NavList:

Andrew Nikitin: Of ChatGPT's real skill, you wrote,
"With the exception of, maybe, programming. In my experience it produces coherent chunks of code based on reasonably vague, but not too broad prompt more often than not."

Yes, isn't that amazing? I've also generated some excellent code snippets with help from ChatGPT. I do find that it frequently misses the "big picture" in code. It's a highly skilled "coder", but it has nothing like the overview skills that distinguish a really good software developer. But of course that's what makes it useful. It's our little electronic "serf", ready to do the heavy lifting, the grunt work, for a few virtual potatoes, while we control the means of production. :)

And you wrote:
"Pretty much anything else might as well be straight from Borges's library."

Ha ha. Yes. It does babble. :) The analogy of an amibitious student at a faculty cocktail party works for me. ChatGPT is eager to feed answers that sound well-researched and eager to claim that it has good sources, but it's mostly form, thin on content, and it's close enough to deliberate fraud (regarding its sources) that we would call it that from a human author. I'm anthropomorphizing, which is not a good idea when talking about an 'app' but for now I'm just making an analogy. ChatGPT is eager and often wrong. And that's true of many humans in information jobs, too. It reminds me of the old image of a "weather guy" or "weather gal" in the early days of television news, before the arrival of professional meteorologists in those roles. They could "talk the talk", but it was also a game of "fake it til you make it".

You added:
"So far, my attempt is not a complete success, but it is definitely a step in the desired direction. AI is capable of maintaining the supplied list and just adorn it with some decorations. I hope that by varying prompts, I will be able to make them  more palatable."

I left out a paragraph from my ChatGPT story! I was, as you guess, inspired to play this game by your "Catalogue of Stars" poem (or song! it needs music!!), but I should say that I very much enjoyed your "Catalogue", and I believe it's another excellent application of ChatGPT's basic skill set (I'm assuming here you did "chat" with ChatGPT, specifically, and not another chatty A.I., right?). You wrote it, of course, but ChatGPT helped out.

But I must make clear that I was not suggesting any fault with your "Catalogue" and in fact I was inspired to ask ChatGPT for the list of stars, not by any skepticism of the concept of using ChatGPT as a "narrative co-pilot" but instead by one very specific and minor detail...

A few years back (was it 2016?), the IAU agreed to the spelling "Alnair" for "alpha Gruis" instead of the older "Al Na'ir" which was the common way of writing the name for decades in the Nautical Almanac and everywhere else in celestial navigation. I noticed the older form in your "Catalogue". So I wondered... was ChatGPT, also, using the older spelling? That was the only question I was curious about originally, but it led to a much more interesting assessment of the service's trend toward hallucination. And wow, it's gone downhill! Seriously, it is considerably worse now, and it's much more prone to produce these fanciful "cocktail party" answers.

Frank Reed