THE WRITER’S LITTLE HELPER: TO CHAT-GPT OR NOT TO CHAT? That is Now a Key Question for Writers Everywhere
Our guest columnist, Dr. Angus Finney, explores the developing AI tools in script writing and analysis, inviting various industry opinions.
As someone who has worked on more than 200 hours of scripted television series, Mark Goffman is no stranger to receiving “feedback”. So when the top writer-producer (The Umbrella Academy, Bull, Limitless) downed tools in the early days of the WGA strike, he reached out to a new AI script tool StoryFit, in a mission “to learn everything I could. I did an early test and wondered what the feedback would look like.”
SCRIPT COVERAGE & ANALYSIS WITH AI
StoryFit is a company that uses AI to mine and collate data on a wide range of storytelling elements within feature and TV scripts. According to founder and CEO Monica Landers, some writers are initially fearful and find it hard to accept, let alone digest and act on, but Goffman found himself “pleasantly surprised” by the way the feedback was presented.
“This is not traditional script coverage. StoryFit is systemized, consistent and data-point driven. It gave me specific analysis on, for example, what characters cared about most, their key dilemmas, and emotional styles. The relationship page was particularly helpful – providing a check on how often they appear with other characters, and for how long. There were no prescriptive suggestions, it just showed me what it found.
“You have to consider that analysis and data, and work out what now? What do I change?” My experience is that StoryFit is pretty objective if used properly, but the key is that the evaluation model is helpful if we all get to share in that knowledge. And that’s the rub. There is an understandable fear around when information is not shared, and creative writers and talent may find themselves operating in the dark.” Goffman feels strongly that core understandings around what he calls “fair dealing” are going to be needed to move us all forward in this new age of AI.
Other leading writers are also concerned about the grey areas and wider issues at stake. Some feel that the three-year WGA settlement feels “like a plaster on a massive great wound,” in the words of twice Emmy award winning UK writer-producer Ashley Pharoah (Life on Mars, Around The World in 80 Days). “Studios and streamers can explore with producers how to use AI, but they cannot generate a writer or a great script from AI right now. Of course, it will improve, month by month. But the building blocks are human. There is a lab rat instinct about testing it out – and I found it rather stiff when I explored – but it’s important to consider the more profound, existential issues that are now at hand.”
On the one hand, (and I quote an award-winning writer who requested anonymity here) the “creative business corporations”, from Studio-streamers to mini-majors through to their commissioning editors, could “feed script material into AI data machines, to throw up the analysis, budget and predictive numbers, and lazy accountants decide on a pass, hold or a greenlight.” On the other hand, much of the existing fear among creative writers “exists before they have actually fully explored the opportunities.”
GROWING AI TOOLS AND VARIED INDUSTRY OPENNESS
One of the most interesting companies that has long been helping writers and producers explore the opportunities is a Swiss-based operation founded and led by Sami Arpa, called Largo.ai. According to Arpa, Largo began with a strong focus on predictive script analysis, categorizing scripts into distinct cinematic DNAs for genre, pacing, character dynamics, casting implications, and anticipated audience reactions.
However, as the film industry's openness to AI has grown, so has his company’s approach. In the past year, Largo has shifted from merely categorizing content to delivering actionable insights, moving beyond neutrality to provide evaluations that speak to the quality of content with a broader range of analytical voices.
The analysis starts with the raw materials of creation: treatments, screenplays, and even completed videos. When it comes to financial or cast predictions, Largo also factors in production parameters like budget, director, and producer profiles. The AI program has been trained on an extensive library of over 400,000 pieces of content, encompassing films, series, and other audio-visual works. This vast dataset educates its system on genres, audience engagement, and the unique cinematographic DNA of actors, enabling a nuanced understanding of potential market performance.
A is candid that not all writers and creative producers have positive reactions to AI script analysis. “Some embrace the technology as a tool for enhancing their work, using AI feedback to refine their scripts before pitching them to studios. Others may be skeptical or feel that creativity should not be constrained by algorithms. While some leverage our AI as a vital enhancement tool, fine-tuning their scripts to better resonate with the audience, others are concerned that creativity could be stifled by algorithmic influence.
“Resistance often arises from fears of diluting originality and undermining artistic integrity, with concerns that AI could lead to a homogenization of storytelling. But we advocate for using AI as a collaborative, not prescriptive, tool.” Largo's philosophy is that AI should serve as a complement to human creativity, not a replacement. This stance has helped maintain our early adopters, who benefit from improved narrative crafting, streamlined editing, and more successful pitching, all leading to sustained engagement with our tools.
Early adopters who find value in AI feedback and can integrate it effectively into their creative process are likely to continue using it. The 'stickiness' depends on the tangible benefits they experience, such as enhanced storytelling, more efficient editing processes, or increased success in pitching scripts.
Other commentators and users of Largo point out that a particularly useful feature is the genre analysis of a script across the running time, visualising the level of e.g., drama, thriller, horror it finds, and how these change across the running time, allowing users to compare a genre analysis of a much loved classic, or other reference film they want to improve upon.
“Exercises like this remind us of how important the writer’s voice is. Authenticity is subjectivity, and no matter how hard the industry tries to create the formula for good television and film, originality is the only thing that endures. ”
A resonant example of a leading author and screenwriter “playing” with AI and sharing his experiences in the public domain, was triggered when a father started co-writing with his son following a severe injury impaired Hanif Kureishi (Buddah of Surburbia, The Mother). Taking co-piloting to a two-tier level, Kureishi and his son Sachin initially used ChatBot GPT4 to help develop ideas for stories. The initial result was a stream of cliches, stereotypes and sentimental Hollywood endings. “It was like interacting with an intelligent child; it impressed me how much they knew for its age, but it couldn’t really help me.” Or could it.
“Then I started treating it like an adult. I began giving it more precise instructions, challenged it if I didn’t like what it was giving me, asked it to be critical of my ideas, to appraise them like an editor. It learns as you teach it, responding to the specificity of the input.” Much of the experimentation didn’t necessarily flesh out, but lots of ideas are what Kureishi states “we need in the beginning, world-building stage. In this dark wood, the chatbot is like a little torch, point us in various directions to go and explore.”
But for all the prompts and co-piloting in the world, Kureishi is clear that right now, and possibly for a very long time to come, AI is just a guide: “Exercises like this remind us of how important the writer’s voice is. Authenticity is subjectivity, and no matter how hard the industry tries to create the formula for good television and film, originality is the only thing that endures.”
Which makes the insight I heard back in March this year in LA (pre the WGA settlement guard rails), from a leading Hollywood genre-driven producer, all the more alarming. We were bouncing around ideas about the levels of utility that might be on the march with AI across the film industry, when he turned round and stated: “But’s what’s so great is we can commission AI to do the first draft, and then the writers can do the polishing. Just think of the time and money we’ll save…”
Not so fast. Not only has the WGA agreement placed a marker on any such move by the Studio-Streamers, but script editors, teachers and dramaturgists remain both intrigued but wary of over-hyped claims that AI-driven Large Language Models (LLMs) can generate outstanding filmable material. What the majority do agree on is the importance of writers and creative producers accepting and learning how to co-pilot and prompt AI language models. But even with a mastery over the ‘art of the prompt,’ there are numerous pitfalls awaiting.
PITFALLS
“We need to be aware of the tight rope that we are walking,” explains Carl Schoenfeld, founder of the Oxford University Innovation incubator Online Screenwriting Academy. “AI programmes can suck you into the cliché without you seeing it, and it demands the human to flex their sentient muscles in order to benefit from AI. A level of cognitive awareness is required, so that we better understand how we do what we do, in contrast to the machine’s work.”
Other experienced script editors concur. “I see writers getting distracted by GPT and there is a danger of over thinking and over using the tools, and there’s an element of laziness at times,” says LA based script editor Sandford Galden-Stone. “GPT generated results remind me of a young student writing an essay in a way that they think a teacher wants to read. It’s not about expressing a personal point of view, it’s about ‘getting the right answers.’ Thus the output is dry and stilted and not particularly creative.
“The craft of writing is about making choices at every point from beginning to end, providing an emotional experience for the reader and/or audience. If you use AI then it needs to be a balancing act between the benefits of a tool on the one hand and your personal instincts on the other. But ultimately machines are only as good as the user.”
And not all are convinced that writers will always be ‘good’ around the WGA settlement and its rules that writers need to be open about their relationship to and direct use of AI. “The rule that writers need to be transparent about if and when they use GPT (or any AI for that matter) was stupid”, states one of Germany’s leading dramaturgists Martin Thau. “They will use it and won’t tell anyone, as it simply devalues their work and they won’t be taken as seriously.
However, ChatGPT’s strongest feature seems to be in the earliest, rougher and time-consuming stages of story mapping, explains Schoenfeld. “This bears the question, once the digital work has been manually expanded into a first draft, if there are any traces of digital origins?” Thau’s response is that there will be “new courses where a coach will work with the writers on how to use GPT without leaving a footprint and having to disclose that fact. Writers will develop their own useful and unique prompts, that they will use on their own material, and they won’t have to disclose.”
“The wider creative writer’s phobia is that they won’t be needed anymore. In fact, it’s probably the other way round. Those who think they can produce a full film within a computer will quickly fail. It's the writer’s ability to choose that is defining. There is always more behind our sub-conscious that writers have the ability (or at least potential) to explore. Violence, sexuality, humour…all the way to tragedies. Indeed, can a language model even grasp the concept of tragedy? Building those scenes, with a sense of danger and enrapturement at the same time? You need a writer for that!”
MOVING FORWARD: HARNESS THE POWER OF AI
Which brings us back to the talented Mr Goffman: “There is a lot of fear out there – some justifiable, but much of that fear exists without writers fully exploring the opportunities. But scenes, dialogue and structure can be played around with [via AI programmes] in a way that allows us to open doors significantly faster and more efficiently. Everyone has different writing methods – no one programme will be and remain the same. But we need to evolve and change as the tools develop and approach it as an interactive exercise. I see [AI’s offerings] as an additional tool that does not replace but adds to the kit.”
Goffman’s recommendation to learn and harness AI is supported by BAFTA-winning screenwriter Tony Grisoni, the vastly experienced British writer who has co-written a number of director Terry Gilliam’s films. “We all need to catch up. AI is here and has been for what are digital centuries. Learn it, harness it, exploit it, subvert it - or whither.”
On Artificial Intelligence: Excerpt from the Summary of 2023 Writers Guild of America (WGA) Theatrical and Television Basic Agreement
We have established regulations for the use of artificial intelligence (“AI”) on MBA-covered projects in the following ways:
AI can’t write or rewrite literary material, and AI-generated material will not be considered source material under the MBA, meaning that AI-generated material can’t be used to undermine a writer’s credit or separated rights.
A writer can choose to use AI when performing writing services, if the company consents and provided that the writer follows applicable company policies, but the company can’t require the writer to use AI software (e.g., ChatGPT) when performing writing services.
The Company must disclose to the writer if any materials given to the writer have been generated by AI or incorporate AI-generated material.
The WGA reserves the right to assert that exploitation of writers’ material to train AI is prohibited by MBA or other law.
Dr. Angus Finney is an executive producer and Fellow at Judge Business School, Cambridge University