- Member Since 5th Dec, 2018
- offline last seen May 4th
SockPuppet
I like writing about the worst day of a character's life; it lets us see the mettle inside. (Pronouns: RB/20 )
SockPuppet
Joined 5th Dec, 2018 ·
Offline
Also LikedSimilarAuthor
- SockPuppet · 1.1k words · 41 8 · 882 views
- Samey90 · 2.3k words · 40 4 · 419 views
- Samey90 · 1.9k words · 125 3 · 1.4k views
- RB_ · 1.6k words · 80 10 · 787 views
- FanOfMostEverything · 2.8k words · 35 4 · 391 views
- SockPuppet · 1.5k words · 64 6 · 995 views
- ARandomLonelyDude · 1.5k words · 39 0 · 482 views
Featured In11
Scootaloo3,267 stories ·2,574 members
Sweetie Belle2,666 stories ·1,642 members
Apple Bloom is Best Pony1,692 stories ·931 members
Children of Equestria1,511 stories ·1,032 members
Sweetiebot136 stories ·702 members
- Rumble316 stories ·204 members
Sweetie Belle is the best pony479 stories ·195 members
CMC Clubhouse2,137 stories ·971 members
Join our SubscribeStar to remove these adverts!
Join our SubscribeStar to remove these adverts!
Stats
Page generated in 0.089 seconds
Total duration
1,054 users online
1,769,225 hits today, 1,870,739 yesterday
FIMFiction
My Little Pony: Friendship is Magic Fanfiction
Designed and coded by knighty & Xaquseg - © 2011-2024
Follow & Support Us
Support us
SubStar
Chat!
Discord
Follow us
Twitter
MLP: Friendship is Magic® - © 2024 Hasbro Inc.®
Fimfiction is in no way affiliated with or endorsed by Hasbro Inc.®
"Kettle Corn is dead!?"
"Yes! She died!"
Anyone else feeling like Miss Cheerilee needs to start teaching some classes in ethics?
No?
Just me?
Adorably daft, and about as well-thought-out and useful to the society as the real automatic plagiarism engines and their extruded poetry-like product.
Edit: Wow, did I offend someone who thinks these "AI"s are worthwhile and valuable?
Good thing to wake uo today, good sir!
Unironically the best haiku I've ever read.
11893796
LOL, must have.
Remember, one should judge their own accomplishments by the enemies they make.
11893875
Although I enjoyed this story, I refuse to become your enemy just to flatter you.
I also think that sometimes people earn enemies through NEGATIVE accomplishments, which this story isn't.
Skeedaddle is an earth pony? I thought he's a unicorn.
Unless you specifically made him an earth pony for this story.
Wut. Why? They couldn't put it anywhere else?
HAH!
11894206
That is the traditional location for 'interface ports', is it not?
One wonders if it's USB or something else.
11894282
Anyone who does that seriously needs a girlfriend, unless it was a girl who made the designs, than I don't know what to say.
All I can say is anyone who does that is a pervert and needs help, nobody puts that there without thinking how "dirty" it is.
Although it is a little hilarious.
Imagine someone made a fic poking fun of ai generated fics
But uses an ai to generate part of the text while someone is commenting on it like "dude really?"
11893875
Eh im kinda neutral on the whole ai generation thing
Yeah i hate passing it off as your own work
But i have seen some cool images made by ai on DA sooo yeah
Sweetie! I've heard about crapping out a story, but...
11893796
Speaking for myself, it's that plagiarism has rather specific definitions that perceptron machine learning is physically incapable of meeting. Even in limited dataset cases where it is trained SOLELY on a single specific body of work, it's still working with patterns derived from the input data rather than the data itself. For many of them, the model doesn't have enough inputs to take the entire image directly, meaning it has to rely on pre-processing and approximation before it can brute-force the identification of patterns, making it even more distant from the original artist's images.
The underlying methodology most resembles making a collage to use as a guide for artstyle imitation. Nobody in their right mind would consider a person who does this to be ripping off the artist they're copying the style of or tracing over any one source image of the collage, but because some nerds figured out how to automate such a process we get legions of technically illiterate lusers (spelling intentional and important) panicking over it.
Edit: The focus on the image generators is because they're the case with a clear analogy. It's much harder to analogize the data convolution of Large Language Models, which function much the same way but with text as the input.
11894461
Nuh-uh-uh-uh, I can confirm the text-based AI generators at least are still clearly plagiarizing, because I've seen plenty of other people who've already checked and found plenty of instances of where selections of the text they produced were very clearly just straight-up lifted from another source.
An example pulled from an article that touches upon this:
cdn.mos.cms.futurecdn.net/r93UEjAUKDTy5Eyp4j4r9U.png
And I'm no expert on plagiarism laws, but I know enough to know doing that without citing sources at all (which most AI do) is very discouraged in the writing world for obvious reasons.
Which does still raise certain implications about the image generating side of the equation since the methodology is still comparable despite the change in medium...but that one is indeed harder to definitively prove in so clear cut a case so I won't press that matter on this occasion.
In any case, a court of law is almost certainly going to have to step in and rule whether it actually is or isn't eventually. But it does still feel kinda scummy that an AI would be allowed to take all that content from other people and repurpose without having to give any credit or acknowledgement for its own use, and often times solely for profit. And when you consider how many people are trying to use AI to pull scams exactly like the one the story tells, either to profit or to defame on somebody else, and expect to get away with it without consequence...it does kinda give one pause.
At the very least, AI definitely needs some ground rules and regulation so to curtail stunts like what our cast of characters in the story tried to pull, if nothing else.
11894530
I like the idea that machine-generated text and images etc. can't be copyrighted. That would immediately kill the appeal for big business, and eliminate most of the issues.
11894530
Only one of those is a complete sentence shared, and "the same sentence exists somewhere on the Internet" is not in fact meaningful evidence of plagiarism. There are only so many ways for text to convey the same information, it is actively inevitable that this will result in accidental quotation. Especially when looking at legal matters like this because the subject has extremely standardized language and frequent use of quotations.
Does the article actually show that comments on Violinist.com could be in the training dataset of the model Google uses? Not even the full chain of scrapers to positively verify such, merely plausible reason for that particular "source" to actually be input? Because the partial overlap with extremely sparse sources that don't actually take the whole sentence looks an awful lot like a rag quote-mining to fearmonger if it doesn't go into this.
Again, legions of technically illiterate lusers. You just look at "articles" and throw a screenshot that's actively bad at proving the point, not a single word of any underlying function. Of course a machine designed to recognize and recombine patterns will generate outputs resembling prior discussion, but for it to be plagiarism that prior discussion has touched by the system during generation, and that's simply not how the technology works.
Legal subjects make constant use of extremely rigid textual composition alongside quotation of phrases without consistent attribution, because the wording needs to very thoroughly represent a specific meaning and yet source placement, method, and frequency vary considerably. If it were plagiarizing, the original context would come readily, but it instead makes up reasonable-sounding fakes whenever people try to get citations out of it.
11894622
Ah, yes - the "But I changed some of the wording" defense. Bet your teachers just loved to hear that one.
11894286
Better a printer port than an emotional support device.
Revenue and ethics tend to have an inverse relationship. In other words, line may go up, but line also go down
But hey, at least that weird Ebon Musk colt didn't get involved. Then the poet's hooves would be able to chop carrots and they'd rust in the rain.
Delightful bit of silliness. Thank you for it.
11894622
For the record, this is the article the image I shared earlier came from--Link--Among other things, it discusses in more depth the significance and the implications of said image, so I'll let it do the talking rather than reiterate it myself.
Whatever the case, the problem isn't so much with the AI itself anyway, but rather how its being used. The fic here (which, btw, was a fun read, thank you for it, Sock
) actually demonstrates a very real thing people are trying to do with AI right now...except they aren't usually waiting until the rights owner is actually dead like this, and it's all just so they can cash in on another's fame and success. Or, conversely, try and defame that other person for whatever reason. I've even seen some that do both. And no matter how you slice it...that's really not cool for somebody to be doing, and as such, I think creators are well within their rights to be objecting to it.
Hence why I'm confident some kind of court ruling one way or another about it is going to have to come about, because currently the laws aren't clear enough on the matter and it's allowing for wiggle room that people are, not without reason, wondering if it should be allowed. Most of these laws weren't written with AI in mind anyway, so even if the courts all universally take a pro-AI stance (we'll see), they'll probably still have to update the laws in some manner accordingly.
11894678
...The actual grammar is altered for the incomplete sentence overlaps, with statements placed differently throughout the paragraph. That is not "changed some of the wording", that is "put in my own words". That somebody else used some of the same words in some of the same order to describe some of the same things at some point does nothing to imply plagiarism without demonstrating use as a source, because a given piece of information can only be phrased so many ways.
I ask again, does the article make any attempt at linking the "sources" to the training dataset? Or are you standing by "somebody else, at some point, somewhere in the world, posted the same sentence online" as proof of plagiarism?
Edited-in response to same-minute reply:
11895159
Oh, so it is just "those words have been typed before, therefor it came from there" without any shred of evidence for dataset presence!
Wooow. You know, I was joking, but the fact you said that unironically, implying you at the very least might actually try that argument on a teacher or similar professional is... honestly kind of impressive.
Still, I can see there's been a misunderstanding. You appear to be operating under the misapprehension that my post was intended as a sincere argument. And, for that matter, I might have tried to put one there a year ago, before it became painfully apparent just how futile any effort always is to try and correct the infantile understanding of theft and plagiarism that abounds within AI enthusaist circles. Indeed, it is tempting to pick up on how clearly you don't understand what is and isn't plagiarism and how you are sincerely arguing "When AIs say the same thing in the same way as the thing they are directly learning from, it could be a coincidence!", but there'd be no point. As such, I am not, nor will I start, arguing with you or explaining anything to you.
I'm simply making fun of you.
Now, you may feel free to object to this, but given how freely you dipped into insults unbidden, it might be a bit hypocritical. Although, if so, while I might at first think to say you, as the saying goes, 'can dish it out but can't take it', well... if 'technically illiterate lusers' is the best you can come up with, I'd only be half-right.
11895188
I think that's a good stance to take on this at this point, so I think I'll join you.
11895188
Plagiarism only applies to outright copying another's work to pass off as your own. Do you seriously think that "These recordings are subject to copyright infringement penalties, regardless of whether they are made for personal use or for commercial gain" is copying "and is subject to civil remedies "to the same extent as an infringer of copyright""?
These have different information (some not shared is bolded) arranged differently (start of new sentence vs. end of sentence continuation), using different words (do I really need to explain that?) and no indication the former has any connection to the latter in particular. How is the former copied from the latter? Why in the world am I supposed to take the article as a whole seriously when such a total lack of resemblance is included with an identical sentence as "plagiarism"?
Failure to cite sources is the problem here, which is not even vaguely against the law in general discourse. In academic or legal contexts it carries certain liabilities, but it is not plagiarism outside the rather narrow case of direct quotation. Which two of the three comparisons cannot be because the grammar structure and syntax differ to the point the meaning is noticeably altered.
Show it's learned from those sources, then, instead of the "these words appeared somewhere else" standard. Show me that the scraper that gathered the training dataset plausibly could include those websites. Do some work to prove it, instead of continuing to make blind assertions followed by insulting the opposition.
Technically illiterate lusers, the both of you. Willfully ignorant of the very important details of the violations you're levying accusations of and the ways the system fails to meet them because you've never had to deal with the back-end. You feel that the people the data was scraped from have been wronged, loosely extrapolate comparisons to plagiarism and theft, then stubbornly refuse to accept any counter-argument that those are inaccurate terms to apply.
Then you openly admit you are just insulting the opposition for "not getting it" when they point out your colloquial use has nothing to do with that back-end, whether legal or technical.
11895592
Oh, how incredibly convenient that your demand just happens to be totally indeteminable due to AI companies not releasing or making any of this available. Well, gee, it's almost like people using copyrighted works in, at the very least, a legally questionable way wouldn't want to reveal that sort of thing! Of course, if they only used non-copyrighted works, there'd be no reason to conceal that, as is the case with Common Corpus, but oh well, sure there's nothing suspicious about that.
As such, let's indeed say that if we can't see the source list, it must be totally fine, and the many, many times LLMs have reproduced copyrighted texts and images pretty much verbatim, not to mention that fact that no one involved with these companies have actually denied the use of copyrighted works (almost like doing so might come back to bite them if they are forced to reveal their datasets), all mean absolutely nothing. Nope, these corporations aren't letting us see how their stuff works, but it must be totally fine!
And the word of the day today, children, is 'projection'. Can you say 'projection'? Very good!
Ah yes, just repeating your unoriginal, unsubstantiated, childish name-calling totally vindicates your use of it and shows your intelligence and argumental superiority. Not that I'm claiming to be some master debater either, of course, it's just you're making this really easy. Although I have to say, calling back to that before immediately chiding someone else for being insulting even after it was explicitly pointed out how hypocritical that would be? That is a level of self-unawareness that is genuinely quite staggering.
All that being said, as fun as it is laughing at your little temper tantrum, I realize the author might not want this taking over his comment section. So,
11893875
I'll just ask: do you want me to stop? If so, I apologize for taking up so much space.