No.6944
Which one of those you find better?
Please vote before checking spoiler; if you want to write some thoughts about any of them, it would be better to do that before reading spoiler as well.
The left one is fully human-made. The right one was created by feeding the left one to gpt-4-1-mini model asking it for opinion on what is wrong with the left poem, then following its offer for a rewrite.Does your opinion change after knowing the method of production? If you want to comment on this aspect, please spoiler it and please don't change your vote.I'm interested to see what feelings and opinions are on AI-generated poetry. I know that the comparison might not be the best way to test that, due to being reliant on left poem's writer skill, but it offers a chance to see whether people spot something "wrong" with AI-generated poetry or not. No.6945
I mean if you wrote a kuso poem on purpose, what do you expect?
No.6946
Voted left.
Didn't vote the right one because it felt less personal about Aya and focused more on the ordinary egg in the last part, the right one reminded me of some stuff I read for high school a very long time ago, in the way it's written. I'm also not very good at interpreting poetry, though.
No.6949
I liked the left one a lot more until "With no future, uncracked it lies," then I feel like its sense of rhythm kind of unravels. The words stop falling into place on the meter quite so neatly and I had to read a few of the lines several times over to make sense of them.
The right one I think doesn't have as many exciting lines as the one on the left does but its rhythm never stumbles (except the one line that's in iambic pentameter for some reason) so it's overall easier to read.
No.6951
>shine can't match your shining eyesLeft.
>>6949Right's easier to read and all but it makes me think of studying poetry in school, like others noted. I hated poetry class...
No.6952
>>6944they're both kind of bad, but left is bad in a less boring way
No.6972
I apparently don't have enough posts to vote, but if I did I'd vote for the right one. In my (uneducated) opinion, that one flows way better, while the left one just feels like it is trying too hard to get fucky with the grammar while failing to make the end product actually sound good.
No.6974
Like anons have said, left for feeling while the right is a little "cleaner."
Specifically, the first two lines of left, a mishmash of left and right for the rest of verse 1, right entirely for verse 2, and for the last verse, left's idea with right's execution. Also, "ordinary" ruins the meter. "Normal" scans much better.
It reminds me vaguely of John Donne. You should hop on Gutenberg and get familiar with his songs if you're not already.
Well dammit. I can kind of tell the difference now. I think the thought of "left's idea with right's execution" on the last verse has something to it. The emotion and sentiment is present there, which is the more important thing as the form can be improved later, whereas right is more superficially "pretty" but means nothing. I prefer your own 2nd verse as well now. That feels very wishy-washy to me, but that's exactly the reaction you said you want to test for here. It also isn't until now I checked the rhyme scheme, how you're actually using something a little more interesting vs AABBCC.
Cheers OP. We should all be writing more crowetry.
No.6978
>>6944First thoughts before spoiler:
Left is a bit awkward (I start to feel like I'd do better) but has a nice flow. So, it's inspirational.
¥ sounds of caws I crave 10/10.
Right aims for the kind of dry mastery I just don't care about.
After spoiler:
Yuup, I see why the right's feeling metadogmatic and rigid. No.6980
>>6950>Poetry is meant to be recitedI don't agree.
No.6981
>>6979Hopefully you didn't get spoiled!
No.6984
>>6978 (Me)
>the kind of dry mastery I just don't care aboutIMO that's also the issue with all
AI "artistry", except the odd curiousities we had seen early when earlier models outputted Lovecraftian horrors.
It could pass as human stuff, but ~99.99% of times it would pass as the human stuff that's just not interesting, unless you're new to the medium. It's evolutionary encoded in our gut that it's a simple truth: art is self-expression; so what "self" non-AGI AI would even try to express? No.6987
>>6980NTA, but I think it's a generally valid expectation.
First of all the origins of all poetry were songs meant to be accompanied by instruments, chiefly of them lyre, hence lyrical poetry. The roots of form are in musicality and sounding good when said aloud.
Second of all, I believe that meaning of any single poem can be fully translated to prose, both in semantic layer and emotional structure. If this holds, then the value of poetry becomes purely in the form the message is presented, not in the message itself. While graphical poems exist, for most poems the value of their form will be in the way they sound when recited.
Finally, somewhat related to the above, but not necessarily requiring such strong thesis as full poetry-prose translation, is the fact that showcasing art is an act of communication and all acts of communication intend to tell something to the receiver. Because art is a less straightforward way of communicating your intended message, requiring much more effort to parse, there must be value to the form of message beside the essence of the message. Both of those poems are basically waifu-posts that could be done instead by writing plainly: "I want to have child with Aya", it's the form that decides their worth. Their form, like for most poems, clearly tries to make the way they sound as pleasing as possible.
No.7012
>>7010If you made the one on the right, why does it have the one random line in pentameter? Please take your craft more seriously in the future, もう...
No.7013
>>7010Well from the thread we can say that you've got a long way to go before you can beat AI, and also the AI's got a long way to go before it can be good.
No.7015
>>7010>telling it to make mistakesGreat going retard. So this whole thread was pointless.
No.7016
>>7013Also me for reference
>>6951I didn't cheat. I just think using the same word twice is unforgivable.
No.7017
File:55.jpg (697.18 KB,800x1127)
>>7010I really hope this is a troll and not the OP. I'm sure I've seen this poet posting before on 4/jp/ and there's no way someone so devoted could use Aya as a front to be so impure and dishonest.
No.7020
>>7018bit that frilly jew-nosed bastard
No.7021
¥Polling and Honesty
¥Honesty
No.7022
>>7016>I just think using the same word twice is unforgivable.but then... why did you do it? i'm with
>>7015, this is rather tarded (and i say that as someone who didn't vote, so no emotional investment)
No.7023
>>7022I'm not a poet lol I just took a thesaurus and made sure it looks like it flows well, I didn't quite even notice the shine/shining thing as a problem or a possible problem until anons pointed it out.
Like I said, the point of this experiment was not to hone my poetic skills, but to see what anons think about AI generated poety. Apparently it has nice sentiment, has a nice flow, is inspirational, interesting, and all things good.
No.7026
>>7024Aya is known for exaggerations and misinformation....sear
ch your feelings, you know it to be true
No.7027
>>7010Conclusions
if true as per (You)r post:
1. OP writes poetry like a soulless pseud.
2. AI content can't pass as human content without extreme human refinement by human sensibilities.
3. AI can't do proper rhythm even when not told to specifically butcher the rhythm. No.7029
An actual OP here, this is
>>7010 some tanuki that must had noticed that I've done a reset of my cookies. Don't believe stinky dog's lies! Per what I've mentioned I can't provide a direct proof of being OP, the next best thing I can do is picrel,
the way I drafted the text (left side earlier).I actually considered telling that the left one was generated, while the right one was human made, to see whether people would spot or at least feel such deception at play, or would they rather give convincing post-hoc rationalisation, but in the end I decided it would be too convoluted at this point and it would be hard to know what exactly was measured. No.7032
>>7030¥ OP had to botfag to get 30+ replies shame-imaru on you
No.7038
>>7037the harder you fight, the greater will be your humiliation and my sexual gratification once you lose to the AI!
oooooho I'm getting all horny just thinking about it, anon!
No.7039
>>7038dude i'm doing a corpo training for making your own agentic networks and even these have premade prompts including "DON'T HALLUCINATE", "DON'T MAKE UP STUFF" verbatim
i'm not gonna lose to this retarded shamanic black box clanker
No.7040
>>7039that's what she said
No.7041
>>7039>"DON'T HALLUCINATE", "DON'T MAKE UP STUFF"How could that help? Stochastic models don't have semantic understanding from what I know, this should be nothing more than some additional weights to output, potentially even detrimental. Unless I'm still in stone age regarding this tech.
>i'm not gonna lose to this retarded shamanic black box clankerYou are asking for instant loss 2koma now, you know this?
No.7042
>>7039what are they trying to use the NN for? The best use I can get out of them is autocompletion for coding and to give me 101 level info on a wide variety of topics while summarizing the first articles in finds in a google search
No.7046
>>6944The one on the right feels more poetic, the one on the left is stilted and hard to understand. Maybe there's charm in spending the time to try to understand it, but if you're asking which is better, the one on the right definitely is.
No.7054
>>7041>How could that help?It just does - it hallucinates in the less important contexts instead
NTA
No.7177
>>7036i can’t believe a /jp/ meme crossed over to japan