The role Artificial Intelligence already plays and will continue to play in our lives has been on the minds of plenty of ghosts, so we asked science writer Gregory Mone, ghost of a forthcoming book on robotics & AI, to moderate a panel of experts, including Nathan Baschez, founder of AI writing tool Lex; Stephen S. Power, executive editor at Kevin Anderson Associates; and Scott Sholder, partner and co-chair of litigation practice group Cown, Debaets, Abrahams & Sheppard LLP to discuss the topic.
The panel opened with a question that cut right to the chase: Is AI going to put ghostwriters out of business?
The consensus? Not yet.
While AI is good at collating facts, Power said, writers are “putting the human in the whole thing” by “pulling out the stories [authors] don’t want to tell. That’s the real skill.”
Ghostwriters particularly bring value beyond what’s on the page. “Part of your job as a ghostwriter, and mine as an editor, is to be your author’s therapist,” Power said. “You get the stories out of the client, you put them down and have them look at them in a new way, and suddenly they remember things they didn’t remember before and bring more color to it.”
Skilled, professional writers, Baschez added, understand “what kinds of stories will be interesting to people, what kind of hooks will be interesting to people right now,” while an AI language model will only generate “the most likely thing, the most generic possible thing.”
Other types of writing and publishing industry roles, however, may not be long for this world, said Power, including copy editors and copywriters. “Publishers just want to cut costs,” he noted, and “AI will start peeling away things that are formulaic.”
But, Power reassured, “As long as you can make a reader care, you will have a job.”
Another question on ghostwriters’ minds: Will clients turn to AI instead of hiring them?
They might try — but they’ll likely be unhappy with the results. AI can “organize the information into a coherent narrative,” Power allowed, but “can it do it in a way someone will want to read it? I think, at this point, probably not.”
What’s more, there are limitations built into AI tools like ChatGPT. For example, Power fed ChatGPT a narrative in which a bear killed a scientist who was going to destroy the world, and the AI refused to write it, opting, after much urging, to have the bear defeat the scientist “using its wiles.”
Also worth considering: Is AI’s process of learning how to write even legal?
As many following AI-related lawsuits may know, large language models (LLMs) are trained using scraped data, a lot of which is copyrighted work used without permission.
While some may argue, as Mone, playing devil’s advocate, pointed out, that writers learn to write by absorbing the work of others — so why can’t AI do the same? — the argument does not apply to AI, according to Holder.
“On a superficial level, that’s how humans learn, but presumably, when they consume that material in the first place, they either paid for it or took it out of the library,” Sholder argued. “They did it with permission,” and, he added, even those who have photographic memories don’t have a copy of a work in their brain.
One potential way forward? “Paying for the work that’s used is a pretty simple solution to solve both the legal and ethical issues” involved with AI, Sholder said. “There should be compensation for past uses” of copyrighted material used to train AI, as well as going forward. “There’s a lot of money in this industry; pay for this content.”
It’s not all doom and gloom, though.
The panelists also discussed how writers can use AI as a tool, including its use in combating writer’s block. Almost every AI aspect of Lex, Baschez said, is to help stuck writers, “to help your process, to help spark ideas, to get the wheels turning.”
Additionally, Baschez uses AI to help shape his freewriting and act as a soundboard because he finds it’s similar to “working with any smart reader or colleague” but on that has “infinite patience and is available 24/7.”
And finally, it’s important to remember we’re just at the beginning.
Compared to the development of the Internet, when it comes to AI, “we’re in 1994,” according to Baschez.
“We are expecting way too much from AI way too soon,” Power added, noting that the timeline of AI’s development is unknowable, and we’d all be best served by working “toward a collaboration” with this technology.
