‘Horrifying and incredible’: one writer’s argument for making use of AI

17 June | The Spinoff

Books editor Claire Mabey talks with Emily Broadmore, who is behind a contentious AI workshop for creative writers, to try and understand the why of it.

On Friday June 6, the Wellington Writers’ Studio held an AI workshop run by Heft Communications (a PR company based in the same building). “This isn’t about using AI to write,” said the blurb on Instagram. “This workshop is about learning about how to super charge the editing, research and refinement process of your writing, safely, with LLM [large language model] AI.”

When the post first went up the notice attracted many comments expressing dismay and anger. There were sad face emojis, aghast emojis, as well as messages of outright rage that a writers’ studio would open up the craft to AI interventions. The comments were turned off and the workshop went ahead.

As someone so existentially threatened by the thieving, ravenous, deceptive nature of AI that I can’t fathom getting into a relationship with it at all, I wanted to ask Emily Broadmore, the founding director of Heft, why she finds it necessary for AI to be invited into the realm of creativity and communication. Because I feel I’m on the losing side of the AI debate, I wanted to hear her out.

Broadmore’s answers suggest that there is, so far, an active division between the application of AI to assist writing in a workplace context (for example, in the communications and marketing sector), and the application of AI to assist writing in the context of literary craft and creative pursuit. What Broadmore told me about AI’s ubiquity in her sector leads me to conclude that hope is now pretty much lost for the workplace.

But can art – man-made, beautiful, blood-sweat-and-tears craft – be held aloft from the march of the human-fed robots? To me, the idea of using AI to influence creative writing (novels, poetry, screenplays, picture books, articles, essays – any creation formed from your own imagination and skills and perseverance) is anathema to art: which is surely about testing and expanding the miracle of a human mind; and to transfer that process-driven effort to another in order to connect and converse on a profoundly human level. A level that, surely, an iterative robot can’t feel or understand, or compete with. It seems to me that the more AI is used, the more valuable human-crafted, non-AI processes must then become.

And yet the boundaries are worryingly blurry as usage is rapidly normalised without any real sense of the rules. I know I will never converse with AI – I love the creative process too much, even the brute, boring, time-consuming bits; and have no interesting in feeding the beast. Broadmore –herself a writer and publisher – feels otherwise.

Am I just suffering from an inability to move from the typewriter to the PC? Read on and decide for yourself.

Claire Mabey: Before the comments were turned off, I saw on Instagram that there were a lot of concerned, angry messages about the fact you were holding the workshop. Why do you think it attracted such comments?

Emily Broadmore: So many reasons, Claire. I was at a conference the other week which was focused on energy. The conversation there was all about the energy consumption necessary for AI. There’s also the fact that the way that the models have been trained is, you could say, unethical.

There are so many problems with the way this technology has come about. The workshop was intended to help our community and to start debate and discussion and actually to raise these issues so we can all talk about them.

CM: How many people attended the workshop?

EB: Maybe 16 or so?

CM: What kind of things did you end up discussing?

EB: People are using LLMs to help them speed up the implementation of work. We wanted to open that up to the writing community.

The first time I realised that there was this massive crossover [between the use of AI in Broadmore’s work in PR, communications and marketing at Heft; and the creative writing world] was when I had this manuscript that I started pre-Covid and was stuck on where to go next with it. Have you ever done a manuscript assessment? Do you remember how much it cost?

CM: Yes I have. They’re variable – and cost anywhere between $200 and $2,000.

EB: So my first manuscript was reviewed by a wonderful author that I know; a professional assessor. And it costs a lot of money, right? I’d been watching what was going on in the comms sector and had this moment of exploration; and I felt sick doing this but I loaded my manuscript into a closed model – one where I’ve got a professional subscription, and I made sure that it wouldn’t be used for training and it wouldn’t share it.

I asked for an assessment of the manuscript on various points, and what came back was both horrifying and incredible. It was a more thorough assessment – and in the space of 10 seconds – than I had ever seen with professional manuscript assessors.

One of the other things that I found has really sped up my process is that AI can help me with research. I’m writing a book set in New York and I haven’t been there in years. So I ask the AI things like: “What would a journalist from Washington Post who’s in her early 30s earn in New York in 2021?” The AI speeds that research up and leaves me with more time to write. I don’t see that as cheating.

One of the writers in the studio is doing some really interesting stuff. He said he was loading his manuscript in sections: like he would load it when there was a plot point, where there’s a twist, or some sort of reveal. He’d asked the model to guess what was going to happen next to make sure that his writing wasn’t doing anything too obvious. I thought that was really clever, especially if you’re working in detective or mystery, right?

So we’re all just exploring; but this is basically what we got together and talked about at the workshop.

CM: What about the tension that arises when people use AI instead of human labour? What about the professional assessors who might lose their work to AI assessment?

EB: Are there many of those, though? The few people that I know that do manuscript assessments are so picky about who they take on. Ultimately if you’re serious about your manuscript, you want that human input. So I don’t think it’s replacing it. It’s meaning that when you’re actually asking for that human to apply their mind to your piece, you’ve got it as good as you possibly can.

But, yes, look, it is fraught. That’s why we’re doing so much work to help people. I saw some stats last week: PWC New Zealand was basically saying people with AI skills are getting an increase in income, and that other jobs and other roles are opening up. This is what happens whenever there’s a big revolution in technology. But I’m not an expert.

I had a conversation with Jane Friedman – who reports on the publishing industry – who said that AI is moving very quickly to support publishers to work through their slush piles: she called it slush pile management. Now, I can see this from both sides: because as a publisher [Broadmore is founder and editor of Folly Journal], on one hand you’ve got 98% of submissions that are completely inappropriate for what you’re trying to do; but on the other hand you don’t want to miss anything that you might really want.

Jane saw the AI development to manage slush as a positive thing: for example, think about how long manuscripts can sit there without being looked at – AI help means that you might hear back sooner, or you might hear back sooner with feedback.

Jane thinks New Zealand’s probably a couple of years behind in terms of our acceptance of AI.

CM: Do you think New Zealand should be more accepting of AI in our creative industries?

EB: We don’t need to be. Creativity is a human art form. People get so tribal about the ways that we do things, and I don’t think it’s helpful.

I saw one piece of research from America that said that 60% of writers are using AI. Maybe there are people who would rather just sit there on Google and figure it out the old fashioned way [Broadmore was referring to research]. But I don’t think it’s helpful to say that we should be more accepting.

CM: One of the reasons there’s anxiety about AI is that it is unregulated and there seems to be this inevitability that AI will get increasingly competent the more it’s used: there are already news stories out there suggesting that AI can write a passable book. Was anyone at the workshop experimenting with using AI to write, as well as assist with research?

EB: I’d assume that the only people that would use it to write are people that are going straight to ebook commercially. I don’t know much about this industry, so I’m probably the wrong person to talk to, but there are people that write to the market.

From my own perspective, I can’t imagine someone like you and me doing that. Writing is so hard, but we do it because we love writing.

CM: You want to discover the capacities of your own mind, and craft, don’t you?

EB: Yes, so why would you want a computer to do it? I haven’t met a single writer that would actually want to own up to not wanting to do the writing.

CM: So, is it fair to say that one of the reasons you’re doing your workshops is to acknowledge that AI is here?

EB: It’s here. We can’t stop it. It’s changing my industry. Yes, there are people out of work, but what I’m working on right now is the young people. It’s the new grads who aren’t getting hired so how do we get those people into the workforce and train them when they aren’t needed any more because of AI? That scares me but I can’t stop it, so where are the good bits?

CM: What kind of things are you doing to try and help the new graduates?

EB: Remember when you and I were young, we would get given grunt work?

CM: We’re still young.

EB: We’d get asked to give a press release a go, or drafting an opinion piece. That doesn’t exist any more. Our clients are using AI. Everyone’s using AI. There’s an expectation we are using AI.

I’ll be completely honest with you. We had our graduate employee leave last year, and I didn’t replace her for six months because AI was doing everything she was doing faster and better. Then I felt really guilty, and I hired two interns. They’re paid above the living wage here at Heft and the reality is we don’t necessarily need them, because AI can do all that stuff. But ethically, taking on young people and training them is going to become like a moral and ethical thing for businesses.

So to your question, what are we doing? We’re redefining grad roles. If you’ve got a young person in the office now, you’re not asking them to draft a press release from scratch. The AI is going to draft the press release from scratch. What you’re asking the young person to do is apply judgment, critical thinking and to understand the brand and the brand voice. So they’re actually being put into a management role. They’re providing judgment of the outcome. So the work that they are doing is looking at content and critiquing it with a red pen, rather than being the person creating the content and being critiqued. It’s completely flipped around.

There are so many issues with this: young people haven’t fully formed their ability to critically analyse work. They haven’t got their 10,000 hours of experience yet. What that means is that we’re in a Socratic dialogue with them, where the conversation isn’t showing them what they’ve done wrong in terms of where you need to work a bit more. Instead, the conversation is: “What do you think’s wrong about this piece of work? Does this reflect the client’s voice? Is this actually her opinion, or was it just made up? Can you go and check that?”

It’s giving young people way more clout. But it requires a whole new way of thinking.