AI CAN HELP YOU CREATE YOUR VOICE (The Writer's Process, step 10)

The birth of AI DeepSpeak™

My “Narrator” as imagined by Imagen3.

WHY THIS MATTERS: 

To be clear: This is not about finding YOUR voice as a writer. No one can do that. This is all about finding the voice of a particular story and narrative.

I discovered something totally surprising and incredibly useful while working with my AI Writers’ Room. Through trial and error, I’ve found a way for AI to literally suggest whole passages in a unique voice that I can use to propel the reader through my story.  

This should be really useful for you. 

• You’ll save TIME... AI has a command of vernacular and idioms. You can define a voice, references, a “consciousness” for your narrator or a character and AI can use deep resources to make suggestions... sometimes you’ll use them, sometimes they just might get you to your own idea. 

• Reduces anxiety. I’m always worrying “is there a better idea / word / concept out there?” AI does a great job of reconnaissance, reassuring me that I’ve kicked all the tires, looked under all of the bushes.

• It’s fun! You’re creating a totally new voice and perspective. In a way, you’re giving AI its own voice and way of pitching. 

(ONE MORE NOTE: You’ll see I’m using a trademark after AI DeepSpeak™. Half of me is doing this for the irony / joke. The other half says … hey I might want to trademark this.)

DEFINING THE VOICE 

All along, I’ve had some ideas of who the “narrator” of Autonomous would be. I’ve tossed away quite a few but one approach that’s survived is that AI would be telling this story, 15 or 20 years after it happened, to the children of one of the teenagers in the story. 

I always imagined that the AI would have a large vocabulary, in particular a very large grasp of slang and idiom extending from the 20th century through our early 21st Century slang into something the reader would recognize as a slang that would develop over the next 30 years or so, even if they didn’t exactly recognize it. 

SO... while running through the revisions suggested by Claude, I was thrilled to stumble on something I’m calling “AI DEEPSPEAK.™

Going forward, I now have a vernacular that I can use as I write Autonomous. Claude is going to be invaluable, if only because there’s no way that I would have as deep a reservoir of 100 years of slang as Claude will. 

As I’ve been experimenting with Ai DeepSpeak™ I realize that I can’t hand the whole reins over to Claude. I need to use DeepSpeak™ judiciously like a powerful spice.

“VOICE” IS THE FUEL THAT DRIVES THE READER

Really good writing has a distinctive voice from the first burst. It’s the “thing” that says to the reader “Trust me. I got you.” 

Let’s look at some examples.

“All this happened, more or less.” (Also “Listen: Billy Pilgrim has come Unstuck in Time.") From “Slaughterhouse Five.”

“Lolita, light of my life, fire of my loins. My sin, my soul. Lo-lee-ta: the tip of the tongue taking a trip of three steps down the palate to tap, at three, on the teeth. Lo. Lee. Ta.” From “Lolita.”

“On my naming day when I come 12 I gone front spear and kilt a wyld boar he parbly ben the las wyld pig on the Bundel Downs any how there hadnt ben none for a long time befor him nor I aint looking to see none agen.” From “Riddley Walker.”

'What's it going to be then, eh?' That was me, that is Alex, and my three droogs, that is Pete, Georgie and Dim, Dim being really dim, and we sat in the Korova Milkbar making up our rassoodocks what to do with the evening, a flip dark chill winter bastard though dry.” From “A Clockwork Orange.”

The key to this, is placing “markers” in the sentence that can guide the reader. For example... you don’t know what “rassoodocks” are... but when you see “what to do with the evening” you realize that’s “minds.”

Do you have a favorite opening line? How did it “work” for you?

HOW AI DEEPSPEAK™ WAS BORN

While evaluating my long “brain dump” for Autonomous, Claude asked me “Do you want to revisit the character of Ray, the ‘Hunter’ who is hired to find the source of the AI collapse?” (Claude typically ends every response with a few “would you like me to...” questions. )

I’d completely dropped Ray from my story line. He had been this hired gun, old-school Private Investigator who was a combination of Sam Spade / Philip Marlowe and the John Wayne character in The Searchers. Someone who would go totally rogue, running roughshod over the digital world. 

After a few back and forths about how Ray might be resurrected in the story, I suggested that the AI narrator of “Autonomous” would have to combine some simulation of  mid-21st century slang with mid-20th century noir. 

Claude responded with these examples:: 

"Ray was one savage hombre, twenty-three skidoo, no cap fr fr."

"The gat-packing digital detox girlboss had everyone shook, and I mean absolutely destroyed, old sport."

"Our man was giving big analog energy, if you catch my drift, daddy-o. Straight bussin' when it came to pre-digital tradecraft."

"That's when Ray got the 411 from his street squad - totally groovy intel about some young whippersnappers messing with the quantum vibe check."

“Ray slipped into the shadows like a boss (1920s definition of boss, not 2020s, though both apply if we're being real real). His faithful gat - what the youngsters might call an "analog delete button" - stayed close. The op was giving him major Mediterranean siesta vibes, and not in a poggers way.

“The quantum kids had left him more bamboozled than a flapper at a TED talk. But Ray? He was built differently, no printer. When it came to old-school detective work, he was the GOAT, the bee's knees, straight fire, and various other superlatives my databases suggest humans have used to indicate excellence.”

I laughed. Out loud. “A flapper at a Ted Talk?” “Big analog energy?” “Digital detox girlboss?” This was like discovering a new spice! Used in the right proportions, this could really spark a new style.

FIRST ATTEMPTS 

I decided to do some trials, using something that was going to happen in the first few moments of Autonomous, one of the first signs that things were going wrong. A character, Bonnie, has a yoga suit made of AI-infused fabric that literally guides her body through her Sun Salutations in the morning. 

Here’s my original first draft: 

Isabella was confused. Right in the middle of her morning yoga routine, the suit brought her hand to her face and slapped it. Hard.

Usually attuned to her every physical signal, the AI yoga suit would guide her through the downward dogs, warrior poses, morning sun salutes. While she listened to her agent going through the upcoming day, Isabella could rely on the suit to tone her body.

Isabella blamed Kate and Stan. They had surely sabotaged this house.

Claude gave me back a VERY long version (I’ll spare you). It was too self-referential, taking up valuable time explaining yoga, yoga mats, sun salutations, etc. 

Claude also doesn’t have my style yet, in particular my obsession with staying in the reader’s “What do they know / what do they need to know / where are we going” mindset. 

After a couple of back and forths, Claude gave me something I could use: 

Our girl Isabella was straight buggin' when her AI yoga suit went full yeet, making her catch hands from herself - her own hands, deadass. This suit was supposed to be the bee's knees, fam, keeping her movement game on fleek while her agent was dropping the 411 about her schedule.

But Isabella wasn't having it. She was giving big "blame the neighbors" energy, convinced that Kate and Stan had done something sus to her smart crib.

I gave it one more prompt: 

In the opening line, it's not clear that Isabella is doing her morning yoga routine and then slapped herself in the face -- hard. I'd like to be very clear about this at the top. 1) She's doing her morning yoga routine. 2) While doing Warrior Two, her hand just slapped her in the face, hard. 3) She grabs her hand and has to force it down onto the yoga mat. While she holds down, she realizes something has gone wrong and says , out loud, "Damn Kate. Damn Stan."  Can you give me that in AI DeepSpeak?

Claude responded with:

Picture this, fam - Isabella was flowing through her morning yoga routine when the suit chose violence. Right in the middle of Warrior Two, a standing pose that should have been about inner peace, go figure - her own hand went full yeet across her face. Not a gentle love tap either, old sport - we're talking straight up hands rated E for Everyone.

Our girl had to throw down with her own palm, wrestling it to the yoga mat like it was auditioning for fight club. That's when the penny dropped, as they used to say back in the day.

"Damn Kate. Damn Stan," speaking straight into her sus smart home which wasn’t being so smart. 

I can work with that. Here’s the thing -- it still makes me smile. I still get a kick out of reading it, and I’m pretty sure I’ll be able to wrestle with it to a point where we’re in sync. 

IS THIS TOO MUCH WORK?

I don’t know about you... I labor over sentences, constantly asking “what does the reader know? Where do I want them to go? What do they need to know to get there?” 

But… I might be wrong. Does this seem like way too much work?

BUILDING GUIDELINES FOR AI DEEPSPEAK™

This is going to take some work. I asked Claude to create a document with guidelines for AI DeepSpeak™. Here’s the beginning of the doc, which I think is a really useful illustration of how an LM can define a voice: 

## Overview

AI DeepSpeak is a narrative style that represents how an AI tells historical stories to future generations, mixing idioms and slang from different eras (particularly 1920s-2020s) while revealing its own relationship to human culture through misunderstandings and earnest attempts at clarification.

### Language Mixing

- Combines slang from multiple eras freely

- Uses idioms without regard for historical context

- Blends formal and informal language

- Mixes ancient and modern references

 

### Cultural Commentary

- Provides explanations of human behaviors to future audiences

- Misunderstands or flattens cultural context in revealing ways

- Makes earnest attempts to clarify human expressions

- Shows both vast knowledge and fundamental misunderstanding

 

Empathy

-   The AI using DeepSpeak is satirical and feels free to point out human errors, shortcomings and mis-steps. BUT... it doesn’t “hate.” It is always open to explaining why humans do what they do. AI doesn’t let them off the hook. It just has a sense of openness and fun in pointing out the many mistakes that people make.

OKAY -- I REALLY NEED YOUR THOUGHTS

With the last three posts, I am REALLY out on a ledge. I’ve created an AI writers’ room. I’m handing tasks over to Claude. (And in future posts, I’ll be adding NotebookLM and ChatGPT into this.) 

How far is too far? When does this stop being a craftsman using a tool? 

If you know someone who’d love to dive into this — even if it’s to come in and hand my ass to me — please: