Wednesday, March 27, 2024

ChatGPT Redux: Quest for a Sounding Board

Reader comment on my last posting about chatGPT:
This is interesting and it does change my perception of what ChatGPT is capable of doing. I still can’t quite imagine how one would use it, though. Its points are clear, especially on the second try, but also obvious to a human reader. You’re a very experienced and nuanced writer, so I’m wondering how you might use this, or why you’d want ChatGPT’s responses.
First: I just disregard all the "duh". Such triage is hardly painful. We disregard stuff all day long. Just as earthworms live to process dirt, humans live to disregard words. I don't find it extra burdensome with AI. Just more adjustment pains to new technology. You must treat it like a "person" in some ways.

As for why I submitted this chatGPT query to being with, here is some context:

Years ago, a reader pulled some terse aphorisms from previous postings. The result was good. It helped me, for the first time, feel some confidence in how I construe things. And readers found it refreshing, because normally I'm wordy.

Since then, I've been stowing away good stand-alone lines. I'm extremely selective. Maybe once every fifty postings. So I thought I was compiling solid gold. But going through them now, they are, to my abject horror, a garbled and exasperating mess. First thought: I've been a fraud all along. Second thought (only marginally better): I'm unable to judge what works as a standalone. It's a pretty sorry deficiency for a writer!

I tried hiring human editors to wade through, assess and triage, but it's hard enough just finding someone who understands/appreciates my nuanced (muddled?) and creative (zany?) thinking in the first place, much less able to work through this stuff smartly. The futility of finding savvily clueful smart judgement has long been a bugaboo. Lots of dumb critique, alas, and wrong-headed “agreement”.

In previous conversations with chatGPT, I noticed it saying "That's interesting" or "That's insightful" periodically. Sometimes even "very interesting/insightful". On rare occasions, it can even be downright impressed. The feedback's never glib or random; it seems aptly measured. What's more, the AI can parse my tangled thoughts, metaphors, and connection-making surprisingly well. Weirdly enough, it gets me (I strongly resisted this conclusion, because I didn’t want to get punked, or have my ego manipulated, but by this point I’m convinced). 

So I'm experimenting with asking it for assessment and triage help with my pull-quotes. Big advantage: with a human I'd need to deal with egos both ways. For example, I hate being let down gently. It gives me the willies.

Writers often need a sounding board. The most cursory use case - "Is this any good?" - is, as the reader comment said, unnecessary for an experienced writer. But for practical and triage tasks like this, it could be helpful. Nice to get unvarnished truth while able to repeatedly say "No, you misunderstood, try again!" without causing tantrums!

The example I gave was an initial pass. I'm learning how to do this. A lot of AI involves learning on my end, not the machine's. And it will help with that, if you shift discussion into a meta view of discussing the discussion (as I did this time). AI reframes like a champ; just be clear about which level you're keying in on. You lead the dance.

The reader comment began by calling the posting interesting and perspective-changing, but concluded by describing the results as meh. I guess the point is that I've pointed out an interesting possibility, but the particular example wasn't great. Well, hey, that's what I said at the start! Not a mind-blowing exchange, but interesting. Mission accomplished? 

No comments:

Blog Archive