Top
Best
New

Posted by Eloissssss 1 day ago

ChatGPT starts blocking direct requests to copy an author's style(arstechnica.com)
90 points | 65 comments
esjeon 2 hours ago|
LLMs are only good at copying the surface, like frequently used words, sentence structures, use of metaphors, etc. They fail at copying the underlying thinking framework. LLMs cannot reproduce this, and they do notice the lack of framework in their writings (that’s how the know the writings are fake), but offer those as successful clones anyways.
malshe 1 day ago||
This is rich considering LLMs can't even write like an average person let alone famous authors
rymiel 3 hours ago||
LLMs cannot write dialogue between two characters where one person knows a fact and the other person doesn't. It will consistently make the character who shouldn't know something bring up said fact and I have not found a model which will not do this. All these years of LLM research and it fails grade schooler level logic, if it is even slightly abstracted behind a story.
internetguy 1 day ago|||
this is rich coming from the company that shreds books for training data
Lerc 9 hours ago|||
There is a really weird perception of the importance of books as if there is a important property to be preserved in every book.

This was hilighted to me by a librarian friend if mine, She said they destroy books all the time in her job, and got rather angry when people suggested that it was intrinsicly bad, because there is nothing sacred about simply being a book. It is about the replacibility. They had a program where, if they destroyed a book they would receive a replacement from the publisher for a tiny cost. If a book got damaged it would be much cheaper to get a freshly minted replacement than it would be to spend time repairing.

If there was a glitch in the matrix and 30 billion Gutenberg bibles suddenly fell in the middle of the Amazon(the wet one), most would be destroyed as an environmental hazard.

I see a lot of comparisons to book burning when the AI scanning is mentioned, but the point of book burning is a symbolic act to indicate that the information within the book should not be shared.

Destroying the book to capture the information within sits at the polar opposite reason for destroying a book.

chrisldgk 6 hours ago||
The difference here I think is that the company (Anthropic, not OpenAI) was specifically selecting rare books for training (and subsequently destroying). I think the value proposition is a much different one in this case. In your example the books are still available in high enough amounts that replacing it is cheaper. If the book is rare enough to not have been digitized for training data yet (which is why Anthropic is interested in it), keeping that book around is much more important.

If Anthropic subsequently released the digitized version of it we’d be having a different discussion, but for obvious reasons they’re not.

giancarlostoro 4 hours ago||
They did it due to some weird copyright argument about copying and destroying the original seeming like a transfer and not copyright violation, which is insane and apparently is mot always interpreted in said way. So they are literally taking a copyright gamble they might lose.
inigyou 18 hours ago|||
[flagged]
classified 19 hours ago|||
Just another propaganda lie, implying that they could imitate any style.
qw2187 18 hours ago|||
The early models could to an extent. They removed it, probably with RL, because the feature reveals the inherent plagiarism.

That is also why ChatGPT blocks it now. The plagiarism is still there of course, just hidden.

kevin_thibedeau 16 hours ago|||
They can imitate distinctive authors.
Eloissssss 1 day ago|||
[dead]
simianwords 23 hours ago|||
[flagged]
Planktonne 21 hours ago|||
> LLMs can obviously write like anyone

This simply isn't true. An LLM can do a weak parody of a sufficiently-famous author, but they're extremely poor at sustained fiction writing even without trying to emulate a specific style.

Perfect grammar is only a small part of writing well.

mort96 21 hours ago|||
If they can "obviously write like anyone", why do they universally write like crap? Wouldn't the AI companies want them to write like someone who can write
serf 16 hours ago||
Because llms can write like anyone in the same way that a type writer can; with a skilled human typist behind it.

You see a lot of crap for the same reason most code is crap - the human who did the thing sucked. Language models aren't magic, they're just a compiler that guarantees an output regardless of input quality and structure.

That doesn't mean much wrt to the capability of the model.

The world is filed with crap literature, movies, stories, and advertisements written by honest to God human professionals in their respective field, and we trained these things in that environment. It's not surprising to me that they need a nudge to get to Hemingway when the corpus average is presumably so low.

akoboldfrying 1 day ago||
Even LLMs from years ago could accurately mimic the style of any sufficiently famous author.
est 1 day ago|||
LLMs from years ago weren't aligned to death like these days. It's hard to get rid of "AI smell" than years ago.
throwawaydfasjf 1 hour ago||
[dead]
Bolwin 1 day ago||||
Yeah and it's degraded significantly since then. Older llms were still mostly language focused and had a lot of latent knowledge about things like writing styles. Now it's crowded out in favor of agenetic work, programming etc.
gwern 1 day ago|||
No, the latent knowledge is larger than ever, as verified by many benchmarks (and instances like people being shocked by truesight of obscure forum posters). This is 100% a chatbot personality/alignment/post-training thing.
HWR_14 1 day ago||
> truesight of obscure forum posters

What does this mean?

gwern 9 hours ago|||
Stylometry on steroids: https://gwern.net/doc/statistics/stylometry/truesight/index
mathgeek 19 hours ago||||
Truesight is a spell in D&D that allows you to see the true being behind an illusion or other deception.
bloqs 22 hours ago|||
It's almost like a claudeism
conception 1 day ago|||
I’m surprised no one has distilled a 4o model yet that’s really good at prose. No money from enterprise users I suppose.
R_D_Olivaw 15 hours ago||||
Can anyone point me in the right direction on how to get these models back? I really like the older ones for NPC dialogue for TTRPG campaigns and the like.

I just looked into it and I think I can still find GPT 3.5, but wonder if it too has been trained out of usefulness.

The characters a styles were quirky and not always perfect, but added a lot more flavor and character nuance that could be corrected with editing.

giancarlostoro 4 hours ago||||
You can pay Microsoft they were exclusive host of GPT models.
fwn 13 hours ago|||
I have no experience with style imitation, so perhaps I'm missing something here. I asked Kimi K3 to rewrite your comment in the style of Ernest Hemingway:

> I would like to get the old models back. If anyone can show me the way, I would be grateful. I liked them for NPC dialogue, for TTRPG campaigns and the like.

> I looked into it. GPT 3.5 is still around, I think. But perhaps they have trained it out of usefulness. It is difficult to say.

> They were quirky, the old ones, and not always perfect. But they had flavor, and their characters had nuance, and what they got wrong you could fix in the editing.

Full conversation with thinking: https://paste.ononoki.org/?2fc049fe1ae7302b#GjzuEn8VmaYJ565M... Password: kimitest

onemoresoop 5 hours ago||
How token efficient is kimi? How much do you average each month for example?
andai 21 hours ago||||
The base models were really good at this. These days even the "base" models are full of awful synthetic data.
casey2 1 day ago|||
At best it can use some of the same vocabulary. Often not even then.
at1as 1 day ago||
I've asked it before to write in a style similar to Orwell (either directly, or by following his published rules for writing). I do hope that continues to be supported.

I just can't stand to see any more "Why It Matters.", and this trick seems to strip out the worst offenders

lethologica 1 day ago||
After it stole literally every authors style in existence…
pseingatl 1 day ago||
Wuddabout:

Works unfinished at the time of the author's death?

Abandoned works?

Lost works? Prompt: Aristotle's work on comedy has been lost. Write a short treatise on comedy, using Aristotle's methods as shown in his Corpus and especially his treatises on Rhetoric and Poetics.

tgv 1 day ago||
Who cares? There probably have been hundreds of such texts by later philosophers and authors. None of these has value as a completion of Aristotle's work. The original's importance is historical, and you can't retrofit history.
pseingatl 19 hours ago||
Some people read for pleasure.
tgv 17 hours ago||
I doubt those people have finished all the other books in existence.
Planktonne 20 hours ago|||
Something that looks a bit like another thing isn't the same thing.

No amount of prompting would get you Aristotle's actual lost work.

watwut 23 hours ago|||
Useless. The work is still lost or abandoned.
pseingatl 19 hours ago||
I would wager that many people are eager to read George R.R. Martin's latest were he unable to finish it. Stieg Larsson's Lisbeth Salander has featured in works written by others after his death. Why not AI?
Planktonne 5 hours ago||
1. AI is incapable of doing this well, and that seems unlikely to change.

2. No one who appreciates those books wants this.

QuadmasterXLII 18 hours ago||
Use a fucking typewriter?
mandeepj 3 hours ago||
How many books were they (LLMs) trained on illegally?
ticulatedspline 3 hours ago|
At this point possibly technically none. Anthropic recently made a deal to basically pay out for their piracy, and all the big groups have purchased shit tons of used books.

wouldn't be surprised if pretty much all the training data is legal now, at least retroactively from having obtained copies later.

raincole 20 hours ago||
Doesn't matter; have DeepSeek. US AI companies are too busy shaving their heads into their own rear sides.

I don't know what would happen when DeepSeek inevitably catch up though. Perhaps that's going to be how this wave of AI hype ends?

obscurette 20 hours ago|
Do you think people behind DeepSeek don't have their own interests? And people/parties controlling people behind DeepSeek?
raincole 18 hours ago|||
When did I say that...? Is this an AI hallucinated comment?
thesmtsolver2 5 hours ago||
Your comment implied that. Don’t have to say everything directly. You don’t need AI hallucination to connect things logically.
colingauvin 3 hours ago|||
...are they going to come to my house and delete my local copy?

Open weights are a one-way ratchet.

mud_dauber 18 hours ago||
I previously asked ChatGPT (albeit several months ago), as an experiment, to answer some general questions in the voice of Don Rickles. I’m hoping this capability somehow survives.
Grimblewald 6 hours ago||
i tend to give ai a "voice" file, which it must use to write. I hope this doesnt spread to providers i do use, since if it does i mugbt just forgoe ai, i cannot stand ai prose.
super256 1 day ago|
Haven't they been doing this for a long time already? I remember trying to copy Hunter S. Thompson's style many moons ago and getting a refusal.
More comments...