Anthropocentric
The lecture, entitled The Complex Intersection between Copyright and Artificial Intelligence, heard that a clear dichotomy is emerging between anthropocentric approaches, as in EU legislation, and certain common law jurisdictional attitudes to computer-generated works.
IMRO director of legal affairs and regulatory compliance Barry Scannell said that originality is central to copyright and the concept of intelligent computers.
“The big intersection lies in the idea of originality and creativity,” he said.
He described AI as any technology which processes and analyses data to look for patterns, and based on those patterns, makes a prediction.
AI is just a tool, despite its sophistication, he said, and originality and creativity cannot be replicated, because they are uniquely human.
Fourth industrial revolution
However, the fourth industrial revolution is driven by big data.
“We are swimming in an ocean of data and AI needs data to survive,” he said, pointing out that the boom in AI is because of the boom in data.
“We need the tool of AI to go through data, but we need data to train AI,” he explained.
When data is fed into AI tools, a human selects the data training sets, he pointed out.
AI selects suggested viewing on YouTube or Netflix but AI reflects the personality of the person who trained it, and also their bias.
This creates an ethical responsibility to ensure, from a cultural perspective, that bias is not introduced.
A clear example was the AI bias in A-Level estimated grades in the UK, which gave better marks to “better” schools.
“We have to ask, was the AI broken, because ultimately the AI accurately reflected the inequality in the British educational system?” Scannell said.
“There is very much a foundationally nurturing element of human involvement in the creative process of using AI as a tool to generate artistic creations,” he said.
In 2014-15, humanity created more data than had ever been created in human civilisation up to that point, he said.
Massive data-production
“We are living in an absolutely paradigm-shifting era of massive data production, and we need AI to be able to navigate that,” he said.
He cites ‘Amara’s Law’ – with all new technologies, the short-term implications are grossly over-estimated while the long-term implications are ignored.
Scannell is currently a UCC PhD candidate investigating AI and music authors’ rights. He also sits on the Law Society IP and Data Protection Law Committee.
The Anglo-American model, including Ireland and the UK, differs from the EU one, because of the history of copyright law in the US.
The US was more concerned with maximum social utility and disseminating information while in continental Europe more weight was given to the notion of individual creator, which gave rise to author’s rights and moral rights.
A European Commission report has said that fully autonomous creation by AI does not currently exist, and will not exist for the foreseeable future.
The commission also asserts that current EU copyright laws are sufficiently flexible to deal with the challenges posed by AI-assisted output, so there is no legal paradigm shift envisaged.
As an indicator of the strategic importance of AI, the five largest IP offices in the world have co-operated in trading views on its evolution – the European Patent Office, the US Patent and Trademark Office, the Japanese Patent Office, and the Korean and Chinese offices.
'New electricity'
AI is often referred to as the “new electricity” for its transformative qualities, the webinar heard, and its use has seen exponential growth of 270% in the past four years.
Legal research, discovery and key processes in litigation, as well as the automation of legal documents, will all be achieved much more effectively with the use of AI.
IMRO adjunct Professor of Intellectual Property Law at the Law Society, Dr Mark Hyland said the interface between copyright and AI blends key themes of technology and law with a strong political dimension.
The area is the subject of much public consultation as AI becomes ever more ubiquitous and omnipresent in our lives, in the shape of personal assistants such as Siri and Alexa, drones, GPS, and driverless cars.
AI in the shape of big data and algorithms are a key part of the fourth industrial revolution which is the confluence of disruptive new technologies and computing, and is a key part of computer science.
Definition
But it is crucial to remember that there is no universal definition of AI, he said.
Copyrighted intellectual property refers to creations of the mind, such as inventions, artistic and literary works, and designs, symbols, names and images used in commerce.
The four classical groups that are protected by copyright are literary, dramatic, musical and artistic works, but it also protects sculpture, software, maps, databases and technical drawings.
Most copyrights last for the life of the author, plus 70 years.
The principal IP rights are legal enforceability; copyrights; patents which protect inventions, and trademarks which protect distinctive signs and logos; and design rights which protect the aesthetic aspects of a product.
The World Intellectual Property Organisation (WIPO) concedes that there is no universal definition of AI but posits that AI is aimed at developing machines and systems that carry out tasks considered to require human intelligence, with limited or no human intervention.
Human intelligence
The UK Intellectual Property Office states that AI constitutes technologies with the ability to perform tasks that would otherwise require human intelligence such as visual perception, speech recognition, and language translation.
The European Commission states that AI systems display intelligent behaviours by analysing their environment and taking action with some degree of autonomy to achieve specific goals. It also says that data and algorithms are key elements of AI.
An important AI definition, set out by a European Commission high-level expert group, is not user-friendly in that its first sentence runs to 72 words, Dr Hyland said.
EU legislation and case law is more anthropomorphic, or human-centred, than in common law copyright regimes which accept the notion of computer-generated original work, Dr Hyland said.
Dr Hyland pointed to this year’s BT Young Scientist competition winner Gregory Tarr, who used AI to detect “deepfake” videos.
Contours of the law
Professor Jonathan Griffiths of Queen Mary University in London said that the extent to which the products of AI can be protected in IP law presents the most interesting and fantastic challenge to the contours of the law.
He said a consensual view is emerging that AI is a tool that, while it is technologically revolutionary, it doesn’t necessitate a paradigm shift in copyrights.
“There are clear advantages if that is in fact the case. Firstly, it doesn’t mess with the structures of the law on which commerce is based.
“It doesn’t rush in quickly to deal with technology before we quite know how it’s going to play out.”
Policy disputes
The creation of new IP rights would also involve difficult policy disputes.
However, Professor Griffiths said some of the problems in absorbing AI outputs into copyright law may be brushed under the carpet.
Significant editing work on an AI output is broadly comparable to arrangements or adaptations of creative works which give rise to copyright under the law, he said.
“AI outputs that have had significant end-of-process work by human beings would clearly be entitled to copyright,” he said, offering the example of a weather report which is generated by AI inputs but effectively arranged for presentation to the world at large, by a person.
The choice of data input for AI could also be described as creativity, he offered, though such “frame-setting” is not generally rewarded with copyright.
While the accommodation of AI within copyright is convenient, it risks diluting the coherence of the concept of authorship more generally, he said.
Scope
Recent UK judgments have been problematical in that they expand the potential scope of the people who can be regarded as authors.
So refusing convenient answers to AI might mean devising related rights to cover AI outputs, in the shape of a tailored publication right.
Professor Griffiths said that, at this stage of technological development, there’s a reluctance to introduce new IP rights. Because of fundamental rights to property, it’s very difficult to get rid of IP rights, even if they prove not to produce the hoped-for results.
An array of related rights might do the job without calling upon copyright, such as the protection to cover sound recordings, and films.
“Much of the output might qualify for sui generis database rights,” he suggested.
An appropriate solution in the context of copyright doctrine might be the use of the concept of a “computer-generated work”, he said.
Distorting authorship
This would avoid distorting the concept of authorship, and avoid attributing entitlement to inappropriate parties.
It distributes entitlement to the business within whose auspices the work has been produced.
“I think that is a lot more promising,” he said.
“It’s an interesting idea that post-Brexit, it may be in Irish law only that the seed for an appropriate solution to this problem is to be found.”