It can write about a spark, but the content has no spark.
Probably my best and most insightful stuff has been produced more or less effortlessly, since I spent enough time/effort _beforehand_ getting to know the domain and issue I was interested in from different angles.
When I try writing fluff or being impressive without putting in the work first, I usually bump up against all the stuff I don't have a clear picture of yet, and it becomes a neverending slog. YMMV.
People asked for permission to repost it, it got shared on social media, it ended up showing higher in Google than a Time magazine (I think) interview of Bill Gates with the same title.
The problem is it isn't easy to detect it and I'm sure the people who work on generated stuff will work hard to make detection even harder.
I have difficulty detecting even fake videos. How can I possibly I detect generated text in plain text accurately? I mean I will make plenty of false positive mistakes, accusing people of using generated text when they wrote it themselves. This will cause unnecessary friction which I don't know how to prevent.
Second thought: Does it _really_ matter? You find it interesting, you continue reading. You don't like it, you stop reading. That's how I do it. If I read something from a human, I expect it to be their thoughts. I don't know if I should expect it to be their hand typing. Ghost writers were a thing long before LLMs. That said, it wouldn't even _occur_ to me to generate anything I want to say. I don't even spell check. But that's me. I can understand that others do it differently.
LLMs are naive and have a very mainstream view on things; this often leads them down suboptimal paths. If you can see through some of the mainstream BS on a number of topics, you can help LLMs avoid mistakes. It helps if you can think from first principles.
I love using LLMs but I wouldn't trust one to write code unsupervised for some of my prized projects. They work incredibly well with supervision though.
Important qualifier there. There's a massive oversupply of contrarian thinking; it's cheap, popular (populist), and incorrect. All you have to do is take some piece of conventional wisdom and write the opposite. You don't have to supply evidence, or if you do then a single cherry-picked piece will suffice.
I'd say something more like "Chesterton's Fence Inspection Company": there are reasons why things are the way they are, but if you dig into them, maybe you will find that the assumptions are no longer true? Or they turn out to be still true and important.
I would much, much, much rather read an article with imperfect English and mistakes than an LLM-edited article. At least I can get an idea of your thinking style and true meaning. Just as an example - if you were to use a false friend [1], an LLM may not deal with this well and conceal it, whereas if I notice the mistake, I can follow the thought process back to look up what was originally intended.
> I run a marketing agency. We use Claude, ChatGPT, Ahrefs, Semrush. Same tools as everyone else. Same access to the same APIs.
Since you use it for your job of course you use it for this blog, and that will make people look harder for AI signs.
Why?
I get using a spell checker. I can see the utility in running a quick grammar check. Showing it to a friend and asking for feedback is usually a good idea.
But why would you trust a hallucinogenic plagiarism machine to "clean" your ideas?
>We're not against AI tools. We use them constantly. What we're against is the idea that using them well is a strategy. It's a baseline.
The short, staccato sentences seem to be overused by AI. Real people tend to ramble a bit more often.
Not exclusive to AI, but I'd be willing to bet any money that the subheadings were generated.