Works ‘created’ by artificial intelligence (AI) cover a range of creative mediums from replica Rembrandt portraits to Shakespearean poetry – but could AI ever write a play with characters, scenes and plot development? New AI-assisted technologies are assisting directors and performers in developing new opportunities for live theatre and performances.
Virtually Immersive Improvisation
A recent study by the University of Kent considered the way video-collaboration tools can replicate the immersive experience of rehearsing and performing with other actors.
Video conferring systems such as Zoom and Microsoft Teams are predominantly designed for business or educational purposes and the study concluded that performers found it difficult to fabricate a world when they were limited by the background of their individual homes. Through the use of ‘Virtual Director’, software which isolates and positions individual video feeds within a 3D environment, multiple performers could engage and improvise in real time exclusively through their screens.
The study focused on improvisational theatre where actors invent new realities on the spot through audience suggestions and by reacting to the real-time behaviour of other performers. This style of performance presented a particular challenge for digitisation as improvisational theatre relies predominantly on collaboration and an awareness of other performers’ verbal and physical cues.
The use of augmented technology allowed for unique opportunities that would not be available in real life such as dynamic backgrounds for scene changes (which responded to the unfolding scene). For example, when a performer indicated they wanted to leave the scene, they would be moved off the screen and then appear in another virtual location.
The performers reported a strong sense of co-presence and physical connection with their colleagues despite being remote, and that they were able to enter a creative flow state. Unlike the compartmentalised user interfaces of ordinary video conferencing systems which require users to split their focus between different windows, Virtual Director placed the performers inside shared virtual environments. This minimised visual distractions and increased visibility of other performers’ non-verbal reactions.
AI has been used in live theatre shows to assist with generating scene ideas, such as Improbotics which debuted in London in 2018.
The improvisational show uses an GPT-3 chatbot to generate and send lines of dialogue to one of the performers through an earpiece which the other performers then incorporate into the scene. The audience typically will not know which performers are ‘controlled’ by AI and which are speaking of their own free will.
The chatbot is based on HumanMachine and is trained on movie dialogue from OpenSubtitles to complete sentences using similar language and topics. By combining human and machine-generated lines, the narrative arc could continue and develop to the surprise of both the audience and the actors.
THEaiTRE, an interdisciplinary research program, has taken AI scriptwriting a step further by using AI to write an entire play. “AI: When a Robot Writes a Play”, which premiered in February 2021, is a play about a robot experiencing everyday emotions and was intended to celebrate the 100th anniversary of Czech playwright Karel Čapek’s play R.U.R, which contained the first use of the word “robot”.
Using the GPT-2 language model by OpenAI with adjustments to assist in generating theatrical scripts, the writers would input a short description of a scene, characters and dialogue and the model would generate 10 lines of dialogue.
The project adopted a human-in-the-loop concept which encouraged the co-operation of machine and human efforts throughout the creative process to ensure that the generated script was coherent and was not too repetitive or contradictory. This design process also ensured that the script made sense for performers, for example by checking that there was not dialogue for a character who had already left the scene.
The end result was a script that was 90 percent automatically generated with only 10 percent written or edited by humans.
Copyright protection of AI-generated works
Dramatic works such as scripts and plays can be the subject of copyright protection provided certain criteria is satisfied.
The Australian approach to copyright requires that a work must originate from a human author as the result of independent intellectual effort. Although AI can be used as a tool to assist the creative process, the Australian position is that in order for a work to be original and protected by copyright, there must be the exercise of sufficient skill or judgment by a human author. This would encompass, for example, the edits to the script by directors or the integration of dialogue from the performers on-stage.
However, the situation is more ambiguous when considering a work which is partially created by humans, and partially by an AI system (such as the examples above). And a play or novel or poem which is wholly generated by AI and would otherwise be the type of work which is protected by copyright, will not be given the absence of a human author.
The relationship between copyright and human authorship can be seen from the genesis of modern copyright law – the introduction of the Statute of Anne in England in 1710. It was designed to be a law aimed squarely at protecting the rights of authors in their creations, and ensuring they were able to obtain financial returns from the property they had created. From this beginning, a strong thread connecting human creativity and originality to the protection of works has carried through the law as it has developed.
However, it has become increasingly clear that the Australian position on human authorship may need to be reconsidered and may be out of step. One option is the path adopted by the UK, where the legislature has introduced a “deeming provision”. Section s 9(3) of the Copyright, Designs and Patents Act 1988 (UK) provides that:
In the case of a literary, dramatic, musical or artistic work which is computer-generated, the author shall be taken to be the person by whom the arrangements necessary for the creation of the work are undertaken.
And, s 178 of the same Act states that:
‘computer-generated’, in relation to a work, means that the work is generated by computer in circumstances such that there is no human author of the work.
A deeming provision such as this effectively creates an author for the purpose of enabling protection of a work. If such a provision was adopted in Australia, a copyright work such as a play, a novel or a poem would still (as a Part III work) need to be original, regardless of the deemed human author. This would mean that not just any computer-generated, or AI-generated work, would be protected by copyright.
However, it leaves open the question of how originality would be assessed in the absence of the human author to whom the law has always looked to determine whether the requisite creative spark exists. To this end, an interesting quirk of the Improbotics project referred to above was that the producers found that the human-generated lines were generally shorter, contained more grammatical errors and were less complex than the lines produced by the AI-assisted technology. And while quality has never been the test for originality, what could be more human than that?
Authors: Alana Callus, Rebecca Dunn and Peter Waters
Read more: Kent Academic Repository