Since AI Dungeon’s heyday, I’ve wondered about a better way to generate Content™. Usually, pondering this, I assume that the purpose of generated content in this domain is to adhere to a couple sets of strict standards a conventional author might like. But after reading this, I’m not so sure.

I usually focus solely on text adventures (Interactive Fiction, for pedants) for simplicity’s sake, since even in this nascent form it gives me the core parameters I’ve been grappling with:

• The content is narrative in nature
• The minimum viable product is cheap to produce
• A large, mature body of existing media to use as reference material
• I bought 50 Years of Text Games and I need to retroactively justify the purchase

Okay with that out of the way… So you know that thing about AI and GPT and all that? It’s cool, it generates legible text and can be applied a little creatively for a system of classifiers- in turn leading to a slew of manners of “gamification”. But let’s throw that all out the window and focus on something else instead:

Let’s say we abducted some authors from all over the world. Language doesn’t matter- they’re locked in a basement. (More practically, we would hire people to do the work described, but the locked-in-a-basement thing is a useful way of looking at it because you don’t have to think about costs, ethics, etcetera.) They sit down at their computers, white text on a black background, and then… Boom. Chained to their chairs, launched right into the Twine user interface. They get a typical text adventure product, sure but something’s different: Their MO is to evaluate and author additional branching options based on global player selections, thereby forcibly birthing an ever-expanding nightmare worm-fetus-thing of human-touch beauty. Using a complex system of organization, graphs, and other doohickies, these authors will be expected to maintain internal consistency, thematic and stylistic cohesiveness in line with the current read-through’s essence, and a few other emergent properties that people care about… Blah, blah blah… Then, boom! Generated content.

The idea behind this is that, barring cost, a systematic infrastructure could well be implemented that well exceeds ChatGPT in terms of quality, and we could have done so in bulk since the 1990s… So why haven’t we just done that?

Oh. Right.

We did.

A lot.

See, this idea of total, absolute freedom in some interactive media isn’t entirely new, it’s just… Not that hot. Not that snazzy! Not that valuable. It’s just… Bulk trash. When you can do anything, it means nothing.

There’s a certain neglect evident in the way the content-churners represent the human interaction with art. We engage with what we read (or “play”, for you gamers), fundamentally and absolutely! Regardless of the actual, literal interactiveness of any article of media! We project our own thoughts and experiences well upon it, and in doing so there is nothing left to be desired. Simply put, this new breed of trash hasn’t changed the game. After all, who hasn’t flipped to a random page in the Library of Babel to reflect?

Generating the content isn’t the point. (Okay- I mean- okay, it’s more nuanced than that, obviously, but it’s not the essence of the point. Friggin’ pedants.) Engaging with it is.

… So, despite all this, you’ll always be able to catch me quaffing down this delicious trash on the regular, because I adore it. But Mr. Cook already covered that.