How to Use AI in your Writing

 I've been hearing a lot of about the use of AI -- large language models, or LLMs -- like ChatGPT -- in writing. For example, game writing. Couldn't ChatGPT come up with a lot of barks? And quests? And stories? And backstories?

The most important thing to know about ChatGPT is that when you pronounce it in French, it means, "Cat, I farted." ("Chat, j'ai pété.")

Okay, maybe not the most important thing, but surprisingly relevant.

ChatGPT is not "artificial intelligence" in the sense of "the machine is smart and knows stuff." ChatGPT takes a prompt and then searches through an enormous database of writing to figure out what is the most likely answer. Not the smartest or the best answers, just the most likely one.

That means if you ask ChatGPT a question, you are getting the average answer. Don't base your medication on ChatGPT results. 

Note that this is *not* how Google answers work. Google's algorithm evaluates the value of a site based on links to other sites -- and rates the links according to the value of those sites. So it is more likely to answer based on what the Encyclopedia Brittanica said than what Joe Rogan said the other day. ChatGPT is more likely to give you Joe Rogan.

What does this mean for writing? ChatGPT will give you a rehash of what everyone else has already done.

Two problems with this:

a.  It is a hash. A mashup. ChatGPT does not know which bits of story go with which other bits. It has no sense of story logic. It does not know if it is giving you a good answer or a bad answer. It is not trying to give you a good answer. It is giving you an answer based on how much of one kind of thing or another kind of thing shows up in its database. Its database is probably The Internet. 

So, for example, if you asked it what happened at a typical wedding, it might decide that the groom has cheated the night before and the couple broke up -- because people generally do not write on the Internet about happy couples getting hitched without a hitch. 

b.  It is what everyone else has already done. Good writing comes out of your creativity. Your personality. Your experience of life. It is filtered through who you are. It has your voice. We are not hiring you to give us clams and tired tropes. We are hiring you to come up with something fresh and compelling. Something heartfelt. Something that gets a rise out of you, and therefore might get a rise out of the reader, or the audience, or the player.

If you ask ChatGPT for barks, it will give you the least surprising barks ever. The most average. The more boring.

If you ask it for stories, it will give you stories you've heard before, unless it gives you stories that make no sense that are based on bits and pieces of stories you've heard before.

So, how can you use large language models in your writing?

Simple:  it can tell you what to avoid.

Ask it to give you barks. DO NOT USE THE ONES IT SHOWS YOU. Make up other ones.

Ask it to give you plots. DO NOT USE ITS PLOTS. THEY ARE TIRED. 

Working with LLMs can teach you to take your writing beyond the expected. It will give you the expected. If you write away from or around what's expected, your writing will become unexpected. Fresh. Original.

Or, you can take its plots and then twist them -- which is what pro writers do all the time. We're not making up plots from scratch all the time. 90% of the time what we write is like something else, but twisted or subverted or redoubled in some way. (This appropriation and adaptation of earlier, better writer's work is called "culture.")

Tl; dr: you can legitimately use ChatGPT and other large language models for writing. But not directly. ChatGPT can only give you (a) nonsense and (b) cheese. But it is useful as a warning. If ChatGPT came up with something, it's probably too tired for you to use it. 

PS  Another reason not to use LLMs: the content they train on is, generally, whatever they can scrape off the Internet. They can't tell whether the content they train on is human-generated or machine-generated. So, the more LLM content there is out there, the more LLMs are training on LLM output. Initially, you may be getting what a computer thinks a human would say next. But after a while, you are getting what a computer thinks a computer would say next. See the problem? 

Malleable backstories

Human beings are perfectly capable of absorbing stories non-linearly. Very often we meet a person and only later find out about their past. Our experience of them grows and changes over time.

That's the best way to present character backstories, too. We need to see the character behave in the present before we care about how they got to be the way they are. We should only learn about their past once we have a sense of them in the present.

Backstories don't need to be set in stone. Some games let you pick the player's character's backstory, and then will adjust dialogue accordingly. 

But we can also assign characters -- even the player character -- different backstories depending on what the player does. If the game flags that the player likes to rush into combat (not a difficult metric to measure), we can give that player character a backstory that reveals why they are so impetuous, or anxious to prove their bravery. If the player likes to stealth and avoid combat, we can give that player character a backstory appropriate to someone who avoids confrontations. Hopefully that leads to the player feeling a kinship with the character they're inhabiting.

A few brief words on backstory

There's backstory that's in the game, and backstory that is in the game bible. I am all for backstory that makes its way into the game, and wary of game bible backstory.

(What I'm writing here also goes for novels and TV shows, but games do a lot more world building, so I'll stick to games here.)

It is easy to write backstory. Backstory, when it is not in the game, costs nothing except the time it takes for the writer to come up with it. People look at The Lord of the Rings and its yards of backstory and think it's not a real fantasy world unless you can trace the hero's lineage back generations, and talk about the wars before the one we're fighting, and so forth. I mean, Tolkien managed to publish an entire book of backstory, The Silmarillion, and it's still in print.

The problem with this sort of backstory is it becomes hard to remember how much of it you have actually put in the game. Writers think that because they've written backstory, they've accomplished something. 

But often no one reads backstory documents. Producers and creative directors are fond of asking for backstory documents, but artists are not fond of reading them. People like to be told a story, and most backstory documents are just recitals of things that happened.

Also, backstory documents often fail to focus on the theme of the game. They are just the writer letting their imagination run amok.

Consequently, if an artist or level designer does read the backstory document, and throws something into the game based on it, sometimes they are not reinforcing the main story, but diluting it, or muddying it. 

Consequently I am very suspicious of spending time writing game bibles. Instead, I'd rather spend time writing descriptions of things that will actually go in the game. Environmental narrative, for example. What is scratched on the walls of that cave? What is in that basement? What is strange about the ruins on the hilltop? What do the names of things tell you about this world? What folk story does that beer label refer to? Why are those tankards shaped so erotically?

We generally put backstory in a game in order to create a sense of immersion. But what creates a sense of immersion? The recital of historical facts? Or bits and pieces of the environment that hint at a broader world? The real world is full of symptoms of historical events, while the historical events, the causes, are buried in books. 

We are used to running into hints of the past; we like speculating about what they mean. In my neighborhood in Montreal, you can see the shadows of centuries: a two story stone wall topped with brick, a diagonal of black pitch running down the side of a building where it used to meet another roof, a bricked up hole that used to be a door, or a window. I love imagining what was there before. 

I think that hints of the past are more engaging than pushing history at the player. The player has an imagination; let them exercise it, and they will create a richer, more personal world than you can provide them.

"Show, don't tell" is a screenwriting maxim. "Hint, don't show" might be an environmental narrative maxim. We can only put so many objects into our world, since each has to be lovingly crafted in the computer. But the few that we can afford to build can stand in for thousands of other objects we didn't put into our world. The trick is to leave gaps in the evidence. Don't tell the whole backstory. Leave hints at a backstory -- possibly ambiguous hints. 

But all this is not game bible backstory. This is in-game backstory. That's what I think game writers should focus on. 

At the late night double feature picture show...

One of the folks I'm working with here in Hangzhou started playing the Rocky Horror Show soundtrack on the taxi's speakers (which is apparently a thing you can do here), and I mentioned I've seen the movie maybe 30 times. She asked why it was such a big deal.

My first answer was that it's a movie for people who feel like they're weird and maybe they're the only weird person in the world, to get together with other weird people and feel comfortable and maybe being weird is okay or even good.

Also, it came out at a time when gender fluidity was not a thing, so it was daring.

But I went home and bought the album on iTunes (third or fourth time I've bought that album) and listened to it. It is sad. The songs that are not about lust are all about longing. They are about having lost something precious. The opening song is about missing the science fiction double features of the singer's youth. Eddie sings about missing his rock'n'roll youth. Dr. Scott (or should I say, Dr. von Scott?) sings about longing for, but being frightened of, living his life for the thrills. Frank sings about going home. 

And they all sing about don't dream it, be it, but in a dreamy, slow way that makes you wonder if it is really possible to be it. 

And the whole thing turns out to be a dream, a science fiction movie, leaving the humans crawling on the planet's face. 

It is elegiac. 

It is about trying to capture dreams in a bottle.

It is a tough soundtrack to hear if you're dreaming it and not being it. 

Thank the Goddess I am married to the love of my life and I get paid to make up stories, which is as much being it as I could ask for. 

What do y'all think?

Unions are good for companies, dammit

Many years ago, I was asked to fly down to Cape Town as the Head Writer on a strange sf tv show called CHARLIE JADE, along with two fellow Canadians, Sean Carley and the irrepressible Denis McGrath, whom we all miss terribly. 

It was a Canadian-South African co-production, which meant we had to hire South African writers for half the scripts. And we had a hell of a time finding good ones. We found one very fine writer, Dennis Venter, but I was never happy with the others, whose scripts we basically threw out and rewrote. (They got the credit anyway. Them's the rules.) There obviously were good ones, but not that many, and they were all busy on the soaps. 

Why couldn't we find good writers? Because the South African writers had no union.

Because the South African TV writers didn't have a strong union, they got paid almost nothing for a script. 

Like, $2,000 a pop for hourlong scripts we got $50,000 to write. You could make more money writing novels.

Novels! I wrote a novel. I've made several thousand dollars from it, mostly from the Canadian government, which pays you $250 a year if your book is in enough libraries.

Unions are good for companies. They make sure that writers can afford to write TV. Writing TV is super fun, but it is very hard work. Not in the "did anyone ask you to dig a ditch?" sweat-of-your-brown hard, but hard as in you are thinking very hard for about 16 hours a day. (Writing is not just typing.) As Head Writer, I was working more or less 6 1/2 days a week. 

If that doesn't pay enough to have a middle-class life, then would-be TV writers will go do something that will. And when you want to get your show written, you'll look around and there won't be any good writers.

Dammit, this ought to be the golden age of being a TV writer. There are more shows than ever before. Some of them have spectacular writing. How is it writers can't afford to be writers? 

Because the companies are greedy, and they're killing the goose that lays the golden eggs.

It's short-sighted. The WGA wants 2% of the profits of the shows their writers create. 2%! Fucksake. 

Do you think there's nowhere else talented writers can go? Because there is this whole other business called video games that also hires writers, and gives us paychecks all year round. Some of us are doing pretty darn well. 

I hope the WGA doesn't have to strike. But God bless them, they will. And everyone will lose a lot of work. And hopefully, the companies will come around, like they did the last time. 

You don't get what you deserve. You get what you negotiate.

Minimum Context

 We're working on an opening sequence. I proposed a couple of short scenes. Other folks came back with scenes that I felt were long and contained too much information. 

As I wrote them, my general principals in approaching the scenes are:

1.  What is the minimum context the player needs in order to understand what is going on? We need the opening sequence to tell us who the main character is, and what his relationship with the other key figures.

2.   It is usually better to provoke the player to ask a question before we answer it. We need to provide a certain minimum amount of context (see above) so the player can ask a question that makes sense. Beyond that, it's better to provoke the question and then answer it later, than to give the player information they haven't asked for yet. This keeps the player engaged -- they pull themselves into the story rather than having story pushed at them.

So long as the player isn't actually confused, let the player wonder why things are happening. 

If we provide more than the minimum context, we don't know which information is the important part. Instead of getting wrapped up in the characters, we're processing information. When we're processing information, we're not asking questions. When we're not asking questions, we're not getting engaged in the story. We're also not getting to the gameplay!

Only resolve a mystery once the resolution is more satisfying than the mystery.

ECGC 2022 How to Take and Give Notes - Alex Epstein

Looks like my talk on how to give and take creative criticism from last year's ECGC is up on YouTube. Has been since August. Who knew?

My dad's passing

My father, Howard Epstein, passed away this afternoon, peacefully, after a long struggle with Parkinson's. 

A boy from Queens, NY, Dad served in the Navy at the tail end of World War II. He used his G.I. Bill to go the Institut d'Etudes Politiques (the "Sciences Po"), coming away a lifelong francophile. 

After a stint as a reporter for the Xenia Daily Gazette (Xenia, OH, is the largest city in the US beginning with "X"), he joined Facts on File, a weekly indexed digest of international news. He became its President, starting up Checkmark Books, an imprint for dictionaries, atlases and encyclopedias, and coincidentally employing many of my friends while I was in high school. He took the company from $2 million a year to $20 million a year, before being unceremoniously fired for trying to protect his staff from layoffs during a merger. 

Dad was a mensch. I'm touched by how many people really loved him. He was a nurturing man, and took care of everyone around him. He was crazy about my mom. He liked sailboats, British roadsters, and old biplanes, though he only owned the one catboat. He appreciated good wine, and made sure his friends got enough of it. He made a hell of a ragout. 

He was working on a book about the French people who saved, or did not save, or hunted, Jewish children during the Nazi occupation of France. 

I've missed him for a while, because Parkinson's steals people from you bit by bit for a long time; his passing was a blessing. He was an atheist, but I'm not, so I get to hope he's in a place where he feels all the love he brought into the world.