In the field of copyright law, the concept of authorship has generally been based upon the assumption that the author is a human being. As businesses have increasingly begun to harness, interrogate and manipulate big data, a new type of author – the machine – has emerged which fundamentally tests this assumption.
By stretching our imagination a little into the future, it is possible to map out four stages in this evolution of authorship:
|1||human being||the work is authored by a human being with no use of machines|
|2||machine-assisted||the work is authored by a human being who utilises a machine in the process (for example, Computer Aided Design)|
|3||insentient machine||the work is originated by an artificially intelligent machine based upon interrogation of data and without human intervention or input in the creative process (for example, music created by Deep Mind software)|
|4||sentient machine||the work is originated by an artificially intelligent machine which is self-aware, introspective and intends to create the work, without human intervention or input.|
The current debate around artificial intelligence and copyright authorship is primarily concerned with stage 3 and, in particular, whether a machine originator of a work should be treated in an equivalent manner to a human author? This question is not merely an academic flight of fancy but one to which the answer will have significant commercial consequences.
Plenty has already been written about the status quo of copyright regimes around the world and how they might apply to works originated by machines. In Europe, for example, the subsistence requirement of the “author’s own intellectual creation” suggests that intellect (and therefore human creativity) is a pre-condition for copyright protection of a work. In the USA, in its most recent ‘Compendium of Practices’ the US Copyright Office has confirmed that it will only register works created by a human being and that works created by a machine that operates randomly or automatically without any creative intervention or input from a human being are uncopyrightable.
If these sort of approaches are applied rigidly, works originated by insentient machines without human involvement in the creative process will be rendered copyright-free. This is an observation rather than a criticism but if such a path is followed it is not entirely clear where it may lead. What, for example, would be the long-term effect on the continued production and value of copyright-protectable works authored by human beings? We don’t really know. A human hand-crafted work of artistic craftsmanship might be valued more highly than a machine-originated equivalent and yet an increase in the supply of such equivalents might also drastically reduce the market price of the former.
Conversely, some jurisdictions already have provisions which could plug this copyright gap (if indeed it is a gap). Section 9(3) of the UK’s Copyright, Designs and Patents Act 1988 provides that the author of a “computer generated” work shall be taken to be “the person by whom the arrangements necessary for the creation of the work are undertaken”. In this context a “computer-generated” work means one generated by a computer in circumstances such that there is no human author. What this means is that the person ‘making the arrangements’ is rewarded with copyright ownership as a result of a creative process undertaken largely or exclusively by a machine.
That might be viewed as somewhat at odds with some of the more traditional justifications for rewarding creators with copyright for their own endeavours, such as those based upon Lockean and Hegelian theories, rather than receiving what looks like a windfall for the creative labour of a machine. It is also perhaps unlikely that such an approach would be adopted as part of a global consensus (for example, by amending or augmenting the WIPO Copyright Treaty to align it with this approach).
A more globally palatable answer may lie somewhere between the ‘gap’ and the ‘windfall’ such as introducing a sui generis right for machine-originated works which rewards the investment of resources – human, financial, or technical – in the machine that ultimately creates those works. While this would result in one of two different legal rights protecting the same type of work depending upon the mode of creation (human or machine), with the ultimate beneficiary a human being, it would also enable a distinction to be made between the rationales for providing that protection and the scope of it. For example, in recognition of the increasing advantages that machines are likely to have over human beings in creating and producing works (such as the speed, accuracy and cost of production) there may be a case to make any sui generis right ‘copyright-lite’, with a shorter period of protection than that available for human authored works.
Whether any differential treatment of authorship between humans and machines could be sustained or justified at stage 4 is a far more difficult question to answer, although one which is unlikely to require addressing any time soon. For stage 4, algorithms would require the ‘singularity’ necessary to be capable of learning as a human would, simulating multiple intelligence types that would make their operation almost impossible to distinguish from human behaviour. That is still something of a pipe dream in 2018.