artificial-intelligence

New Writers Are Learning With AI Training Wheels

/ Logan Sidwell / Leave a comment

I’ve been a part of my local writing chapter for a little over 5 years. It wasn’t long after I started that ChatGPT was released into the world. Being the tech person I am, I fiddled with it. I learned how to use it for work, then shared what I learned with my fellow writers. At the time, it was a novelty. There was a sense of looming doom whenever we discussed it, but that looming doom was in the distant future.

Boy the distant future came up quick.

An AI Priest

I don’t use AI in my writing. I don’t use it in my editing. Unfortunately, as a young person that wants to be able to maintain a career ten years from now, I am forced to regularly experiment with the new AI models and learn how they work. The knowledge that experimentation has given me made me the tech expert in every writer’s room. That expertise turned me into something I didn’t expect, a priest of sorts.

Last month, during one of my chapter’s write-ins, someone came up to me to discuss their own experiments with AI. They spoke in a hushed voice, slid the work to me like it was contraband, then asked my opinion on what it all meant. I felt like I was a priest in a confessional, giving absolution to the sinner. “No, you don’t have to be ashamed of a test.”, “Yes, your AI story is disqualified from the contest”, “I wouldn’t trust an AI on opinions of taste, but that’s just me.”

In the writing world, there’s a deep shame associated with using AI. The community, without saying it out loud, has quietly endorsed a system of exile for the AI writer. I understand the instinct. I learned how to write without AI, all the best writers in my chapter learned without AI. How could anyone using AI become a ‘real’ writer?

After five or six confessionals, I realized something. It wasn’t the experienced writers that were dabbling with AI. It was the rookies.

The Next Generation of Writers

Do you remember when you first started writing?

I remember it was really hard. My mind was brimming with ideas, but when my fingers reached the keyboard, I had to push my wondrous scenes through the meat grinder of the page. Everything got all messed up, it was never as beautiful as I envisioned, and sometimes the product was incomprehensible. I ended up shoving an entire novel through that meatgrinder, now I hide that novel away. Things only started working when I brought in a professional to guide me.

In games, they call it a difficulty curve. When the game is really hard, they call it a difficulty cliff. Learning to write well is a steep, steep mountain. Most of us that scaled that mountain have forgotten what it was like.

My step-father picked up writing last year. He had never written a story before that. He’s a very tech-savvy guy, so rather than climb the cliffs of storytelling, he wrote his stories using a tool. Novelmaker. In one week, the AI wrote six books.

You might recoil reading that number. Six books in one week. I would have too, except that I knew he’d never try to publish them. It wasn’t about the money, it wasn’t about the fame, it was an expression of his creativity. He wanted to see his ideas play out on the page.

He took a break after the sixth book. Since then he’s spoken to me about writing more, and about writing better. He’s taken an interest in the craft and he’s started to see the weaknesses of AI writing. In time, he might even write a few stories of his own. I think it’s wonderful.

Last year, I presented on AI again to my writing group. I warned them that five or six years from now, new writers would learn to write not by trying and failing, but by using AI to generate their first stories. My estimate was wrong. It wasn’t five or six years, it was twelve months.

Half the new writers in my writing group have dabbled with AI. Some have asked for ChatGPT’s opinion on their work. Some have generated entire stories using it. I’m not talking about seventeen year olds that were raised on AI, either. These are adults entering the writing space using AI from the get-go.

The difficulty cliff in writing is too steep for newcomers. AI has softened the experience. On the one hand, it means there’s a glut of terrible writing. On the other hand, it means more writers are discovering their voice than ever before.

Gatekeeping Will Kill Communities

Every person is entitled to their opinion, especially on a topic as controversial as AI. It’s a values-based decision, and each of us has their own set of values. Some people think using AI for anything at all means you’ve violated some unspoken code of conduct. Some people think generating a book that’s a paraphrased copy of a different book is good business. The truth has to lie somewhere in between.

It’s okay for each person to draw personal boundaries. Personally, I’d never use AI to generate or edit my works. I think we get into trouble when we start drawing lines for others. New writers don’t ask for feedback on a one-page story that’s horribly written anymore. They show up with a ten-page story that’s bloated with filler words and ChatGPT cliches. It won’t be long before all prospective writers begin their journey using AI. As a community, we can’t just show those people the door.

That doesn’t mean we start allowing AI-generated stories to be submitted to our anthologies. It doesn’t mean we don’t point out that a story was AI-assisted. It just means we must continue to be welcoming to newcomers, no matter how they got into the art. When it becomes hard to empathize, I imagine a teen posting a forum: All the stories I’ve ever written were created with AI, I don’t feel like I’m a real writer. How do I get better?

Compassion and patience are the answers. From what I’ve seen as an accidental Writer-Priest, AI writing is like training wheels. Once a rookie learns to pedal, they’ll be eager to go without.

The Historic Strategy Games That Built My Book

/ Logan Sidwell / Leave a comment

For thirty years, strategy game players have been reckoning with the harsh reality that a computer might be able to play a game better than them. Beginning in 1997 with Kasparov vs Deep Blue and ending with Lee Se-Dol vs AlphaGo, AI inched ahead of human performance year by year, culminating in their total victory.

I love that tension, the open question that floats in the air with every game, ‘Can humanity win?’. Every victory and every defeat carried enormous weight. It’s the heart of my novel, The Human Countermove, strategy games and the fight against a mentally superior enemy.

The challenge with writing a strategy book is creating strategies that feel authentic and clever. The kind of ideas that are convincingly grandmaster in skill, but understandable to the general public. In order to achieve that, I had to learn from the best.

Kasparov vs Deep Blue (1997)

This game is the seed at the center of my book. The tipping point for humanity, the moment we realized computers could out-think people. In 1996, Kasparov won 4-2.

In 1997, they had a rematch, Deep Blue won 3.5-2.5.

Those two matches record the exact year engineering overtook training.

My favorite moment from the 1997 match comes in game 2, when Kasparov accused the Deep Blue team of cheating by having a Grandmaster help with a move. Even a computer can get illegal assistance from time-to-time it seems.

But the conflict of the moment is what really captures me. On the one hand, we want to believe a person is capable of outperforming a computer. On the other, what an incredible feat it is to reproduce the mind of a genius with a bit of code and training. Caught in between, the audience cheers both sides, athletic feat against human ingenuity.

Kasparov has a list of mistakes he says he regrets about that match. Moments he could have snatched a draw from a defeat, a victory from a stalemate. The thing is, if he had won, all it would have done is stall the inevitable. Instead of discussing the 1997 Kasparov vs Deep Blue match, we’d be discussing the 1998 Kasparov vs Deep Blue match.

It’s all of this I try to capture in my book. The tension, the conflict, the regret, and the determination to beat the unbeatable.

Now when Chess Engines and AI models face off against one another, they are a tier beyond our best players. A mentor for grandmasters like Magnus Carlsen, and something beyond the rest of our comprehension.

The Opera Game (1858)

This is a lighter game. The Opera game was played by Paul Morphy and The Duke of Brunswick over a century ago. It’s one I draw inspiration from in my novel not as a strategic tool, but as a piece of chess culture. The Opera Game represents the beginning of a chess student’s education, one of the very first games a novice will be introduced to.

Paul Morphy makes strong, understandable decisions against a much weaker opponent, rapidly gains the advantage, and wins in style. But it’s not just a game, it’s a story. The best in the world dragged into the Duke’s box to play a chess game in the middle of an opera. For beginners, it weaves a romance around chess, and attaches a narrative to one of their first lessons.

In my book, the protagonist Zouk does a lot of teaching on the side, as many professional players find themselves doing. When an opportunity to lecture to a big audience comes around and he realizes the inexperience of his listeners, he abandons the esoteric analysis had prepared, and leans on a tried and true classic with a fun story, The Highway Game.

Go: Lee Se-Dol vs AlphaGo (2017)

Lee Se-Dol vs AlphaGo ended in a 1-4 result. For those of us that had been tracking the development of computers since Deep Blue’s game against Kasparov, seeing AlphaGo take its victory wasn’t a surprise. Go is much more computationally difficult than chess, but Moore’s Law is a powerful force.

But did you notice the scoreboard? Lee Se-Dol won the fourth game. That was an upset.

Against Google’s best engineers and decades of neural networking and algorithmic design, a human being managed to snatch victory, and it all came from a single move. Move 78.

That move has been gone over, analyzed, and studied for years. It’s believed Move 78 pushed the game into a uniquely complicated position, a position AlphaGo couldn’t calculate. A blind spot in the computer’s play that drew out blunder after blunder.

Lee Se-Dol was like a grandmaster Quality Assurance tester, noticing where AlphaGo was weak and pushing it further and further down that path until its behavior was sub-par. Basically, Lee Se-Dol found a bug.

Even when it seemed impossible, a person beat the unbeatable.

The Hippo and Various Anti-AI Strategies

Since Kasparov vs Deep Blue, a thousand Chess engines have burst onto the scene. Anyone willing to run a bit of code on their computer and risk getting banned can play like a grandmaster. To beat such unsavory characters, grandmasters have had to develop a special set of tools. First and foremost is time.

Consider two games. One gives each player an hour to make all their turns, the other gives each player a minute to make their turns. The first game is deeply thought out, with strong moves that remove all chances of counterplay. The second is superficial, moves borne more from training than thought.

In tight time controls, using a chess engine becomes a liability. The grandmaster can play from their subconscious, but the cheater is stuck waiting for the ‘perfect answer’ from the machine.

Thus we meet The Hippo. The Hippo slows the game down to a crawl. Pieces only move forward a square or two, then build a near-impenetrable fortress. As the opponent approaches, the grandmaster makes every effort to close down the position, keeping the number of moving pieces to a minimum.

With each move, the cheater loses a little more time, and the walls close tighter around them.

As their time dwindles, the cheater is forced to throw in a few of their own moves. These usually turn out to be of a significantly lower quality than what a chess engine can put out. Once the grandmaster has stripped the cheater of their chess engine, they unravel all the complexity of The Hippo and go in for the kill.

Once again, complexity and time as weapons to beat an overthinking machine.

The Battle of Cannae and Real-Time Strategy Games

I love real-time strategy games. The feeling of making a plan, facing the hard truths of reality, making adjustments, and turning the battle in your favor is exhilarating. And they’re so different from a game like Chess or Go. In Chess and Go, the entire shape of the board is transformed in a single move. 

In Real-Time Strategy games like Starcraft, you’re making a new move every second, and it’s only when you add all those little decisions up that you end up with a result.

And in games like that, there’s one particular battle result that everyone is chasing.

During the Second Punic Wars, Hannibal faced a much larger Roman force and turned the battle completely in his favor. The trick? Draw the enemy in, encircle them completely, then tighten the trap.

The game in my book, LINE, isn’t like Chess or Go. It’s a little more practical in nature. In theory, the game is playable on a field, not that most people would enjoy the feeling of being shot by a rubber bag. Because of the practical realities of squadrons facing off against one another, tactics like Chess’ fork and pin don’t translate.

But what does translate, is the greatest military trap of all time. Let the enemy over-extend themselves, wait for the right moment, and strike.

Final Words

There are plenty of other strategy games I no doubt pulled inspiration from. Things like the Total War games, Role Playing Games, X-COM, but Chess was my guiding star. It’s funny, once you open your mind to a question like, ‘how does a person beat an AI in strategy?’, you realize how many other people already pondered the same question. 

AIs have been kicking Mankind’s collective butt for thirty years. It’s nearly impossible to imagine a person turning it around on them. But nearly impossible is still possible, it only takes the right person and the right techniques to turn things around. Even when the robot brains out-think us on every front, we can still squeak out a victory every now and then. Especially when we’re learning from everything that’s available.

In The Human Countermove, my protagonist Zouk Solinsen is the right person with the right techniques. The skills to outsmart computational genius.

My debut novel, THE HUMAN COUNTERMOVE is now available for purchase!