Blog writing | Three AIs have a hack
Having had a tryout with three AIs in the last six weeks since I discovered them, it might prove interesting to put them head-to-head, so I asked them to comment on and give me their best version of a blog post.
It’s been revelatory, mostly about how important the prompt is in setting the parameters for analysis.
Qwen was helpful. It told me what worked, what needed to be tweaked (I had deliberately left a few dictation mistakes), but then threw in two major changes that twisted what I had to say.
The first change is pretty bad. It added part of the prompt into the text, embellishing a simple description of my writing process into a personal philosophy. Just not mine.
Second, it “punched up” my simple ending into a sermon, which is not exactly my style. Don’t change my intent, buddy!
ChatGPT was an overbearing, pedantic editor who knew better than me. Yuck!
It added a phrase that ideas “rattle around in my head”, which is not a good metaphor. Empty skull, no brain, shrivelled peas instead. Surely my ideas are the streamlined expression of a deeply experienced mind!
Its proposed title was pompous and more like clickbait. ChatGPT also included my prompt instructions in the text. It was like being back in school with the teacher lecturing me while adding a few hundred words in a narrative arc it invented to “help” me. It rewrote my blog post completely, stifling my own writing style, and trivialising my experience.
So, Gemini. This pointed out the dictation errors, correctly assessed that I needed a “parting thought” that would resonate rather than a “part thought,” but it didn’t change that, or much else. It was the necessary light touch.
Gemini concluded: An AI tool should clean the dust off the window, not rebuild the house.
And I have to agree with that.