AI Created Works and Authorship under Irish Copyright Law: I'm Afraid I Can't Copyright That

• Date: 01 January 2020

1

AI Created Works and Authorship under Irish

Copyright Law: I’m afraid I can’t copyright that

KEVIN FLOOD*

I. Introduction

All technology starts from humble beginnings; from computers the size of a

room to smartphones, oppy disks to streaming services and digital storage in the

‘Cloud’. Technology that would have been unimaginable y years ago is treated as

a normal part of everyday life. e problem that arises is when the advancements

made begin to outpace our ability to control and regulate the technology itself.

Regulating technological advancements provides challenges to policy makers.

e benets of new technologies can be endless, that is undisputed. Yet, all

consequences need to be considered, with careful thought and analysis being a

must.1 One technological advancement that is beginning to require attention is

articial intelligence (‘AI’). AI has been a common science ction trope for many

years but it is becoming increasingly integrated into everyday life as well as more

and more advanced.2 As AI continues to progress, the processes it engages in

require less and less human input. As a result, the question arises: what happens

when an AI ‘creates’ a nished product and more importantly who owns that

creation; and who can enforce the rights, if any, that follow from this hypothetical

creation? While these questions at one point in history were merely academic, this

is now a real intellectual property law consideration.

is article will analyse these questions aer addressing a numb er of key preliminary

points. Part II will examine what an AI is for the purposes of this discussion, and

attempt to ground the language that surrounds this developing area of technology.

Part III will address some underlying rationales for having copyright protections

at all. Part IV will give a summary of copyright law in Ireland and address some

elements unique to the questions concerning AI and copyright. Part V will see an

analysis of the potential ownership of AI created products. Part VI will seek to oer

some recommendations and address the conclusions of this discussion.

* LL.B. (Dubl.), Solicitor.

1 In an open letter from the Future of L ife Institute entitled ‘Research Priorities for Robust and

Benecial Articial Intelligence’, it is noted ‘[ b]ecause of the great potential of AI, it is important to

research how to reap its benets while avoiding potential pitfalls’. is letter has been signed by over

8,000 signatories including the late Stephen Hawking.

2 For an interesting discussion with Ian Goodfellow, the creator of the GAN, or ‘generative adversarial

network’ where AIs work against each other to further learn and improve in their performance

of tasks see an-whos-given-

machines-the-gi-of-imagination/> accessed 16 April 2020.

2  

In this article, the use of ‘product’ when addressing the creations of AI is used to

prevent confusion between the hypothetical pieces created by an AI and the legal

term ‘works’, which when used in a copyright related context has a clear legislative

denition. ‘Create’ is used in the general sense of ‘to bring into existence’.

II. What is AI

e average layperson’s understanding of AI is typically rooted in science ction.

AI has been a common trope in media, with one of the earliest explorations of AI

in ction seen as far back as the 1860s.3 Nevertheless, AI is very much a reality.

A simple denition of AI is ‘a system capable of performing tasks that would

normally require human intelligence such as recognition, decision-making,

creation, learning, evolving, and communicating’.4 e notion that a computer

could learn beyond its programming is not a new or novel concept; Alan Turing , a

pioneer in computer science and part of the team that deciphered the Enigma code

used during the Second World War, stated that AI learning would ‘be like a pupil

who had learnt much from his master. When this happens, I feel that one is obliged

to regard the machine as showing intelligence’.5

Russell and Norvig reference eight dierent denitions in their discussions on

AI.6 ey are grouped under four main headings: inking Humanly, inking

Rationally, Acting Humanly and Acting Rationally. inking Humanly in this

context means anthropomorphising AI, and seeing it as a machine with a mind.

inking Rationally means viewing AI through the lens of being able to perceive,

reason and act accordingly. Acting Humanly is concerned with AI doing tasks that

humans normally do. Acting Rationally sees AI as being intelligent things; while

intelligent, it is still a thing or an object.

With regard to how an AI operates, there are inputs and outputs. As summarised

by Poole, there are a number of relevant inputs and they will vary depending on the

task the AI has to carry out.7 One example used is an infobot, which has the task

of extracting information from a variety of sources which includes the internet and

encyclopaedias. e input is a request to act a certain way or use certain things in

its task. Some inputs that would be needed include:

3 Samuel Butler, ‘Darwin among the Machines’ e Press (Christchurch, 13 June 1863) examined the

notion of ‘machines’ evolving beyond what we can perceive.

4 Shlomit Yanisky-Ravid, ‘Generating Rembrandt: Articial Intelligence, Copyright and Accountability

in the 3A Era – e Human-like Authors are Already Here – A New Model’ (2017) Michigan State

Law Review 659, 673.

5 L ecture to London Mathematical Society on 20 February 1947. uringarchive.org/

browse.php/B/1> accessed 16 April 2020.

6 Stuart J Russell and Peter Norvig, Articial Intelligence: A Modern Approach (3rd edn, Prentice Hall

2009) 2.

7 David Poole, Alan Mackworth, Randy Goebel, Computational Intelligence: A Logical Approach

(OUP 1998) 15.