fund Modular Asset Management, Singapore

Quote No. 2: Who (needs to) control who?
“Financial transactions are fraught with agency problems. One solution is to
replace human judgement with smart, systematic processes. Another
solution is to deploy algorithms that monitor human decisions in search the
for biases. In both cases, AI has an important role to play.”
Dr Marcos Lopez de Prado, True Positive Technologies LP, New York

Governments start to pay attention to AI ethics

Even China is in on it…

Zooming in on Financial Markets
Our chapter focuses on operationalizing principles to facilitate the ethical
use of AI in Financial Markets
AI principles for
fair use in
financial
markets
Fairness
Privacy
Transparency
Explainability
Accountability

Fairness
Background What is the issue? Recommendation example
Treating stakeholders impartially and
just, without taking unfair advantage
of them is not a new matter in
markets
Perhaps the fairness issue is as old
as financial markets themselves
Reinforcement learning algorithms
get compensated through a feedback
loop depending on their
performance.
They will learn about patterns in the
training data suggest these practices
to exploit them profitably.
When building AI trading algorithms,
we recommend defining boundaries
to delineate fair from unfair trading
activity
ML algorithms should not be allowed
to engage in spoofing, wash trading,
stock bashing or pump & dump / run-
type of activities
Financial Market regulation does this
today for human traders

Privacy
The Cambridge Dictionary
defines privacy as
“someone's right to keep
their personal matters and
relationships secret”
(Cambridge Dictionary,
2019).
The centralized financial
marketplaces of today
display weaknesses when it
comes to private data
protection.
Hong Kong’s Central Bank reported that in 2018,
cyber attacks on financial services providers
doubled (Hong Kong Monetary Authority, 2019).
It is easy to imagine how hackers will use AI to
exploit markets unfairly by getting access to private
data to use it to their advantage in the markets.
For example, the market participants may have to
disclose ultimate beneficial ownership data of their
executives. If illegally obtained, such data could be
used to augment the dataset fed to an AI algorithm
to achieve superior trading results without
observing privacy rights.
One solution to preserving the
privacy of actors in the financial
markets could be the use of
Zero-Knowledge Proofs (Simari,
2002).
Secret information does not
have to be revealed for
counterparts to establish
probabilistic certainty of, for
example, identity-related data. If
stored on a blockchain, data
owners could selectively share
Zero-Knowledge Proofs on data
attributes more securely and as
required to interact in the more
private markets of the future.

Transparency
Transparency is the quality of
making something obvious
or easy to understand
(Merriam-Webster, n.d.).
In our context, it refers to the
public’s view of how and if
AI’s functionality is aligned
with reality.
In 2018, an AI hedge fund, Tyndaris,
lost $20 million of an investor’s
funds. The investor sued the fund
because the losses wouldn’t have
occurred if the machine learning
algorithm in use was as
sophisticated as the salesperson
claimed it was (Beardsworth et al.,
n.d.).
The case has not been settled in
court at the time of writing. However,
there is an expectation gap between
the investor’s understanding of the
technology and its actual capability.
To address this, we recommend investment
managers using AI to be transparent beyond
mandatory disclosures in both the investment
prospectus and regular reporting.
We also believe additional education is
necessary, too: This could include
‘Fundamentals of Machine Learning in
Finance’ (similar to the classes for novice
investors in derivatives products).
Beyond such education, we believe a more
detailed explanation of biases (such as
herding) that algorithms may display, seem to
be most relevant to support the investor in
understanding the AI related risks.

Explainability
Explainability is the ability to
explain the rationale behind
an action (in our case, an
investment decision). The
objective is to enable
humans to understand,
appropriately trust and
effectively manage what
artificially intelligence
partners do (Gunning, 2018).
The more sophisticated
machines are, the less
explainable the mechanisms
are.
An investor wishes not to invest in
the defence sector. This decision has
been reflected in the investment
constraints of the fund.
However, upon an act of war, an AI
algorithm decided to short a
particular stock (think Airlines
around September 11 2001) to
maximise trading profits.
While every investor’s moral
compass works differently, this may
or may not be acceptable to the
investor.
We recommend for AI algorithms to be
regularly stressed-tested to help establish
behavioural patterns, that can form part of
their decision-making process.
Side-note: We should not forget that often a
human decision-making process is not entirely
explainable too.

Accountability
Accountability is the ability to
hold an individual / an entity
accountable for an act. It
includes the ability to admit
a flaw as a consequence of
an action or behaviour.
Accountability goes hand in
hand with explainability in
Artificial Intelligence. If a
decision is explainable,
actors can more naturally be
held responsible for their
actions.
AI is only an enabling technology
used by a financial services provider
The management can be held
accountable similarly to when a
human staff member errs and
causes economic damage.
Data sources become more and
more plentiful.
Financial markets entities using AI should
enact a data policy to limit risks of data
related incidents.
Such a policy governs privacy but also data
quality and integrity.
Additionally, the severity and probability of
potential harm created by an AI algorithm in
the markets should drive the human
involvement and controls required.

Accountability

Conclusion
• The pace of technological and data related change accelerates
exponentially
• Machine Learning cannot be considered an emerging technology
anymore
• It is not free of ethical risks that need to be managed carefully
• It is essential to operationalize ethical principles by thinking about how
they specifically apply to the financial markets

Conclusion, continued
• We suggest operationalizations for transparency, explainability,
accountability, privacy and fairness
• We should, however, refrain from implementing more rigorous controls
for AI than the ones we put in place for humans
“I see no distinction between AI and people since the latter are just
biological AI”
Henry Young, ex-CEO of High-Frequency Trading Technology firm TS-Associates

Q&A

THANKYOU
&PLEASEGETINTOUCH
info@lightbulbcap.com
www.lightbulbcap.com
Lightbulb Capital (Singapore) Pte. Ltd.
28C Stanley Street
Singapore 068737
Check out the Singapore FinTech
Champions 2018 here:
https://www.lightbulbcap.com/singapore-
fintech-champions
www.facebook.com/lightbulbcap
www.linkedin.com/company/lightbulbcap

9
Demos, slides and video available on QuAcademy
Go to www.qu.academy
9

10
Instructions for the Lab:
1. Go to https://academy.qusandbox.com/#/register and register using the code:
"QUFALLSCHOOL"

Thank you!
Sri Krishnamurthy, CFA, CAP
Founder and CEO
QuantUniversity LLC.
srikrishnamurthy
www.QuantUniversity.com
Contact
Information, data and drawings embodied in this presentation are strictly a property of QuantUniversity LLC. and shall not be
distributed or used in any other publication without the prior written consent of QuantUniversity LLC.
11

Qu speaker series:Ethical Use of AI in Financial Markets

More Related Content

Qu speaker series:Ethical Use of AI in Financial Markets