This presentation was given at the Empowerment Summit and is about the current of regulation with respect to Articifical Intelligence (AI) and contains some prospective insights about the elements that AI regulation should take into account

1. Artificial Intelligence.
Technological Singularity.
Law.
Florian Ducommun, HDC, Attorney-at-Law
1

2. < 1 > Artificial Intelligence (AI)
2

3. Algorithm
Set of instructions within computer programs that determine how these
programs read, collect, process, and analyze data to generate some readable
form of analysis or output
(Cormen, Thomas H.; Leiserson, Charles E.; Rivest, Ronald L.; Stein, Clifford (2009).
Introduction to algorithms (3rd ed.). Cambridge, Mass.: MIT Press. p. 5)
3

4. Instructions
Every algorithm must satisfy the following criteria:
 input;
 output;
 definiteness;
 finiteness;
 effectiveness.
4

5. Algorithmic bias
 Algorithmic bias
 Bias can be introduced to an algorithm :
- during the assemblage of a database;
- as a result of algorithm design.
5

6. < 2 > Technological Singularity
 The technological singularity (also, simply, the singularity) is the
hypothesis that the invention of artificial superintelligence (ASI)
will abruptly trigger runaway technological growth, resulting in
unfathomable changes to human civilization.
6

7. Technological Singularity
1950s : John von Neumann
7
1945: Isaac Asimov
1993 : Vernor Vinge

8. Where are we?
 AlphaZero (2017)
 Deep Neural Networks
 General Reinforcement Learning Algorithm
 Google Brain
-> develops AI that allegedly builds AI better and faster than humans can
8

9. Dystopian Future
9
 Utopian describes a society that's
conceived to be perfect.
 Dystopian is the exact opposite
— it describes an imaginary
society that is as dehumanizing
and as unpleasant as possible.

10. <3> Legal
How to
regulate AI?
10

11. Challenges
 AI's moral maze and the infinite scenarios that are still to be addressed;
 Exponential speed of technological progress;
 International/transnational/global regulation required
11

12. Legal questions
AI will raise legal questions in three areas:
 Rights: Are/Should algorithms ever be protected by IP Rights?
Could algorithms create IP rights?
 Accountability: Who or what should be liable if A.I causes harm ?
 Ethics: How can we create and enforce moral codes for AI?
12

13. Rights
13

14. Copyright
 Idea/ expression dichotomy
 Balance between fostering
creativity/innovation and public good
 Originality
 Author = physical person (Art. 6
Copyright Act)
14

15. Copyright
15
Copyright protects the “literal expression” of computer
programs (source and object code), it does not protect
the “ideas”underlying the computer program.
Art. 2 para. 3 Swiss Copyright Act: Computer programs
are literary works.
-> not applicable to algorithms

16. Open Source
 a type of computer software whose source code is released
under a license in which the copyright holder grants users
the rights to study, change, and distribute the software to
anyone and for any purpose
 Relates to the copyrights in and to the computer software
 Increased transparency and allowed collaborative software
development
16

17. Patents
17
To be eligible for patent protection, an invention must meet several
criteria:
(i) the invention must consist of patentable subject matter;
(ii) the invention must be capable of industrial application (or, in certain
countries, be useful);
(iii) it must be new (novel);
(iv) it must involve an inventive step (be non-obvious); and
(v) the disclosure of the invention in the patent application must meet
certain formal and substantive standards.

18. Patents
18
Abstract algorithms cannot be patented: the organization and
ordering of a series of math equations is not human invention.
However, a patent can be granted on an application of an abstract
idea
Dividing line between unpatentable abstract ideas or algorithms and
patentable inventions, which include processes.

19. Patents
19
-> Protection during 20 years of a business method
or process encapsulated into an algorithm
-> Disclosure but full control by the patent owner
-> Not suitable for algorithms

20. Trade secrets
20
Broadly speaking, any confidential business information
which provides an enterprise a competitive edge may be
considered a trade secret. Trade secrets encompass
manufacturing or industrial secrets and commercial
secrets.
.

21. Trade secrets
21
Trade secret means information which meets all of the
following requirements:
- secret in the sense that it is not, as a body or in the
precise configuration and assembly of its components,
generally known among or readily accessible to
persons within the circles that normally deal with the
kind of information in question;
- commercial value because it is secret;
- it has been subject to reasonable steps under the
circumstances, by the person lawfully in control of the
information, to keep it secret.
-> Algorithms can be protected by trade secrets

22. Trade secrets
22
Trade secret law lacks any expressly delineated justice
balancing mechanism between the rights holder and the
public, unlike patents and copyright.
But:
-> Privatisation of algorithms by companies that developed
them for a unlimited period in time.
No disclosure of the underlying instructions, criteria -> no
possibility to discover potential algorithmic bias.

23. Antitrust Law
 Algorithms may foster tacit collusion, adversely affect consumer choice, even pose a threat to
pluralism.
 Holders of algorithms as trade secrets could be in a dominant position.
 The spreading of “static”, non-learning algorithms may facilitate collusion in markets that were not
prone to such conduct.
 Dynamic, deep-learning algorithms may, once they are broadly implemented, require an
adjustment of competition law concepts such as causality, awareness and intent.
23

24. Data Protection
 Algorithms treat data -> consent required for such treatment.
 Art 71 Recitals / Art. 22 para 3 GDPR:
Right not to be subject to a decision, which may include a measure, evaluating personal aspects relating to him or her
which is based solely on automated processing , such as automatic refusal of an online credit application or e-
recruiting practices without any human intervention.
Such processing should be subject to suitable safeguards, which should include specific information to the data subject
and the right to obtain human intervention, to express his or her point of view, to obtain an explanation of the decision
reached after such assessment and to challenge the decision.
 right to explanation is a right to be given an explanation for an output of the algorithm.
24

25. Open Algorithms
 Open Algorithms: open source for algorithms
 Google claims to license the algorithms on Open source licenses
 Initiatives:
- OPAL (for "Open Algorithms") is a non-profit socio-technological innovation developed by a group of partners
aiming to unlock the potential of private sector data for public good purposes by “sending the code to the data” in
a safe, participatory, and sustainable manner (https://www.opalproject.org/)
- Open marketplace for algorithms in a range of domains (https://algorithmia.com/algorithms).
25

26. Accountability
26

27. Accountability
27
Accountability implies an obligation to report and justify algorithmic
decision-making, and to mitigate any negative social impacts or potential
harms.

28. Five core principles
28
Responsibility: For any algorithmic system, there needs to
be a person with the authority to deal with its adverse
individual or societal effects in a timely fashion
Explainability: any decisions produced by an algorithmic
system should be explainable to the people affected by
those decisions.
Accuracy: sources of error and uncertainty throughout an
algorithm and its data sources need to be identified,
logged, and benchmarked

29. Five core principles
29
Auditability: algorithms should be developed to enable
third parties to probe and review the behavior of an
algorithm.
Fairness: all algorithms making decisions about
individuals should be evaluated for discriminatory effects.
The results of the evaluation and the criteria used should
be publicly released and explained.

30. Robots or AI as
“electronic
person” ?
30
Sophism

31. Taxation of
robots?
31
Sophism

32. Ethics
32

33. Isaac Asimov's "Three Laws of Robotics"
 A robot may not injure a human being or, through inaction, allow a human being to come to
harm.
 A robot must obey orders given by human beings except where such orders would conflict with
the First Law.
 A robot must protect its own existence as long as such protection does not conflict with the First
or Second Law.
33

34. Oren Etzioni’s proposal
 AI system must be subject to the full gamut of laws that apply to its human operator;
 AI system must clearly disclose that it is not human;
 AI system cannot retain or disclose confidential information without explicit approval from the
source of that information.
34

35. Google guidelines
Any set of rules for AI should focus on predicting harm, mitigating risk, and ensuring safety is a priority
 Avoiding Negative Side Effects: How can we ensure that an AI system will not disturb its environment in
negative ways while pursuing its goals?
 Avoiding Reward Hacking: How can we avoid gaming of the reward function?
 Scalable Oversight: How can we efficiently ensure that a given AI system respects aspects of the
objective that are too expensive to be frequently evaluated during training?
 Safe Exploration: How do we ensure that an AI system doesn’t make exploratory moves with very
negative repercussions?
 Robustness to Distributional Shift: How do we ensure that an AI system recognizes, and behaves
robustly, when it’s in an environment very different from its training environment?
35

36. EU Ethical Guidelines on AI
 As with any transformative technology, artificial intelligence may raise new ethical and legal
questions, related to liability or potentially biased decision-making. New technologies should not
mean new values.
 The Commission will present ethical guidelines on AI development by the end of 2018, based on
the EU's Charter of Fundamental Rights, taking into account principles such as data protection
and transparency
36

37. Working Group
1. Should we regulate AI? If yes, at which level?
2. Should algorithms be protected by IP rights (trade secrets)?
3. Should companies have an obligation to disclose algorithms?
4. What should be the guiding principles of regulation?
37

38. Guiding principles (European Group on
Ethics)
 Human dignity
 Autonomy
 Responsibility
38

39. Guiding principles (European Group on
Ethics)
 Justice, equality and solidarity
 Democracy
 Rule of law and accountability
39

40. Guiding principles (European Group on
Ethics)
40
Security, safety, bodily and mental integrity
Data protection and privacy
Sustainability

41. Merci pour votre attention!
Florian Ducommun
Avocat ; LL.M
Av. Auguste Tissot 2bis
1006 Lausanne
021/310.73.10
ducommun@hdclegal.ch
41