Subheading and dot point spam, low density writing (the opposite of standard technical english), including useless detail (like enumerating stats on Shopify's scale), using contrastive parallelism, and other llm-isms. Even if it's not AI it's bad writing done by someone who has picked up AI's worst ticks.
For example this subheading:
> "The real bottleneck: connections, not CPU"
That's two AI smells. AI likes to say vacuous punchy statements like "the final takeaway" or "here's the rub". Then contrastive parallelism "connections, not CPU". Contrastive parallelism should scarcely exist in technical writing, regardless of whether it's AI generated.
I had never heard the phrase "contrastive parallelism" before and I love it. I mean, I can't stand how often LLMs do it, but having a word for my annoyance is nice.
I really wish that slop writing was disincentivized in whatever RLHF they do. I don’t want to read the weird LinkedIn pop-sci tone for the rest of my life in such amounts.
I wonder if the average reader is also getting annoyed like this or whether they just don’t care - especially seeing what seems to get upvoted on your run of the mill social media sites. They probably collectively shape things more than I do.
I don’t know how you stop the slop writing. What the LLM writes has the voice of whatever has been RLHF’d into the weights. There will be a voice. It will have ticks.
They can change it but it will be there.
I do like that I have a term for that “this, not that” phrasing that’s nails on the chalkboard for me after several years of reading slop.
The LinkedIn “techbro” writings are the worst. They talk like a cryptobro “APPLE JUST PUBLISHED A CODE REPOSITORY ON THEIR HAND CRAFTED LLM ANYONE CAN USE THIS TO COMPETE WITH OPENAI AND ANTHROPIC” and its some random simple LLM on github that handles very little compared to OpenAI and Anthropic.
That's two AI smells. AI likes to say vacuous punchy statements like "the final takeaway" or "here's the rub". Then contrastive parallelism "connections, not CPU". Contrastive parallelism should scarcely exist in technical writing, regardless of whether it's AI generated.
Aren't the LLMs trained on a massive corpus of human written texts? If that stands, then they are doing what they were asked, kind of? I am also not a fan of the fluff words that Claude pollutes the context with, it's a bit much, I could live 30% less but it doesn't want to read its instructions in the .md
If the author wanted, they could ask their "AI" to write the article as a caveman, no? I, simply for fun, slop coded up a skill so that I can ask Claude to reply as Samual L. Jackson, or Kramer, etc.
I also wonder if there is an "AI language barrier", likely not in this particular article, but let's say there was a published article by a researcher whose native language isn't English. What happens with the translation? I realize I am asking naive things :)
There are distinctive patterns of language use that can be powerful when used sparingly. LLMs trained on a massive copus of human communication, picked effective patterns, and overuses those patterns to the point that it feels both artificial and underwhelming.
I wonder if the human training process amplified this effect. I like to imagine there's an OpenAI RLHF trainer in Kenya whose name we'll never know who really likes this style of writing and is solely responsible for LLM voice. I know that's not how it works, but it would make for a fun story.
And you know what? That's not annoying--that's brave.
>Aren't the LLMs trained on a massive corpus of human written texts? If that stands, then they are doing what they were asked, kind of?
I think you might be interested in reading in training data generation, training, and post training papers/articles. I think you might be surprised at how much intervention there is on some of these levels.
Just? With all the Claudisms it is hard to distill what really happened. My guess:
- Our oversell protection was a gross hack that broke ACID.
- MySQL has a new feature that allows us to remove parts of the gross hack.
- Question: To what extent does the hack still exist?
Instead you get garbage like "The answer is often in the plumbing, not the engine." and "Crucially, this wasn't about making reservations fast. It was about making them safe neighbors."
But this crap is what Lütke wants, so they deliver.
For example this subheading:
> "The real bottleneck: connections, not CPU"
That's two AI smells. AI likes to say vacuous punchy statements like "the final takeaway" or "here's the rub". Then contrastive parallelism "connections, not CPU". Contrastive parallelism should scarcely exist in technical writing, regardless of whether it's AI generated.