Issue 79

On today’s quest:

— How is Claude at mimicking my voice?
— Japanese author used AI to write part of her award-winning novel
— How creatives can double down on the skills AI will never replace
— Immunology professor says AI is revolutionary
— International sports fans want more AI-driven content
— Grammarly expands its AI features
— Chinese AI isn’t constrained by scarce energy

How is Claude at mimicking my voice?

After running the GPT-4o versus GPT-5 experiment last week in which I attempted to get these models to write an item for the newsletter in my voice, I thought, “Why stop with ChatGPT?”

Claude 4 Sonnet

I’ve often heard Claude is a good writer, so I gave it a go with identical prompts, and the results were interesting because it addressed one of the things I said ChatGPT couldn’t do — before it began to write, it asked me to define the focus of the piece and gave me two suggestions.  

Both versions of ChatGPT just wrote up a general item, but Claude realized there were multiple ways to approach the topic!

Unfortunately, the writing sounded like AI — the language was much more cutesy than mine — and the piece was also far too long. Worse, it made quite a few stronger, more emphatic statements than I normally do.

Claude Opus 4.1

But … I was also using the free version, and it didn’t seem like a fair test since I was using the best ChatGPT model, so I bit the bullet and subscribed to the Pro version and tried it again with Claude Opus 4.1, which is supposed to be a better writer.

It also asked a follow-up question, but this time it was how I wanted to sum up the item (instead of the focus). The result was a much more appropriate length, but the writing had about the same amount of writing that felt “off” to me.

But the unsettling thing Opus 4.1 did that none of the other models did was speak as if it were me — using the phrases “but what really stuck in my mind,” “the most mind-blowing detail,” and “captures something I’ve been thinking about for a while.” I mean, this actually is how I write, so it’s accurately matching my style, but it feels like a violation to me — a step too far — especially since they weren’t the things that stuck in my mind or were mind-blowing.

I got what I asked for — it mimicked my style in a “using the first person” way — so I guess I shouldn’t be surprised. Maybe the surprising thing is that the ChatGPT models didn’t use the first person.

I’ll paste the outputs at the bottom for those who want to see the whole thing.

Japanese author used AI to write part of her award-winning novel

Novelist Rie Qudan says she used ChatGPT to write about 5% of her novel, “Sympathy Tower Tokyo,” which won the prestigious Akutagawa Prize in 2024 for new or rising authors in Japan. In the novel, a character chats with ChatGPT, and she used ChatGPT to write those parts.

But in a Guardian interview, she also says she, “‘gained a lot of inspiration’ for the novel through ‘exchanges with AI and from the realisation that it can reflect human thought processes in interesting ways.’” Yet, she doesn’t think AI is even close to being able to fully write a novel that rivals a human author.

How creatives can double down on the skills AI will never replace

I love to see people writing about ways writers can distinguish themselves in this age of AI, and a new piece at Creative Boom went beyond the typical surface tips. Here’s an excerpt from the “Life Experience” section:

Start learning AI in 2025

Keeping up with AI is hard – we get it!

That’s why over 1M professionals read Superhuman AI to stay ahead.

• Get daily AI news, tools, and tutorials

• Learn new AI skills you can use at work in 3 mins a day

• Become 10X more productive

Sign up and start mastering AI

Immunology professor says AI is revolutionary

Using sweeping, hyperbolic language, Jackson Labs immunology professor Derya Unutmaz said, “This isn’t an ‘incremental improvement.’ This is a revolution!”

His post on X says GPT-5 Thinking analyzed trial data in five minutes that took his team more than a month. Further, he says the highly detailed prompt led the model to identify all the important points they did and “uncovered several discoveries we missed.”

International sports fans want more AI-driven content

IBM backed a new study about AI in sports with the following results: “85% of respondents seeing value in integrating the technology into their sports experience and 63% expressing trust in AI-generated sports content.”

I was initially shocked by these numbers until I saw it was an international study, with people in the U.S. making up only ~10% of the respondents. It’s a good reminder of how much more AI is accepted in many other parts of the world.

Grammarly expands its AI features

I’m noticing a lot more education-focused announcements as school is starting, and Grammarly is out with new AI features designed for both students and educators. The one getting the most attention is the AI Grader, which the company claims can predict the grade a student will get on a paper.

Some of the other tools are things that are already easy to do with any LLM, such as Expert Review (“get feedback inspired by subject-matter experts so you can craft stronger arguments the way an expert would”) and Reader Reactions (“See how your writing might land with readers and get tips to revise your writing for the impact you want”).

Further, I find it deeply cynical that they are launching both a Humanizer, to make AI writing sound more human, and an AI detector, to catch students trying to pass off AI writing as human. I mean, it makes sense from a business perspective, but … ugh.

Finally, they are calling these features “AI Agents,” but as far as I can tell, most don’t fit the definition of “agent” used by AI companies. Agents are tools that go out in the world and do things for you without supervision, not something like a plagiarism checker. Grammarly’s new features are mostly just that — features. It feels like they’re just trying to jump on the agent hype bandwagon.

Chinese AI isn’t constrained by scarce energy

Fortune had a fascinating article about Chinese energy infrastructure this week. Here are two quotes that jumped out at me:

• [China] has consistently maintained at least twice the [energy] capacity it needs, Fishman said. They have so much available space that instead of seeing AI data centers as a threat to grid stability, China treats them as a convenient way to “soak up oversupply.”

• In China, renewables are framed as a cornerstone of the economy because they make sense economically and strategically, not because they carry moral weight. Coal use isn’t cast as a sign of villainy, as it would be among some circles in the U.S. –  it’s simply seen as outdated.

Quick Hits

Using AI

14 ways Googlers use AI to work smarter — Google Blog

Climate

America's power grid is no match for the AI data center boom — Quartz

Bad stuff

AI is making skilled hackers better and faster — NBC

Other

Three years in, one of AI’s enduring impacts is to make people feel like they’re losing it. — The Atlantic

We are gen Z – and AI is our future. Will that be good or bad? — The Guardian

Advertisers are increasingly using generative artificial intelligence to make their commercials. — New York Times

Draftsmith

A special thank-you to Draftsmith — an editing assistant in Word that supports the way you work. Draftsmith has supported the newsletter by signing on as a premium sponsor for a whole month of Mondays. 🙂 Get 30% off with code SIDEQUEST30.

What is AI Sidequest?

Are you interested in the intersection of AI with language, writing, and culture? With maybe a little consumer business thrown in? Then you’re in the right place!

I’m Mignon Fogarty: I’ve been writing about language for almost 20 years and was the chair of media entrepreneurship in the School of Journalism at the University of Nevada, Reno. I became interested in AI back in 2022 when articles about large language models started flooding my Google alerts. AI Sidequest is where I write about stories I find interesting. I hope you find them interesting too.

If you loved the newsletter, share your favorite part on social media and tag me so I can engage! [LinkedIn — Facebook — Mastodon]

Written by a human (except the examples below, obviously)

Claude Writing Examples

I used the same prompts I used to have ChatGPT learn my newsletter writing style and then write a newsletter item in my voice. In all cases, I wrote the first paragraph and asked the model to complete the item based on my notes. These are the results from two Claude models.

Blue text: Writing that sounds like AI to me
Red text: Writing where Claude presented an opinion as mine

If you want to see the earlier ChatGPT writing tests, you can find them in the August 13, 2025 newsletter.

Claude Sonnet 4 (the free model). 572 words

Many of the AI critics I hear don't seem to consider the global nature of AI. For example, they think the copyright cases in the U.S. can make AI go away. But world-class Chinese models already exist, aren't bound by our laws, and the technology will continue to advance. Supply chain expert Cameron Johnson on the latest Odd Lots podcast highlighted some of the ways China is ahead:

The infrastructure advantage alone is staggering. While we're debating data center locations and environmental impact, China built out 380 gigawatts of electricity capacity last year. The U.S.? Maybe 60 to 80 gigawatts. That's not even close — it's the kind of gap that gives them a massive runway to build AI infrastructure while we're still figuring out permitting.

But the real kicker is the talent pipeline. Cameron described visiting a booth called Semi-U at a semiconductor show in Shanghai, an organization focused on getting graduates into emerging tech fields. Their goal isn't modest: 100,000 graduates per year going into AI, quantum computing, robotics, semiconductors, and batteries. That's more than the entire Western world combined. By the end of the decade, they want half a million graduates in these fields — more than the entire world produces now.

To put that in perspective, China has 50 graduate programs specifically for battery chemistry and metallurgy. The U.S. has "a couple of professors," as Cameron put it. When a country is that serious about building expertise in specific technical areas, it shows.

And this isn't just happening at universities. Starting this fall, AI education is mandatory for every kid in China from kindergarten up. The entire educational apparatus of 1.4 billion people is being oriented toward AI literacy. Meanwhile, we're still debating whether kids should be allowed to use ChatGPT for homework.

The most telling part of Cameron's interview was his visit to a consumer goods manufacturer in southern China. The factory already had DeepSeek integrated into their operations, and they were using the current slow business period to train their 100+ R&D and technical staff on how to use AI tools throughout their supply chains. Not just in one department — throughout the entire chain, from raw materials to final components.

Cameron's observation was stark: "So you're seeing this massive compound throughout all of the supply chains, and that's just one company. Is Boeing doing that? Is Dow Chemical doing that? Is BMW doing that? I would argue no."

This gets to what might be the most important distinction: the U.S. excels at breakthrough discoveries — what Cameron called "the metaphysical" — while China is focusing on integration and adaptation. As he put it, "We're going to build the digital god first," but "the technology and the adaptation and the integration is all going to be Chinese."

DeepSeek was a watershed moment, but Cameron pointed out there are "100 DeepSeek equivalents coming online." While we're celebrating individual model releases and arguing about regulation, they're systematically building the infrastructure, talent, and integration capabilities to make AI ubiquitous across their entire economy.

I'm not saying this to be alarmist, but critics who think U.S. copyright cases or regulation can slow down AI development globally are missing the forest for the trees. The technology is already distributed, the infrastructure is being built at scale, and entire supply chains are learning to use it effectively. Whatever happens with AI, it's going to happen with or without us.

Claude Opus 4.1. 416 words

Many of the AI critics I hear don't seem to consider the global nature of AI. For example, they think the copyright cases in the U.S. can make AI go away. But world-class Chinese models already exist, aren't bound by our laws, and the technology will continue to advance. Supply chain expert Cameron Johnson on the latest Odd Lots podcast highlighted some of the ways China is ahead:

The infrastructure gap alone is staggering. China built out 380 gigawatts of electricity capacity in 2024, while the U.S. managed maybe 60 to 80. That's the kind of runway you need for massive data center expansion, and they have it.

But what really stuck in my mind from Johnson's interview was the talent pipeline. He visited a booth at a semiconductor show in Shanghai called Semi-U, which is helping young graduates enter key industries. Last year, they were doing introductions and internships. This year? They're placing 100,000 graduates annually into five fields: AI, quantum computing, robotics, semiconductors, and batteries. That's more than all Western countries combined. By the end of the decade, they want half a million — more than the entire world combined.

And it's not just graduate programs. Starting this fall, every kid in China — from kindergarten on up — will have mandatory AI education. Meanwhile, Johnson points out that China has 50 graduate programs in battery chemistry and metallurgy. The U.S.? A couple of professors.

The most mind-blowing detail to me was Johnson's visit to a consumer goods manufacturer in southern China. They had DeepSeek running in their factory. When he asked why, the owner explained that business was slow because of the trade war, so he had his 100+ R&D and technical staff figuring out how to use AI while they had time. But here's the kicker — the tech team wasn't just experimenting internally. They were talking to their entire supply chain about using DeepSeek and other AI tools, from raw material producers all the way to final component manufacturers.

Johnson calls this "massive compounding knowledge throughout all of the supply chains," and then asks the uncomfortable question: "Is Boeing doing that? Is Dow Chemical doing that? Is BMW doing that? I would argue no."

His final point captures something I've been thinking about for a while. The U.S. might be ahead in breakthroughs — what he calls the "metaphysical" discoveries. We might even "build the digital god first." But the technology adaptation and integration? That's all going to be Chinese.