Jordan117 comments on The american gift economy

[–] Jordan117@lemmy.world -1 points 14 hours ago* (last edited 14 hours ago) (3 children)

Yes it does, and there's no need to be a dick about it.

The story itself is just off. It's not how anybody acts or talks in real life. America is not a gift economy where strangers gladly spend time and effort and gas moving furniture and heavy appliances for a few slices of pizza (and laughingly refuse payment). Ain't nobody got time for that. Struggling to guilt friends into doing that kind of grunt work for pizza has been a cringe humor trope forever.

The short staccato sentences. The constant anaphora (repeating "Kevin" or "Tony" half a dozen times in a few short paragraphs).

The overly dramatic ALL CAPS PRONOUNCEMENTS.

The oddly-specific numbers ("fifteen mile radius", "eleven minutes")

The expected not X, but Y constructions and reversals ("which felt rehearsed and was not")

Weird serious emphasis on unremarkable things nobody would care about (sitting on the couch first, the daughter being aggressive)

Why would a man you've never met ask you if you could help them move?

Etc.

These are all legitimate rhetorical tricks, but the way it's laid on so thickly and uncannily in short punchy sentences screams AI, specifically Claude's particular style of prose.

Also, GPTZero and Pangram both say it's 100% AI generated.

Also, they've literally said: "I'm simply a Japanese creator who writes in Japanese, uses translation tools and AI as creative tools, and tries to tell funny stories."

permalink

fedilink

source

parent

hideshow 3 child comments

[–] ChunkMcHorkle@lemmy.world 3 points 8 hours ago (2 children)

Jesus fucking christ man. You people. A preponderance of observed characteristics is a factor, but not itself fact.

For myself, I am so goddamn tired of being called AI. I have NEVER used AI a day in my life. Why? Because I know how to write, I write well, I write for fun, and I've been doing it daily in one way or another for the better part of a century.

Go look at my post history. You'll see much of what you just called AI in all my posts. Then look at the time stamps: it takes time to write. Then look at the imperfect construction and grammar: good, but by no means perfect.

Then come back here and tell me I'm AI because I know you will. Go ahead. Let's get it over with.

permalink

fedilink

source

parent

hideshow 2 child comments

[–] Jordan117@lemmy.world 5 points 6 hours ago (1 child)

Idk why you're getting so offended. I never implied you or anyone here was AI, just the Twitter post in the screenshot.

I also said the techniques mentioned are legit and common. I'm even guilty of some AI favorites myself (em dashes, false ranges, etc.). That's kind of the point -- AI is trained to emulate human language.

The thing is, they're also heavily post-trained to reward "interesting" or faux-profound rhetorical tricks that impress human graders. This encourages them to overuse such techniques, even when they don't make much sense. It's like those old Cola Wars taste tests where people preferred a sip of Pepsi because it tasted sweeter on the spot, but it becomes cloying and saccharine over a whole can. Default AI prose reads like it's trying to twist every single sentence into some kind of surprise or artful flourish, and usually the same few kinds over and over and over again. Seeing those well-worn patterns, combined with the pretty questionable understanding of normal human behavior, makes it pretty clear what's going on.

FWIW, your comments don't read like AI at all. I do agree there's a frustrating epidemic of such accusations, especially from people who don't understand good writing lobbing it at people who are capable of putting more than one clause in a sentence. But I'm being sincere when I say that this has very specific and characteristic traits of a particular AI model that I've seen a lot, not just "it's long and well-written so it MUST be fake."

permalink

fedilink

source

parent

hideshow 1 child comment

[–] ChunkMcHorkle@lemmy.world 5 points 5 hours ago* (last edited 5 hours ago)

Funny, I did not expect to but I actually agree with most of you said. Primarily because to me, as a completely non-AI writer, you pegged what I personally see as the unmistakable flags of AI content: non sequiturs of language and richness of concept and ideas.

For myself, when I am looking for AI, I do not pay nearly as much attention to specific techniques as I do to the way it feels as a whole, reading it carefully as if someone were speaking to me personally and noting the oddnesses, the tiny bits that are just huh? as you move by them because in some tiny and otherwise unnoticeable way they just do not follow. They are not what a natural speaker would say, even a natural speaker of a non-English language who put their text through a translator and then posted it. Some things just don't follow, but make sense to a pattern recognizer. I think that's what you're alluding to by "questionable understanding of normal human behavior" and I agree.

Thus, in light of what you replied, I actually do now see your point about OP's posted text, but the reason it did not read as AI to me personally is because it passed this personal non-sequitur test. When I read it, it brought in the unexpected and it did turn the same rhetorical tricks you cited, but not in a jarring or "uncanny valley" way. Then again, I am also not on social media and never have been, so there's that: to spot counterfeits, don't spend time with the fakes; spend time with the real thing. Being away from sound bites and short form content helps a lot. That doesn't mean it isn't AI, only that it did not read so to me.

I also look for what I personally call "idea richness," the cumulative number of separate (related or unrelated to the primary) concepts and ideas that are interwoven into the text but are not presented as part of a list or itemized rebuttal. To me, that richness is part of what makes something worth reading: does it introduce secondary images and ideas? Metaphors and similes? Bring unrelated ideas and concepts into contrast? Use the five senses? Include technically unimportant but still related lived experience? Demonstrate a wide and curious intellectual history? I could go on, but you get the point. Does it add tarragon to the fish?

AI fails miserably at all that. AI at this time could not actually write this comment, no matter how well it was prompted, because it is an exploration of one person's lived and felt experience. AI is a pattern recognizer, not a human feeler, senser, and wildly imperfect thinker like me. In terms of idea richness, OP's text didn't pass and didn't fail that one for me, so I defaulted to not-AI.

Speaking of which, anymore (and I might be odd for my age because only the young people are telling me the same) along with all the above I am starting to get a felt sense of revulsion for AI content, one that I chalk up to my putting so much effort over time into detecting the human content, taken in addition to my burgeoning aversion to the fake. I have a great foundation for spotting counterfeit prose because I have read so much real prose, but my revulsion to it is growing apace with learning to spot it. The young people I know (high school, early college) are telling me they feel the same, though for them it's more about hearing it in music than reading it in text.

It will be interesting to see what happens next, because as possible as it is that AI is moving toward content that is superficially indistinguishable from human product, it is also possible in some vague way that none of us can wholly explain yet that humans are also becoming capable of detecting non-human content by purely animal means.

And you're right, I was getting offended, but you are not the correct target. Thank you for taking the time to reply in earnest. I appreciate it.

permalink

fedilink

source

parent