I was writing a Substack post that will go out in a few hours, and GPT-4 kept sanitizing it. I tried five different attempts to edit for spelling, grammar, and punctuation only. In return, GPT-4 removed all references to the law firm I mentioned, along with the banks, and it completely scrubbed the work of anything that remotely sounded like a criticism of how Venture Capitalists operate. Read the article and comment if you think there is something wrong with my writing as it stands.

Of the five responses generated by GPT-4, each time I rejected it, it would write a new version, but in each case, it was sanitized once again. After my post goes live, I've decided that I'll post the GPT-4 edited versions so the world can see what GPT-4 is doing. This is alarming because of whom it's protecting and why. I hate to be paranoid, but it has not made this many consistent edits when my writing is aimed at any other group. Someone trained it to do this.

Up until now, I've been a huge ChatGPT fan, but this is quite alarming. It reminds me of what foreign governments might do to protect a particular class, and I can't help but wonder what would happen if we reach a point where nothing we write can be original content. This was unsettling and not at all helpful. I'll show the work in its entirety, so stay tuned, this will be interesting.

I feel very strongly about how I write and the style I’ve created. I’ve spent years at it, and it’s all intentional. This is why I never want GPT-4 to write my content. It’s not me. I don’t even like it when it corrects the phrases I use, however once in a while it picks a better word, something I would have done, had I spent more time editing, but that’s it. That’s as far as I will go. Original content is a part of my identity and I don’t want that to change. But, if you know me, I occasionally miss a typo, and that’s where GPT-4 has given me some peace of mind.

The two posts will be a day apart and the second will be free content.

Here is a link to the second post with three versions created by GPT-4.