Ben Affleck explains why AI won't replace writers, and it's brilliant
Affleck goes 'Good Will Hunting' on the AI industry, and it's worth three minutes of your time.
Exciting news: My upcoming second book—Super Skill: Why Storytelling Is the Superpower of the AI Age—was named an Amazon #1 New Release thanks to your pre-orders! If you like this newsletter, you’ll love this book. You’ll discover:
How an unlikely neuroscientist discovered the storytelling tactics that build trust and drive sales.
How an laid-off banker built a media empire by practicing one active listening technique.
A powerful three-step framework—Unlock, Unleash, Upgrade—that will empower you to build a following, future-proof your career, and master storytelling in the AI Age.
The feedback from early readers has been incredible. Here’s what AJ Thomas, f. Head of Global Talent Experience at Google X, had to say:
“Super Skill is a lively, modern guide for anyone navigating the age of AI—told through a series of sharp, memorable stories that make big ideas feel approachable and fun. With humor, heart, and a front-row seat to how technology is reshaping our world, Lazer shows how storytelling isn’t just for creators but for all of us: leaders, parents, learners, and everyday humans trying to make meaning. It’s part masterclass, part story collection, and all inspiration for a generation learning to communicate, connect, and create alongside intelligent machines.”
I’m giving away bonus chapters, storytelling courses, workshops, dope socks, signed books, and more to those who pre-order. Get yours now!
Now onto today’s edition:
Every 12 months or so, Ben Affleck pops into my feed to explain the state of AI, creativity, and media in a way that leaves my jaw on the floor.
The latest moment came this past week, when Affleck and Matt Damon stopped by the Joe Rogan Experience to promote their new movie, The Rip. Affleck delivered a three-minute monologue that’s required viewing for anyone who cares about how AI will impact media, marketing, and the future of work.
The AI-savvy Hollywood legend makes four points that are worthy of closer consideration:
1. AI writing disappoints because it reverts to the mean
“If you try to get ChatGPT or Gemini or Claude to write you something, it’s really shitty,” Affleck says. “And it’s really shitty by its nature because it goes to the mean. Now, it’s a useful tool if you’re a writer, and you’re going, ‘I’m trying to send something up, and they’re supposed to get a letter, and it’s delayed two days,’ and it’ll give you some examples of that. But I don’t think it’s actually very likely that it’s going to write anything meaningful, or that it’s going to be making movies whole cloth.”
Some people will bristle at this characterization of AI writing, and sure, “really shitty” may be a little harsh. After all, AI is pretty good at formulaic, derivative content; if I give it a research report and ask it to write a landing page to drive sign-ups, it’ll do a B+ job. But AI does struggle mightily with original storytelling and thought leadership, thanks to how it’s trained.
This summer, I broke down why that is in my essay, Beware the AI Vortex of Mid:
Large Language Models (LLMs) like ChatGPT, Claude, and Gemini are mid by design. (If you don’t speak Gen Z, mid means mediocre.)
LLMs learn to generate text from the most common patterns in their training data, which is comprised of everything our Silicon Valley overlords could scrape off the internet, copyright be damned. Most of that training data is poorly written—technical, jargony, and flat.
The LLMs are then fine-tuned through Reinforcement Learning from Human Feedback (RLHF for those in the know), in which humans rate the AI output, often in “digital sweatshops” in Africa and Southeast Asia. These human trainers are instructed to steer the AI towards safe, inoffensive outputs and to rate jargon-laden technical answers higher.
This is why AI writing hasn’t made the same leap as AI coding. With coding, AI can test its code until it produces something that works when deployed. With writing, AI systems are sabotaged by a training process that makes it sound like an insufferable junior McKinsey analyst from Connecticut named Brett. Sure, AI is better than the average human writer, but it’s far worse than the best human ones. AI is inherently an averaging out of all the shitty content on the internet, and as a result, it creates a vortex that sucks you back to the middle when you rely on it too heavily.
I call it the AI Vortex of Mid.
If you laid out all the content on the Internet on a 2x2 graph spanning Low-High Originality and Low-High Quality, most AI-generated or heavily AI-assisted content would fall in the middle. Not bad, necessarily, but not good enough to make a significant impact.
Six months later, this dynamic hasn’t changed—even with the release of ChatGPT 5, Claude 4.5 Opus, and Gemini 3. AI can write, but it can never be a storyteller. Which brings us to the next point Affleck makes.
2) AI’s writing and storytelling capabilities are progressing more slowly than expected.
“It turns out the technology is not progressing in exactly the same way they presented, and really, what it’s going to be is a tool,” Affleck continues.
When ChatGPT came out three years ago—featuring OpenAI’s GPT-3.5 model—people lost their minds, in large part because it made a significant leap over previous large language models like GPT-3, which powered widely available AI copywriting tools like Jasper.
GPT-3 was like Steve, a C-minus freelance copywriter who you’d probably fire after two or three jobs. ChatGPT was like Ted, a B-minus copywriter with tons of ideas and the occasional flash of brilliance. We’ve all had Teds on our team and thought, “Yeah, he’s okay.” So once AI crossed the Ted Chasm, people went bananas.
The other reason was the hype. AI leaders like OpenAI’s Sam Altman and Anthropic’s Dario Amodei preached that AI would improve exponentially, thanks to AI’s scaling laws. Sure, it was Ted, a B-minus copywriter now, but it’d be Toni Morrison in a matter of months.
In reality, though, AI’s improvements as a writer and storyteller have stalled. A large part of that is due to the training process. But as I wrote in my deep dive essay on the trajectory of AI development—Storytelling Is the New Coding—there’s another factor at play.
As Dan Balsam—co-founder and CTO of Goodfire, one of the hottest AI research labs in Silicon Valley—told me, the producers of AI models have been forced to make trade-offs on what the models get better at, thanks to limitations in power, microchip supply, and cost.
AI giants like Google, OpenAI, and Anthropic are following the money and focusing on making AI better at coding and other technical tasks, like data analysis in Excel. After all, there’s much more money in replacing software engineers and data scientists than copywriters.
It’s also much easier to judge objectively good code than objectively good writing. With coding, AI can simulate tests of its code until it produces something that works when deployed. The training data for AI code, while not perfect, is also considered much higher quality.
That’s why there’s good reason to expect that AI’s improvements in writing will continue to plateau or increase incrementally. It’s a challenging problem, and since AI can already execute jargony business writing at an acceptable level, there’s not a huge incentive to improve it.
3) AI companies are incentivized to scare the crap out of us.
This is the point that Affleck made that made me sit up and think the most:
“There’s a lot more fear because we have this sense of existential dread,” he says. “It’s gonna wipe everything out. But that runs counter to what history shows—adoption is slow, it’s incremental. I think a lot of that rhetoric comes from people who are trying to justify valuations around companies where they go, ‘We’re going to change everything! In two years, there’s going to be no more work!’
Well, the reason they’re saying that is they need to ascribe a valuation for investment that can warrant the capex (capital expenditure) spend they’re going to make on data centers, with the argument that ‘Oh, as soon as we do the next model, it’s gonna scale up, it’s going to be twice as good.’ Except that ChatGPT-5 is only about 25% better than ChatGPT-4, and costs about four times as much in terms of electricity and data. When AI came out, the line went up very steeply, and now it’s kind of leveling off.”
You can quibble with Affleck’s numbers here. ChatGPT-5 is probably more like 40-50% better than ChatGPT-4 based on (flawed) AI benchmarks, and GPT-5 might be eight times as expensive as GPT-4, although it’s actually cheaper in terms of what OpenAI charges you to use its API.
But his larger point is poignant: the CEOs of Silicon Valley giants have a massive incentive to scare the crap out of us.
The AI boom is propping up our entire economy, and our economy now runs on narrative. As Kyla Scanlon has written, our foundational economic substrates used to be things like land, labor, and capital. You used to start by developing a product or project, building its infrastructure, raising capital, and then acquiring attention. Now, it’s reversed. The foundation of any project or product is attention and narrative. Your ability to capture attention becomes a prerequisite for getting anything done, and your ability to tell a compelling narrative determines the flow of money, sentiment, and momentum.
OpenAI is worth $500 billion on only $13 billion in revenue. It plans to burn $17 billion this year alone. The only story that justifies the hundreds of billions of dollars we’re investing in AI overall is one of massive disruption, with AI replacing workers en masse. A story of AI as “normal technology”—with most companies struggling with org-wide digital transformation, and most workers slow to adopt it in a way that drives real productivity gains—isn’t one you can even consider telling if you’re Sam Altman or Dario Amodei.
For what it’s worth, I don’t think the hype is nefarious. I’ve spent the past five years working for AI companies, and most people in the AI and tech industry genuinely believe that AI will imminently replace jobs en masse. I get it. If you’re a power user of AI at work—like I am—it’s unfathomable to imagine that AI won’t massively disrupt the world of work in the next 12-18 months.
But what most power users miss is that they’re in the top 1% of AI adoption and living in a bubble. Change comes much more slowly at the average company in corporate America.
The incentive to induce fear is still there, though. Even if Sam Altman woke up tomorrow and had an epiphany that AI’s impact is going to be much slower and more incremental than he previously thought, it’s not something he can say. So when you see fear-mongering headlines about how AI is going to take your job, take a breath and remember the incentives at play.
4) Most people aren’t using ChatGPT for work—and that should color how we frame AI’s impact.
When OpenAI released ChatGPT-5, Sam Altman hosted an AMA on Reddit. Users revolted, begging him to bring back GPT-4. Was it because ChatGPT-5 wasn’t as helpful at work? Of course not. It was because OpenAI had reengineered GPT-4’s sycophantic personality, and users felt like they’d lost their best friend. As Affleck recaps:
“A lot of people were like, ‘Fuck this! We want ChatGPT-4!’ The vast majority of people who use AI are using it for companion bots. There’s no work, there’s no productivity, there’s no value to it.”
This may sound hyperbolic, but Affleck is right—73% of ChatGPT usage is non-work related, according to OpenAI’s research. The heaviest ChatGPT users seem to be using it for companionship. In the coming weeks, we’re likely to hear that ChatGPT has reached 1 billion users, accompanied by breathless prognostications from LinkedIn influencers about the imminent arrival of mass unemployment.
When that moment comes, ask a simple question: What are they using it for?
What you should read next:
Recommended
AI’s Memorization Crisis (Alex Reisner / The Atlantic): New research shows that AI models can reproduce 95% of Harry Potter books. This could completely upend the copyright infringement cases facing AI giants, altering the course of AI development.
The 26 Most Important Ideas for 2026 (Derek Thompson): You’ll be ready to crush some cocktail conversation after reading this.
In 2026, We Are Friction-Maxxing (Kathryn Kezer-Morton, The Cut): Does friction-maxxing sound like a term invented by a gaggle of drunk frat bros? Yes. But it’s also easy to see this trend catching on.
I’m the best-selling author of The Storytelling Edge and the upcoming book Super Skill: Why Storytelling Is the Superpower of the AI Age. Subscribe to this newsletter for fresh storytelling and audience-building strategies in your inbox each week, and pre-order your copy of Super Skill to unlock bonus chapters, workshops, courses, and more.
Once again Affleck proves he is highly intelligent, yet nobody will ever let him be as such.
I love Ben's passion, and I love your take, Joe. But we don't know. I guess I'm thinking that it's not AI vs. human. It's going to be some type of cohabitation of creativity, as Ray Kurzweil talks about. At some point in the very near future, you're not going to know what's come from human and what's come from AI. It will just be creative. Or story. We'll look back on these conversations and laugh.
My concern is that Ben sharing this gives a lot of writers and content creators security that it's all going to be okay instead of actively looking at how to change their processes now. And if they look at everything, and they look at the system, and they look at how the game is going to be played, and say, "I don't want to use any AI at all," then I think that's just fine.
Keep up the great work.