On the Potential, Shortcomings, and Other Aspects of Writing with Artificial Intelligence On the Potential, Shortcomings, and Other Aspects of Writing with Artificial Intelligence On the Potential, Shortcomings, and Other Aspects of Writing with Artificial Intelligence

To address this question, I launched two rounds of “Writing with Artificial Intelligence” research in 2018 and 2021 with the support of the National Arts Council in Singapore.

In the second round of research, I established a creative framework and invited four human writers to each create a script, while I used AI models to create scripts based on the same premise. I then put together these eight scripts for a table read, during which neither the actors nor the audience knew which scripts were created by humans and which were created by AI. Afterwards, the audience and actors voted and held a discussion on AI writing.

If you are familiar with the history of AI, you can see that this research borrowed the concept of the Turing test. The Turing test, proposed by Alan Turing, tests whether a machine can exhibit human-like intelligence. In the test, a tester interacts with both a human and a machine, and if the machine can deceive the tester into believing it is human more than 30% of the time, it is considered to possess human-like intelligence.

Of course, passing the Turing test is not the ultimate goal for artists and artistic creation. What interests me more is how AI will impact human creativity. Will humans be able to collaborate more deeply with new technology? How can artistic creation challenge the boundaries of technology?

Scriptwriting is a process of creative writing, which in some ways, is a process of anti-system and extreme individuality. Therefore, it may be a false proposition to try to program the creative process of writing. The road for AI to carry out scriptwriting may still be a long one.

Firstly, there is a need for understanding the artistic form. For scriptwriting, in particular, a thorough understanding of the dramatic vocabulary such as characters, dramatic actions, events, and conflicts is necessary for traditional plot writing. Currently available AI models do not possess a comprehensive understanding of this. If it involves theatrical performances, AI even requires a deep understanding of spatial structure and its application, and in the case of TV and film, AI also needs to understand more about camera language.

Secondly, the structure is essential. Traditional plot writing requires a strong control over the overall structure, and the development, direction, and changes in the plot are very complex. It remains a challenge whether AI can well connect the context, make the plot form a complete cycle, and this is still an unsolved problem.

Based on the above two points, AI is currently unable to complete a “qualified” scriptwriting independently. In my experiment, researchers and audiences repeatedly pointed out that compared to scripts written by human playwrights, AI scripts appear “incoherent.” It still relies on the cooperation of human scriptwriters, or human scriptwriters can use AI to assist in their creation. In other words, the key phrase in the issue of “Writing with Artificial Intelligence” is not “artificial intelligence” or “writing”, but “with”.

So, what are the advantages of human creators over AI in writing?

Firstly, the establishment of a symbol system. Compared to AI’s “non-intentionality” in creation, human creators are clearly conscious and intentional in their creative process – to some extent, it is the establishment of a symbol system. AI’s selection of elements is based on data rather than literary quality. In contrast, human scriptwriters are better able to consider the potential and space for symbol interpretation when selecting elements, which is often strategic, more rational, logical, and coherent.

Secondly, the ability of self-reflection. Human scriptwriters can read their own writing, generate criticism, and make modifications; or listen to others’ opinions and suggestions, filter and absorb them. Moreover, humans can create works with “meta” settings, such as writing a play about scriptwriting. Can AI create reflective plays? There is currently no answer. But human reflection and ability is what constantly drives us to progress and develop.

Thirdly, involvement of other elements. Human scriptwriters can have a more comprehensive understanding of artistic forms, consider more effects of script presentation, and the integration with other elements. In contrast, scripts written by AI tend to lean towards literary texts. Although AI models can understand other elements with appropriate guidance, it is not like humans who can have real experience with theatre or film operations, so it’s only talk on paper .

Lastly, as long as there is data, there is a data centre; and as long as there is a data centre, there is an ideological and capitalistic composition behind it. In my workshop, a student discovered that the dialogue written by AI was “very Hollywood” – we can infer from this that certain AI models have a certain data composition behind them. Whereas humans are diverse, and human self-awareness can bring us possibilities for reflection.

Art has always responded to the era and society we live in, including the technological developments of this era. The invention of electricity has changed the creation and form of theatre, and the invention of the Internet has once changed the research and creation of creators…Technology is always gradually influencing how we think, live, move, and create. Perhaps technology and art are expanding the boundaries of human creativity in this game.

As a committed practitioner with a desire to push artistic boundaries, Liu is regarded as a promising figure at the forefront of the experimental theatre scene in Asia. He received the Young Artist Award awarded by the National Arts Council of Singapore in 2016.