Stock

I have noted in passing the current controversy over AI-generated art, with artists asking their fellow creatives to make sure that any art illustrating their work is created by an actual person, and that any stock art being used is not created by an AI.

I sympathize, and I wish all creatives well. I suspect the horse has long ago escaped the barn— twenty years ago probably, when Google indexed every piece of art and every photo online, and did so without crediting the author or offering compensation. It’s images like these that the various AIs are trained on.

But my sympathies are always with creatives, and when, earlier this year, I wanted new covers on some of my books, I hired an actual Hugo-winning artist to create them.

But here’s a problem: many artists license stock art, which is then used or modified or re-envisioned somehow in their compositions. (Full disclosure: most of my ebook covers employ stock art.) This stock art is licensed from a dealer. But can an artist or cover designer know how the stock art was created?

In the recent case involving Tor Books’ cover for Christopher Paolini’s latest, the origin of the stock art was obvious to viewers, but that won’t always be the case. How is the end-user to know for certain?

As it happens, my family room wall has several pieces of computer art by a local artist, Lance Beaton. I specifically licensed one of these from the artist for one of my homemade ebook covers. I don’t think I was in the wrong to do so, most especially because the images are obviously generated by a computer. There’s no attempt by Lance or myself to make them “realistic” in any way. You’ll never mistake it for work by John Picacio.

I’m not any kind of specialist in visual arts or their creation. I’m a prose guy, and for someone as self-absorbed as myself I’m pretty good at predicting future trends, and what I’m seeing in AI-generated prose is a career-ending apocalypse.

Not my career, I hasten to add. I think my works are clearly the product of a certain sensibility and level of craft, and it would be hard to imitate my quirks. Tell an AI to write something “in the style of Walter Jon Williams,” and the computer would be perfectly right in asking, “Which Walter Jon Williams?” I contain multitudes.

ChatGPT, from OpenAI, has been available for less than a month, and it’s already turned into a tidal wave looming over online publishing.

It’s clearly a game-changer for academe. Ask it for “a 500-word essay on the influence of William Carlos Williams on the Projectivist Poets,” and this is what you’ll get:

William Carlos Williams was a leading figure in the development of the projectivist school of poetry, which emerged in the mid-20th century as a reaction against traditional poetic forms and techniques. Williams, along with other projectivist poets such as Charles Olson and Robert Creeley, sought to create a new kind of poetry that was more closely tied to the rhythms and patterns of everyday life and language.

One of the key influences that Williams had on the projectivist poets was his emphasis on the use of everyday speech and language in poetry. Williams believed that poetry should be rooted in the ordinary, everyday language that people use to communicate with each other, rather than the elevated and formal language of traditional poetry. This idea was reflected in his own poetry, which often featured colloquial speech and slang, as well as references to common objects and events from everyday life.

Another important influence that Williams had on the projectivist poets was his focus on the visual aspect of poetry. Williams believed that the way a poem looked on the page could be just as important as the words themselves, and he often experimented with different ways of arranging his poems on the page in order to create a visual impact. This emphasis on the visual aspect of poetry was taken up by other projectivist poets, who also sought to create visual interest through the use of unconventional line breaks and spacing.

In addition to these specific influences, Williams also played a key role in the development of the projectivist school of poetry more generally. Along with other poets such as Olson and Creeley, he was part of a community of writers who were actively working to create a new kind of poetry that was more closely attuned to the realities of modern life. This sense of community and shared purpose was an important factor in the development of the projectivist school, and Williams’s contributions to this community helped to shape the direction and goals of the movement as a whole.

Overall, the influence of William Carlos Williams on the projectivist poets was significant and far-reaching. Through his emphasis on everyday language and the visual aspects of poetry, as well as his role as a leader and mentor within the projectivist community, Williams helped to shape the direction and goals of this important movement in 20th century poetry.

Not bad, ne? It took the computer less than a minute to generate that.

It’s possible that teachers may have to go back to making their students write their essays in class to make sure it’s them what are doing the work and not a bot, but apparently there’s another program by OpenAI designed to detect when the author is using ChatGPT, so maybe the problem will solve itself.

Another giveaway would be lack of citations, or a very scant list of citations.

Chat specializes in dialogue, though when I asked for a dialogue between Montaigne and TS Eliot on William Carlos Williams’ influence, all I got was an error message.

You can also ask questions, and— here’s a useful item!– it can generate outlines for projects, which might be useful if you’re out of ideas or can’t figure out what comes next.

But the bot can also generate fiction, and I see this as a problem for a certain kind of writer.

I see a lot of these writers on Kindle Unlimited. They do genre: thrillers, mysteries, romance, adventure, and Cowboy Sheriff Daddy (Hawkins Texas Age Gap Secret Pregnancy Romance Series) (A real thing.)

The plots are standard for the genre, the protagonists are amiable and rarely wrong, and the prose is serviceable, though never to be mistaken for the work of Henry James. Writers who are successful on Kindle Unlimited have to write a lot— ten books a year is not unknown— and the books all follow the same plot, more or less. (I’ve known KU writers to get quite bored with having to write the same book over and over.)

Full disclosure— this last year I read a number of KU-type historical adventures while I was sick, or traveling, and in need of light entertainment. I quite enjoyed myself, though I did get a little bored with protagonists who always triumphed and never made a mistake.

ChatGPT has the potential to wipe these writers out. It can generate the plot, create acceptable prose, and write the dialogue. It won’t be perfect, and the operator— you can’t call him a writer anymore, can you?— will have to patch over the rough spots and fill in the lacunae. But still you can generate a colossal amount of fiction this way— replacing one kind of adequate novel with another— and essentially flood the market and out-produce the competition.

The first users to figure out how the bot works will have the advantage and fill up every possible market niche, effectively blocking others out of the market.

I’ve survived a number of publishing apocalypses in my long career, and I suspect I’ll be surviving this one, but I’m guessing there is going to be a Sorting that a lot of writers aren’t going to like.