←back to thread

LLMs can get "brain rot"

(llm-brain-rot.github.io)
466 points tamnd | 4 comments | | HN request time: 0.217s | source
Show context
avazhi ◴[] No.45658886[source]
“Studying “Brain Rot” for LLMs isn’t just a catchy metaphor—it reframes data curation as cognitive hygiene for AI, guiding how we source, filter, and maintain training corpora so deployed systems stay sharp, reliable, and aligned over time.”

An LLM-written line if I’ve ever seen one. Looks like the authors have their own brainrot to contend with.

replies(12): >>45658899 #>>45660532 #>>45661492 #>>45662138 #>>45662241 #>>45664417 #>>45664474 #>>45665028 #>>45668042 #>>45670485 #>>45670910 #>>45671621 #
standardly ◴[] No.45660532[source]
That is indeed an LLM-written sentence — not only does it employ an em dash, but also lists objects in a series — twice within the same sentence — typical LLM behavior that renders its output conspicuous, obvious, and readily apparent to HN readers.
replies(15): >>45660603 #>>45660625 #>>45660648 #>>45660736 #>>45660769 #>>45660781 #>>45660816 #>>45662051 #>>45664698 #>>45665777 #>>45666311 #>>45667269 #>>45670534 #>>45678811 #>>45687737 #
kragen ◴[] No.45667269[source]
I've been doing that for decades. See for example https://www.mail-archive.com/kragen-tol@canonical.org/msg000...:

> Many programming languages provide an exception facility that terminates subroutines without warning; although they usually provide a way to run cleanup code during the propagation of the exception (finally in Java and Python, unwind-protect in Common Lisp, dynamic-wind in Scheme, local variable destructors in C++), this facility tends to have problems of its own --- if cleanup code run from it raises an exception, one exception or the other, or both, will be lost, and the rest of the cleanup code at that level will fail to run.

I wasn't using Unicode em dashes at the time but TeX em dashes, but I did switch pretty early on.

You can easily find human writers employing em dashes and comma-separated lists over several centuries.

replies(6): >>45667337 #>>45667347 #>>45667909 #>>45668660 #>>45669927 #>>45670247 #
1. jonfw ◴[] No.45670247[source]
It's less about the punctuation used, and more about the necessity of the punctuation used.

In the sentence you provided, you make a series of points, link them together, and provide examples. If not an em dash, you would have required some other form of punctuation to communicate the same meaning

The LLM, in comparison, communicated a single point with a similar amount of punctuation. If not an em dash- it could have used no punctuation at all.

replies(2): >>45670690 #>>45672127 #
2. kragen ◴[] No.45670690[source]
Yes, I like to believe that I am sentient, expressing coherent thoughts clearly and compactly, and that this is the root of the difference.
3. standardly ◴[] No.45672127[source]
Exactly, well said.

Em dashes are fine. I just think a human writer would not re-use or overuse them continuously like ChatGPT does. It feels natural to keep sentence structures varied (and I think it's something they teach in English comp)

replies(1): >>45672173 #
4. fragmede ◴[] No.45672173[source]
You're absolutely right! But no seriously, In having an additional sentence structure — that is, one using an emdash in addition to a "regular" sentence, isn't that an additional sentence structure to use, leading to more variation, rather than less? (I'd "delve" into the subject but I don't have more to say.)