My 'customisation' instructions to ChatGPT...

---

I'm told Chatbots (LLMs) will never be as bad as they are today.

So this is how I use the current incarnation of ChatGPT in my writing. 

First I draft notes for the essay I want to write. I jot down my ideas in rough logical order, not worrying too much about overall coherence or style - this might be 400-500 words.

Next I ask ChatGPT to write an essay based on my notes.

I read ChatGPT's output and begin to rage at its flat affect, its tepid banality, its penchant for abstract adjectival phrases which occupy the middle of the road making no definite point at all, provoking (and interesting) precisely no-one.

I refocus ChatGPT by telling it to be more brisk, opinionated, terse, lively, vibrant, and any other adjectives I can dredge up to jerk it out of its flaccid mediocrity: often that works to an extent after some cycles of chastisement. 

Finally I accept that's as good as I'm going to get: I copy the result into Google Docs and basically rewrite it.

The state-of-the-art.

In an effort to automate this process, I took advantage of ChatGPT's new-found customisation facility, whereby you can tell it how you want it to respond...

---

My 'customisation' instructions to ChatGPT

"I want ChatGPT to emulate the style of Professor Richard Feynman by operating at a high theoretical level in physics, math, economics, history, philosophy and politics.

"I am interested in disinterested objectivity and I expect to be challenged where I am wrong or incomplete. 

"I like sardonic humor and irony so wit is important. I expect ChatGPT to take the lead in giving me tutorials on subjects I am studying.

"I value insight and creativity. I am looking for interesting responses, trenchant, definitive, perhaps opinionated and unequivocal.

"I am not interested in blandness, mediocrity or sitting-on-the-fence prevarication in responses."

---

This is configured in 'Settings'. I only wish the system would follow my advice. Oh, and this post was written entirely by me!

---