The Machines Are Thinking. We’re Just Forwarding the Messages.

The Machines Are Thinking. We’re Just Forwarding the Messages.

A few years ago, I published a little book about the great mathematician Bernhard Riemann, written by mathematics historian Jeremy Gray. It was part of our Simply Charly Great Lives series. Riemann is known for many things, but perhaps his greatest claim to fame is the Riemann Hypothesis, a mathematical puzzle involving prime numbers that has stumped some of the brightest minds for more than 160 years.

When the book came out, mathematicians were eager to read it. Even Sir Michael Atiyah, one of the greatest mathematicians of his generation, got in touch to ask for a copy. He was working on the Riemann Hypothesis himself. In 2018, he presented what he believed was a solution. It was an impressive effort, but his proof didn’t hold up. He came close, but not close enough. The problem remained unsolved.

Until now, perhaps. Yesterday, Levent Alpöge, a mathematician at Anthropic, the company behind Claude, announced that he believed he had discovered a proof of the Riemann Hypothesis with the help of AI. He also claims the work extends to the Generalized Riemann Hypothesis, an even broader mathematical problem. He released a computer-verifiable version of the proof written in Lean, a programming language used to check mathematical arguments. There is a big catch: the proof has not yet been independently verified, so the Riemann Hypothesis cannot be declared solved. But think about what's being claimed here. A mathematician working with an AI system may have cracked one of the most famous problems in the history of mathematics, something that has defeated generations of brilliant minds. That’s remarkable, even if the claim ultimately doesn't hold up.

This isn’t entirely out of the blue. Anthropic had previously revealed that one of its researchers gave an unreleased version of Claude the Riemann Hypothesis and asked it to take a serious crack at the problem. Claude didn’t solve the original, but it made progress on a related mathematical question that had stumped researchers for decades. OpenAI has also reported an AI-assisted proof of a weaker version of the hypothesis. The machines, it seems, are getting closer. And it raises a much bigger question: What happens when machines begin doing things we once considered uniquely human?

I'm less worried about mathematicians than about the rest of us, because the change is already showing up in our everyday lives. AI is writing books, composing music, solving problems, answering questions, and carrying on conversations. It can clean up our writing, organize our thoughts, and make us sound a good deal smarter than we might actually be. That’s all well and good. But at what point does a little help turn into complete dependence? We ask one machine to write something, another to polish it, and then send it to someone who uses a third machine to write a reply. Two people think they’re having a conversation, but neither may actually be doing much talking.

I got a taste of this recently while messaging with someone I’d just met. My messages were ordinary enough, with the occasional typo, a sentence that wandered off, and a comma in the wrong place. Her responses, on the other hand, were long, perfectly polished essays. I was chatting. She seemed to be writing a dissertation. Eventually, I told her I suspected she was using AI to respond to me. I wasn’t looking for a perfectly constructed argument. I wanted a conversation with another person. A little spontaneity, a joke, an awkward sentence or two. You know, the usual signs of life. It was like phoning someone and getting their speechwriter. There’s something strange about sending someone a few casual thoughts and getting back five paragraphs of perfectly organized prose. You start wondering whether the person is actually talking to you or simply feeding your messages into an AI program and forwarding the results. Whatever happened to just talking?

And it isn’t just casual chats. Students are handing in essays they didn’t write. Professionals are sending reports they didn’t draft. Businesses are automating customer service. People who once struggled to write a birthday card can now produce a sentimental masterpiece in three seconds. Maybe that’s progress. But there’s a difference between helping someone express an idea and doing the thinking for them. Writing isn’t just about putting words on a page. It’s how we figure out what we think in the first place. Some of our best ideas come from getting lost halfway through a sentence, taking a wrong turn, or making a mistake that leads us somewhere unexpected. That’s part of being creative. If we hand all of that over to a machine, what exactly are we left doing?

There’s another problem. When millions of people use the same handful of tools to write and communicate, everything starts to sound alike. We end up with beautifully organized sentences that nobody remembers. The machines get better at sounding intelligent while we get worse at being interesting. We clean up our mistakes, remove our awkwardness and polish away our personalities until we all sound like the same instruction manual.

Aldous Huxley wrote Brave New World nearly a century ago. I’m not sure our new world is particularly brave. It may just be a little creepy. Huxley imagined a society where people were kept so comfortable and entertained that they no longer had much reason to think for themselves. Perhaps we’re heading somewhere in that direction, only with better technology and considerably more screen time. Why bother thinking when a machine will happily do it for you?

Maybe the real problem isn’t that machines are learning to think. Maybe it’s that we’re getting too lazy, too impressed or too comfortable to keep doing it ourselves. We used to worry about machines taking our jobs. Now they’re taking our words, our conversations, our creativity and possibly even our love lives. Before long, two people may fall in love through AI-generated love letters without either of them having written a single word. That’s quite a romance.

We wanted machines to make our lives easier, and they probably will. But being human was never about getting every sentence right. It was about getting it wrong, trying again, making a mess, laughing at ourselves and occasionally saying something so dumb that it turns out to be brilliant. I’m happy to let a machine do the hard math, and I’d gladly let one do my taxes. But when I write, have a conversation, argue about something that matters or fall in love, I want to know there’s a real person on the other end.

Otherwise, we’ll end up in a world where the machines write everything perfectly, and nobody has anything interesting left to say.

Recommended Reading


Casa Carlini Simply Riemann audiobook variation, featuring the product packaging with branding and design elements.

1 comment

Will there now be a new edition of the book?

James Klagge •

Leave a comment

Please note, comments need to be approved before they are published.