chatbot - Steven R. Southard (Poseidon's Scribe)

Home » Tag Archive: chatbot

Get Ready for AI Bestsellers

Ask a chatbot to write a story and it will do so. You’ll find the result contains the correct story elements. However, if you do that today, the story won’t move you. You’d rate it at junior high school level, certainly not a classic.

That describes the state of artificial intelligence story-writing in mid-2024. From this, you might well conclude that AI will never write stories as well as the best human authors do.

Prediction

Indeed, author Fiona M. Jones is the latest to draw that conclusion. She asks “is there any realistic prospect of AI ‘improving’ to a point where it becomes indistinguishable from the work of creative writers? Maybe you can imagine it. I can’t.” She goes on to state “I am not afraid that AI bots will take my place as a writer.”

I intend no disrespect to Ms. Jones. Others share her opinion. I accept the possibility that her contention may prove correct.

Superiority Complex

However, it occurs to me that the history of our species includes several symptoms of a shared superiority complex. In each case, people erected a metaphorical pedestal for humanity, only to have science tear it down.

In cosmology, early depictions of the universe showed Earth at the center. Today, astronomers relegate our world to a backwater.
In zoology, humans have long regarded themselves as lords of the animal kingdom. We claimed to possess the largest brain, and to be the only creature that feels pain or happiness, that talks, that uses tools, that is self-aware. The march of science seems to be trampling this pedestal as well.

Story-writing Today

For now, humans stand, undisputed, atop the Best Story-writers pedestal, at least on this planet. We’ve stood there for thousands of years, so it seems natural to regard the honor as permanent.

At this moment, AI seems unlikely to unseat us from that perch. In that, I agree with Fiona M. Jones. We humans have written stories for thousands of years, and told them verbally far longer than that. AI chatbots have written stories for a much shorter time, a few years at best. Hardly a fair comparison.

AI chatbots learn fast. Very fast. They can memorize the entire internet. They do not die, and therefore don’t have to teach the next generation of chatbots to write. I expect them to write with more skill and originality soon. But could they surpass us?

Ms. Jones offers many good arguments, but they boil down to the fact that human authors write about the human condition, and chatbots can’t possibly understand the human condition as well as humans do.

Maybe. But consider that human fiction writers often convey the thoughts and emotions of non-human characters in their stories. These characters include gods, animals, plants, even inanimate objects. Given similar creativity and imagination, chatbots might become capable of conveying human thoughts and emotions in a convincing way, even though they’re not human.

Story-writing Tomorrow

You could measure a story’s quality by the intensity of emotion it produces in the reader. Once AI chatbots understand us better, what’s to prevent them from crafting stories evoking strong emotions?

When and if they do, what will that mean for human readers? For human writers? I’ve already explored some of these implications in a previous blogpost and won’t repeat them here.

I suggest we should not assume our present superiority will last. We may not remain forever at the center of the writing universe, at the pinnacle of writing prowess, standing atop the Best Story-writer pedestal.

It should not surprise us when the first AI-written novel tops the Best-Seller list. Even then, we should not dismiss that achievement as a novelty, a fluke, unlikely to repeat.

In the meantime, fellow human writers, I suggest we write and publish the best stories we can, while they still sell. That’s the course steered by—

Poseidon’s Scribe

July 7, 2024Poseidon's ScribePermalink 2 Comments

Learning to Write Stories—Analysis or Practice?

What’s the best way to learn how to write stories? Should you just start writing a lot and work to improve? Or should you study the works of the best writers and understand their techniques before setting fingers to keyboard yourself? Or a combination of the two?

A writer friend enrolled in a literary master’s degree program and took a short story workshop class. The instructor told the students to dissect a literary work and analyze it. My friend discovered the entire workshop would consist of these analyses, and suggested to the instructor that students wouldn’t actually learn to write stories that way.

Picking a good metaphor, my friend said you can’t learn to build a house by taking apart other houses and studying them. You have to learn by doing.

The instructor disagreed, leaving my friend dissatisfied with that conclusion to the argument.

Let’s call the instructor’s way the ‘analytical approach’ and my friend’s way the ‘practice approach.’ (Note: I don’t mean to imply my friend only wrote and never read—this student objected to the 100% analytical approach imposed by the instructor.)

Who’s right? Both approaches seen to hold some merit, unless taken to extremes. A person who just analyzes famous writer’s works may develop expertise in analysis but never write a story of value. A writer who never reads seems equally unlikely to produce enjoyable prose.

I envision an experiment performed in two classrooms of second or third graders. One class simply writes stories without prompts. The other spends a year studying high quality children’s literature and discussing those books, and then the students write a story at the end. Which classroom’s students would end up crafting the best stories?

Imagine a line, a spectrum, with the pure ‘analytical approach’ at one end and the pure ‘practice approach’ at the other. My guess is, few of the great authors cluster at either end. They learned to write classic stories by some combination of approaches—by analysis and by practice. Perhaps an optimum exists on that curve, and I suspect it’s past the midpoint, toward the ‘practice approach’ end.

We might gain further insight on this by considering the artificial intelligence program ChatGPT. You may ask this chatbot to write a short story, and even prompt it with a subject, setting, mood, and style. The program will produce a short story for you in minutes.

How does ChapGPT do that? From what I’ve read, ChatGPT’s developers gave the chatbot many, many such prompts, graded the results, and provided feedback to the program regarding the grades. This seems analogous to the practice approach.

To produce a short story for you, ChatGPT scours the internet for information about the words in your prompt (for example, the subject, setting, mood, style, or other parameters you provided). That research seems analogous to the analytical approach.

Thus it appears ChatGPT learned to write short stories by some combination of approaches, someplace between the ends of the spectrum.

Note: ChatGPT does much more than write short stories. I don’t mean to sell it short. It also writes poems, essays, the answers to questions, and accomplishes many other tasks involving text.

In the end, my friend learned little about how to write a short story from the course. The analysis of classic short stories seemed, to my friend, better suited to undergraduate or even high school level, rather than a master’s degree course.

When learning to build a house, examining other houses helps, but so does building one yourself, and that’s similar to learning to write.

An appropriate mix of the analytical and practice approaches seems the best choice, at least for—

Poseidon’s Scribe