Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

LLM Writing Tropes.md

tropes.fyi

Post reply on HN
Print view View original

171–180 of 211 posts

Re: LLM Writing Tropes.md

#171

Post by kristianp »

capnrefsmmat wrote: post #15

I work on research studying LLM writing styles, so I am going to have to steal this. I've seen plenty of lists of LLM style features, but this is the first one I noticed that mentions "tapestry", which we found is GPT-4o's second-most-overused word (after "camaraderie", for some reason).[1] We used a set of grammatical features in our initial style comparisons (like present participles, which GPT-4o loved so much tha…

> It's something in instruction-tuning that's doing it.

Isn't the instruction tuning done with huge amounts of synthetic data? I wonder if the lack of diversity comes from llm generated data used for instruction tuning.

Re: LLM Writing Tropes.md

#172

Post by jari_mustonen »

Some of these are neccesary parts of LLM's. They use the content they create to direct what they are going to say. This applies to patterns like "In conclusion, ..." and what the author calls "Fractal summary". Turn them off, and the general quality of the AI thought gets lower.

Re: LLM Writing Tropes.md

#174

Post by BoredomIsFun »

FiniteIntegral wrote: post #79

A subtle tell for generated text is just how damn flat it is to read. Not that technical documentation require some form of grand prose, but how unspecific the text can truly get. Reading a high school persuasive essay can have more detail, and those are often just written for a grade. I can understand someone needing help with writing but getting an agent to do the job for you feels like a personal defeat.

[flagged]

Re: LLM Writing Tropes.md

#175

Post by BoredomIsFun »

devonkelley wrote: post #163

[dead]

Here what kimi thinks:

The post is moralizing theater masquerading as craft wisdom. “Just write it yourself” ignores the actual quality curve. Give a modern LLM a one-paragraph brief, ask for 600 words, then spend three minutes deleting the three most obvious adjectives and one “delve.” The result is already clearer, better structured, and more grammatically airtight than what 80 % of English-literate adults can produce in twenty distracted, coffee-spilled minutes. That isn’t speculation; it’s what every A/B test in every newsroom, ad agency, and SEO shop shows when copy is anonymized and editors pick winners. The average human twenty-minute draft loses—every single week.

Re: LLM Writing Tropes.md

#176

Post by andelink »

BoredomIsFun wrote: post #174
FiniteIntegral wrote: post #79

A subtle tell for generated text is just how damn flat it is to read. Not that technical documentation require some form of grand prose, but how unspecific the text can truly get. Reading a high school persuasive essay can have more detail, and those are often just written for a grade. I can understand someone needing help with writing but getting an agent to do the job for you feels like a personal defeat.

[flagged]

Great example of parent comments point.

Re: LLM Writing Tropes.md

#179

Post by Lerc »

jug wrote: post #170
Lerc wrote: post #149

Earlier quoted context omitted.

I have seen people suggest that the problem is that LLMs let you express any of your ideas, but the number of people with ideas worth expressing is limited. In a sense I think this is accurate, but not inevitable. I think there is a lack of creative thinking, but it has come from a world that doesn't value it and suppresses difference. There is a brilliant line in Treehouse of Horrors IV where Principle Skinner says…

I think it has more to do with LLM's being statistical models than human creativity lacking in the input. The creativity and millions of voices and tones may be there, but since these models tend to go for the most likely next words, polishing this away becomes a feature. A text by a human mind may be seen as a jagged crystal with rough edges and character. Maybe not perfectly written but it's special. An LLM takes a…

>I think it has more to do with LLM's being statistical models than human creativity lacking in the input. The creativity and millions of voices and tones may be there, but since these models tend to go for the most likely next words, polishing this away becomes a feature.

I have always thought this is a rather misguided view as to what LLMs do and indeed what statistical models are. When people describe something as 'just statistics' I feel like they have a rather high-school-ish view of what statistics represents and are transferring this simplistic view to what is going on inside a LLM. Notably they do not find the most probable next word. They find the probability of every word that could come next. That is a far richer signal than most imagine.

And ultimately it's like saying that human brains are just chemical bonds changing and sometimes triggering electrical pulses that causes some more chemicals to change. Complex arrangements of simple mechanisms can produce human thought. Pointing at any simple internal mechanism of an entity without taking into account the structural complexity would force you to assume that both AI and Humans are incapable of creativity.

Transformers are essentially multi-layer perceptron with a mechanism attached to transfer information to where it is needed.

Re: LLM Writing Tropes.md

#180

Post by chmod775 »

Lerc wrote: post #179
jug wrote: post #170

Earlier quoted context omitted.

I think it has more to do with LLM's being statistical models than human creativity lacking in the input. The creativity and millions of voices and tones may be there, but since these models tend to go for the most likely next words, polishing this away becomes a feature. A text by a human mind may be seen as a jagged crystal with rough edges and character. Maybe not perfectly written but it's special. An LLM takes a…

> I think it has more to do with LLM's being statistical models than human creativity lacking in the input. The creativity and millions of voices and tones may be there, but since these models tend to go for the most likely next words, polishing this away becomes a feature. I have always thought this is a rather misguided view as to what LLMs do and indeed what statistical models are. When people describe something a…

> They find the probability of every word that could come next.

If we're being pedantic, they find a* probability for every token (which are sometimes words) that could come next.

What actually ends up being chosen depends on what the rest of the system does, but generally it will just choose the most probable token before continuing.

* Saying the probability would be giving a bit too much credit. And really calling it a probability at all when most systems would be choosing the same word every time is a bit of a misnomer as well. During inference the number generally is priority, not probability.

Post reply on HN