Character Brainstorming with AI

Here's a WRITER'S DIGEST article about how an author might use ChatGPT as an aid to composition without actually having the program do the writing:

Using AI to Develop Characters

The author, Laura Picklesimer, describes her experiment in workshopping character ideas with the help of generative AI. She began by asking the program how it might be able to help in character creation, and it generated a list of ten quite reasonable although not particularly exciting possibilities. She then implemented one of the suggestions by requesting ideas for characters in a thriller set in 1940s Los Angeles. The result consisted of "a host of rather stereotypical characters." When she asked the AI to suggest ways to subvert those characters, she was more impressed with the answers. Reading that list, I agree something like it might actually be useful in sparking story ideas. Her advice to writers who consider using such a program includes being "as specific as possible with your prompts, making use of key words and specifying how long ChatGPT’s response should be." She also points out, "It may take multiple versions of a prompt to arrive at a helpful response."

I was intrigued to learn that a program called Character.AI can be set up to allow a writer to carry on a conversation with a fictional character, either from literature or one of her own creations. The article shows a couple of examples.

Picklesimer also cautions potential users against the limitations of systems such as ChatGPT, including their proneness to "hallucinations." When she asked the AI about its own limitations, it answered honestly and in detail. Most importantly for creative writers, in my opinion, it can easily perpetuate stereotypes, cliches, and over-familiar tropes. It also lacks the capacity for emotional depth and comlexity, of course. If an author keeps these cautions in mind, though, I think experimenting with such programs a brainstorming tools could be fun and potentially productive -- just as a search in a thesaurus might not turn up the word you're looking for but might surprise you with a better idea.

It's worth noting, however, that this essay links to another one titled "Why We Must Not Cede Writing to the Machines" -- which Picklesimer, of course, doesn't advocate doing.

Do Not Go Gentle

Margaret L. Carter

Please explore love among the monsters at Carter's Crypt.

Gray Goo Doomsday?

Could runaway nanobots take over the planet?

"Gray goo is a term used to describe a lifeless world completely occupied by self-replicating nanomaterials that have consumed the energy of all life forms due to uncontrolled replication."

The complete explanation:

Definition of Gray Goo

A longer, more technical treatment on Wikipedia:

Wikipedia: Gray Goo

I came across the term in a short piece in the BALTIMORE SUN this past Sunday. Discovering how long this idea has been around, I was surprised I hadn't heard of it before. Unrestrained nanobot proliferation is compared to runaway generative AI. The example given in the newspaper refers to ChatGPT trying to be funny. When asked to tell a joke, the program falls back on the same twenty-five jokes over and over, about 90% of the time. If this example is typical of the effect of artificial intelligence on communication, could ever-increasing dependence on AI lead to decreasing originality and creativity? The sidebar in the SUN is an excerpt from this essay:

If Generative AI Runs Rampant

While I don't necessarily think we're doomed yet, this hypothetical scenario about the long-term effects of overuse of AI in creative work does raise disturbingly plausible concerns. As far as the basic viewing-with-alarm "gray goo" scenario is concerned, there's an obvious counter-argument: Nanobots couldn't reproduce uncontrollably unless we first invent them and then release them into the wild without safeguards, similar to Asimov's Three Laws of Robotics. So we probably won't have to worry about getting smothered in goo anytime soon.

Margaret L. Carter

Please explore love among the monsters at Carter's Crypt.

AI Sermons?

To follow up the topic of "creative" artificial intelligence programs, here are some clergy-persons' thoughts about sermons composed by chatbots:

Sermons Written by ChatGPT

Not surprisingly, the consensus from representatives of several different faith traditions is that AI-composed sermons have no "soul." This is one genre in which the personal, human element remains essential. A rabbi in New York comments, “Maybe ChatGPT is really great at appearing intelligent, but the question is, can it be empathetic? And that, right now at least, it can’t.” A pastor in Minneapolis writes about the program's attempt to compose an essay on maintaining one's mental health during the stress of the holiday season, “Although the facts are correct, there is something profound missing. . . . AI can’t understand community and inclusivity and how important those things are in building a church.”

On the other hand, New Testament scholar Todd Brewer asked ChatGPT to write a Christmas sermon based on the Nativity story in Luke's gospel, "with quotes from Karl Barth, Martin Luther, Irenaeus of Lyons, and Barack Obama." He was taken aback when the resulting composition was “better than many Christmas sermons I’ve heard over the years.” However, judging from the listed criteria, the requested product sounds more like an article than a sermon. Brewer himself, again not surprisingly, said it lacked "human warmth." Given that reservation, can the AI really be said to "understand what makes the birth of Jesus really good news"? Not to mention the unlikelihood that artificial intelligence in its present stage of development can literally "understand" anything -- raising a whole other complex question, whether intelligence can exist without consciousness.

From reports on ChatGPT from people who've tried it, I get the impression that it can produce creditable essays on factual topics, if fed enough sufficiently specific data, although they tend to be "bland." In more creative endeavors, as might be expected, the program falls short. And it wouldn't be ethical to present the program's raw output as one's original work anyway.

Since I'm a slow writer and first-draft composing is my least favorite phase of the writing process, I've often wished that a word-processing program existed that would take my detailed outline—such as those I've constructed according to the plan in Karen Wiesner's excellent FIRST DRAFT IN THIRTY DAYS—and expand it into a fleshed-out draft of a novella or novel in my own style. I could take it from there with editing and revision. While it's possible to instruct ChapGPT to create a writing sample "in the style of" a particular author, I strongly doubt that procedure would work for fiction anytime soon. So for the time being I'll just have to continue tackling the laborious stage between outlining (which I enjoy) and revising (which I don't mind, up to a point) the hard way.

Margaret L. Carter

Carter's Crypt