Showing posts with label Isaac Asimov. Show all posts
Showing posts with label Isaac Asimov. Show all posts

Friday, March 17, 2023

AI Writes Like a Drunk Middleschooler

Artificial Intelligence. AI. It sounds so innocuous unless you grew up in the 60s and early 70s watching cheesy scifi matinee movies about rogue robots going on rampages. Maybe if you read Asimov and realized that the entirety of his writing career was spent coming up with the laws of robotics and then BREAKING them.

Sure. At the moment, we're talking about using AI to generate words or art for us (though I guarantee that AI is already in use a ton of other places that impact you already - you just don't know it.) It's a big gap between cribbing someone else's work and dodging Skynet. 

I am already using AI (specifically Chat GPT in this case)  to write with. It's for the day job where I'm part of a research group working on using the power of AI to transform our business model. I'm taking classes from those smarter and more experienced in the arena that I am so I can learn how to wrest forth the best of what AI has to offer. Benefit: I'm producing technical writing content for clients that is roughly 1/3 AI written based on prompts I give the AI. Problem: AI writes like a drunk sixth grader. Maybe a seventh grader. There isn't much that AI writes that can remain untouched. I cannot simply copy and paste wholesale and move on. Second problem: I MUST know my subject matter because AI is pulling information from the web. Some of that information is outdated. Some is dead wrong. Result: AI has shifted me to being a knowledge worker rather than someone who sits around and thinks up words to write for clients.  It's not all bad. Drafting is my weak spot. Editing is my strong suite.

AI favors writer/editors who have a grasp of their subject matter and who know how to match a brand's tone and voice. The moral of the story is that no matter how much or how little AI content I include in a piece of work, I will always have to tweak it or rewrite it. Always.

Chat GPT is good at distilling information from the dark, dusty corners of the WWW and bringing back something reasonably cogent. Mostly. To generate the best content, you'll need to understand what Chat GPT needs as a prompt. It's powerful for nonfiction. It's a little less useful for fiction. Like too many of us, it seems to want to avoid conflict.

I'm experimenting with AI (Sudowrite, in this case) in novel generation. I'm finding it is a much better fit for fiction. I do not like it at all for nonfiction. At least not the kind I'm working on where I need to bring hard data to a paper.  Sudowrite also requires super robust prompts that are loaded with details around your story. There are classes available for this. Some are free. If the tech interests you, the classes are worth the time. Benefit: Collaboration with that drunk middle schooler I mentioned earlier. Sudowrite takes a prompt and generates 2-3 text options. Problem: The writing is pretty terrible. Unless you declare a POV character in your prompt, Sudowrite defaults to omniscient POV. It's all tell and no show. It's pretty bare bones. The power of Sudowrite is that you can change all of that with different options in the interface. Maybe you want more action. Sudowrite can rewrite for more action. Or more description. Or more intensity. Sudowrite isn't afraid of conflict or blood. And as you write or paste in your own writing, Sudowrite learns to match your style a little better.

Sudowrite isn't capable of generating a novel without a writer to knit everything together and direct the story. Like one of the instructors said in class - think of AI as a junior writer. You're still senior. You know the craft. You know the story and the characters. You will always have to supervise and direct the junior writer's efforts. Just like with Chat GPT, I cannot take big chunks of text from Sudowrite and import it to my WIP. Just can't. I can pull a cool turn of phrase or a sentence or two. But mostly, Sudowrite's power is in making me question how I'm thinking about my story and the direction it needs to go. I have not successfully completed a book with it. I do know people who have and who continue to use it to spur their writing.

So. AI. Evil? Benign? Beneficent? Eh. Yes. To all of it. There will be good. There will be bad. Most of it will be neutral. But AI is a genie that escaped the bottle. There's no getting it back in there, now. We're going to have to learn to cope with it. The way that generations before us had to learn to cope with the evil televisions rotting our brains and making us all go blind because we sat too close. Or computer games. Or cell phones. Or whatever other technology destroyed hearts and minds and the modern family and civilization as we know it.

I do believe that AI will change the shape of work. It already is and has. You can figure out how to work with it, or you can ignore it like most of us ignored crypto. I expect that publishing will break up into camps. One will expound the evils, the other will tout the benefits, and somewhere in the middle, the rest of us will just try to finish our stories and get them out into the world.

Friday, March 3, 2017

The Gaps in the World

I'm sitting beside someone playing The Division. It's a computer game set in NYC just after Christmas - a Christmas wherein someone infected dollar bills with a genetically engineered virus. He put the bills in circulation and the virus wiped out a broad swath of the population. The premise of the game is that it's January. The player is an agent in a clandestine organization called The Division. The player is activated as a means of reclaiming the city and lending aid to the remaining survivors.

Amazing world building. The story tellers/game developers seriously thought through the threats, the challenges, and the ways people would react to the disaster. They really considered how long it would take for essential services to break down - how long it would take official government agencies to sweep in and cordon off the city before they, too, started dying.

But there are some serious gaps, if not in world building, then in logic. All of the firetrucks and police cars have their lights on and flashing. The cars are abandoned, their hoods crumbled, but their batteries are still good, by God! However. People have been dead for so long that all the dogs in the city have become feral. The single biggest miss by the dev team? The dogs aren't roving in packs.

Do you know how long it takes dogs to revert to feral? It's a researchable parameter. Not to mention that anyone who's watched a single National Geographic episode knows that dogs are social animals. They require a pack. In our homes, the accept their human families as pack substitutes. Were that family suddenly taken away and a dog had to fend for itself, it would have to have another pack. The dog would automatically seek other humans and ask for help. If that didn't work and the dog didn't simply starve, it would, for its own psychological survival, have to join a pack of other free-roaming dogs.

It seems like a little thing, doesn't it? But it's indicative, to me, of some lazy world building within the game. Someone simply went, "Cool element! Feral dogs!" But no one bothered to ask a simple question. "Hey. If the dogs went wild, why are the crumpled cop cars still flashing red and blue? I mean their car batteries died in the first 72 hours, right?"

It's proof that misses don't have to be great big hairy things. It's the little things that build up over time and really start to bug people. Ask Walking Dead fans whose visages harden and whose lips thin ever so slightly whenever the Rick and the group drive past an obviously cultivated field or a mowed lawn 2 years after the zombie apocalypse. (Also, why do the cars still start? Have you ever had gasoline varnish in an engine after a single season??)

World building is very much akin to the cultural iceberg - we only see the tiny bit above the surface, but there's an entire huge structure underpinning what we see. That invisible structure requires deep thinking if you're going to be creating it. It is where the 'why' comes from for your world. If the 'why' is firmly in place, a writer is less likely to miss the kinds of world building elements that'll get books tossed across the room.

I, personally, have a pet peeve about time travel stories. I have yet to see a movie with time travel as an element that didn't shoot itself right the timeline. (Meaning the story creates an Asimov Paradox even after Isaac Asimov described the paradox so writers could avoid it.)

What world building/consistency misses rub your fur the wrong way?