When someone does your push-ups, can you feel your muscles flex?

In June, IPO-bound Anthropic accused Alibaba of stealing from it. Or, to use the Silicon Valley euphemism, 'learning' from it. Oh, the exquisite irony. Anthropic trained on the lifetime work and knowledge of others without paying them a penny, and is now complaining about others training on 'Anthropic's knowledge' without paying them a penny. Apparently, unauthorised data harvesting is a brilliant tech miracle when they do it, but industrial espionage when it happens to them.

Better still, the people whose work Anthropic trained on for free are now being charged to use Anthropic. It's the first business model I've seen where somebody borrows your notes, sells them back to you, and then asks for a tip.

I was recently on the receiving end of similar treatment. AI has read a lot of what I've written, and I'm personally guilty for some of that input. So, when I entered an article I wrote entirely myself into an AI detection tool, it thought 27% of it had been written by AI. Apparently, after spending 20 years developing my own voice, I now sound suspiciously like a machine that learned how to sound like me.


It's a fascinating business model. A bit like a tribute band accusing the original artist of sounding too much like them. The next logical step is for karaoke singers to start issuing cease-and-desist notices to Adele.

Personally, I'd quite like a button that lets me analyse AI-generated content and declare that 27% of it appears to have been written by Papa CJ. However, the whole thing raises an interesting question: are we using AI? Or is AI using us?

I use AI regularly, mostly for research. Given my historical relationship with grammar, I remain grateful that AI arrived before anybody created a public archive of my school essays. What fascinates me is something else entirely. The more powerful AI becomes, the more tempting it becomes to outsource the part of life that actually matters: the thinking.

Ask AI a question and it immediately produces 17 perspectives, 12 frameworks, 1 historical overview, 3 counterarguments, and a summary so comprehensive that you begin to wonder whether having an original thought is now considered inefficient. I wouldn't be surprised if at this rate, independent thinking will, at some point, be referred to as a 'legacy feature'.

But somewhere along the way we've started confusing access to information with insight. They're not the same thing. Information tells you what people think. Insight comes from thinking yourself.

For most of human history, this was unavoidable. If you wanted an answer, you had to sit with the question, read, reflect, and occasionally change your mind when you discovered that your strongly held opinion was complete nonsense. Nowadays, you can skip directly to the conclusion and have AI deliver your intellectual Happy Meal.

What happens when you are deprived of the process? You may get the answer. But you don't necessarily become the person capable of finding it. And it's the struggle that helps you become that person. That struggle isn't wasted time. It's where you learn how to do the thing, how you evolve and grow, and often, what you learn about yourself along the way.

The value of solving a problem isn't simply getting the answer, but developing the ability to solve the next one. Otherwise reading a map would be the same as making the journey.

AI may eventually be able to write a book for you. But what it cannot do is care, doubt, or lie awake at 3 a.m. wondering whether it's wrong. Usually 5 seconds after pressing 'Reply All.'

So, the next time AI gives you an answer, pause before accepting it. Not because the machine might be wrong. But because it might be stopping you from discovering something for yourself. And that would be a very human thing to lose.

Personally, I prefer to write without the aid of AI because as an artist, I have to feel it. That can only happen when I create it. Because what AI creates, doesn't have me in it. Well, actually, it does. But only 27%.