I wasn't crazy about the Woodrell passage as a whole (although, yeah, "flourish upon flourish of small popping sounds" is pretty great) -- but then as you suggest, I bet if I'd read the preceding 25% of the book I might feel differently. To my taste, the passage sounds like the product of a writer who has learned too well the lesson of using vivid verbs. (I wonder if Ron Charles has read 'Winter's Bone'?)
One of the things I'm finding interesting about using Chat (Claude, in my case) is that while it can analyze prose brilliantly, it can't actively use its analytic skills to vet its own writing in the process of writing it. Most (?) human writers keep the analytic machine going in the background while they're writing, so when they write, say, "I was there," something immediately clicks in their head which tells them the sentence just doesn't say enough. I once asked Claude if that split in function (writing vs. editing) is actual or imagined and it said yes, that's how it works.
Since then I've submitted some of Claude's longer responses to Claude itself to see if it picks up on the same things that sometimes bother me about it, stylistically. (Claude's "incognito" mode works well for this, because it then won't know it's analyzing its own product.) But I'd really like to come up with a prompt sequence which says something like, "Generate [a full-length response to Prompt X], but don't share the response with me right away; save it, analyze it [as, say, Tom Shroder asked Chat to analyze various writing samples], and FIX it according to the results of that analysis; and, finally, present the fixed response to me as if it were the initial response to [Prompt X]." No idea if that's possible though.
I’d guess something like that would be possible. But I think it’s easier to point out all the parts of its responses you fault and express what you don’t like about it and ask it to avoid in the future. It has responded to that well.
Thanks. I keep seeing admonitions not to tell Claude what NOT to do -- e.g., provide lists of taboo words, phrases, and grammatical constructions -- because then it expends too much energy trying to come up with substitutes. But then that displaces onto me the need to provide alternatives to use, which seems to defeat that purpose of having an LLM at my disposal.
You're a generous guy, Tom. You must've been a delight to write for.
Well, in El Paso, Texas, area we have three AI behemoths under construction - two civilian, one military. The sources of water and power are a little vague. Let's conserve energy! Don't tell Claude what words to avoid.
I like the phrase "look at this line." My mother and I email back and forth about malaphors and mistaken or overwrought writing, including what we call unfortunate collisions with a thesaurus.
Just yesterday I read an article that stated someone's recovery was a long road but he finally got to the other side. Perhaps Chat could have said huh? in reviewing before publishing.
This is of course a high level of literary analysis. I guess it is. Would I be interested in such an evaluation, by human or machine, of my favorite novel? Probably not. But I am very impressed.
I wasn't crazy about the Woodrell passage as a whole (although, yeah, "flourish upon flourish of small popping sounds" is pretty great) -- but then as you suggest, I bet if I'd read the preceding 25% of the book I might feel differently. To my taste, the passage sounds like the product of a writer who has learned too well the lesson of using vivid verbs. (I wonder if Ron Charles has read 'Winter's Bone'?)
One of the things I'm finding interesting about using Chat (Claude, in my case) is that while it can analyze prose brilliantly, it can't actively use its analytic skills to vet its own writing in the process of writing it. Most (?) human writers keep the analytic machine going in the background while they're writing, so when they write, say, "I was there," something immediately clicks in their head which tells them the sentence just doesn't say enough. I once asked Claude if that split in function (writing vs. editing) is actual or imagined and it said yes, that's how it works.
Since then I've submitted some of Claude's longer responses to Claude itself to see if it picks up on the same things that sometimes bother me about it, stylistically. (Claude's "incognito" mode works well for this, because it then won't know it's analyzing its own product.) But I'd really like to come up with a prompt sequence which says something like, "Generate [a full-length response to Prompt X], but don't share the response with me right away; save it, analyze it [as, say, Tom Shroder asked Chat to analyze various writing samples], and FIX it according to the results of that analysis; and, finally, present the fixed response to me as if it were the initial response to [Prompt X]." No idea if that's possible though.
I’d guess something like that would be possible. But I think it’s easier to point out all the parts of its responses you fault and express what you don’t like about it and ask it to avoid in the future. It has responded to that well.
Thanks. I keep seeing admonitions not to tell Claude what NOT to do -- e.g., provide lists of taboo words, phrases, and grammatical constructions -- because then it expends too much energy trying to come up with substitutes. But then that displaces onto me the need to provide alternatives to use, which seems to defeat that purpose of having an LLM at my disposal.
You're a generous guy, Tom. You must've been a delight to write for.
Well, in El Paso, Texas, area we have three AI behemoths under construction - two civilian, one military. The sources of water and power are a little vague. Let's conserve energy! Don't tell Claude what words to avoid.
"Are they intoxicated by their own language?"
I so wish I could have wielded that one when I was a junior editor trying to corral far older and wilier correspondents.
I like the phrase "look at this line." My mother and I email back and forth about malaphors and mistaken or overwrought writing, including what we call unfortunate collisions with a thesaurus.
Just yesterday I read an article that stated someone's recovery was a long road but he finally got to the other side. Perhaps Chat could have said huh? in reviewing before publishing.
This is of course a high level of literary analysis. I guess it is. Would I be interested in such an evaluation, by human or machine, of my favorite novel? Probably not. But I am very impressed.
Yes it’s true you don’t want to do an autopsy on every beautiful work of art. Kind of kills the buzz. But if you need to dissect, AI is proving adept.
This is one of those instances where I read the book and much preferred the movie
I’ll look for it in streaming
Jeez, who needs a book club anymore?
Exactly. If you’re mainly interested in the wine drinking part though, you may want to keep the human version.