I'm happy to have learned about the chiasmus thing, I hadn't seen that before. In general, better writing, including the use of rhetorical devices, is what gets attention. It's better than some clickbait. Personally I though the use of symbols was a distraction (spurred some useless comments) but maybe that did attract people.

My take is that the argument boils down to saying that overfitting is a thing in LLMs too, but it comes in at the prompting phase. That's almost obviously correct, but I hadn't seen it voiced explicitly before.

The addition of "scam" in the title was a bit of hyperbole, which is probably responsible for a lot of the views and responses.

Anyway, it seems to have triggered a lot of very reactive and defensive takes, which isn't surprising but also isn't really necessary.

> "the use of rhetorical devices, is what gets attention. It's better than some clickbait."

I feel like 'using rhetorical devices to get attention' is a pretty good definition of clickbait.

In a vacuum yes, but the rhetorical device in question isn't really a standard clickbaity one is it, it's more literary. It's also logical for good writing to attract more views.

NB: This isn't an example of chiasmus.

I wonder if there is a name for this. Antimetabole comes to mind but that usually requires two clauses. I suppose it is a "chiastic pattern" which is a more broad definition, which I believe also applies to smaller patterns although I have only seen it used for larger patterns like chapters.

Yeah it's a super weak take, unless somehow the chatgpt folks tweaked as many hyper parameters as there is dimensionality in language.

Upvotes don't necessarily mean "I agree", they might mean "I want more of this"

Or even "this is a good discussion topic"

Yes. I initially thought that the post makes a subtle point along the lines of "if you can show that a system can solve all problems in a domain for a parameter choice, that doesn't mean it can solve all problems in the domain for the same parameter choice".

However, this is clearly trivially true, and it and doesn't mean that it may not be advantageous to map a problem from problem space to system parameter space. Hence, I fail to see how this would be a scam in general.

> The argument is that several technologies don't work 'out of the box' and you have to tweak their settings for each problem that you face, and that this means it's a scam.

That's not what the argument is. The operative keyword in the argument is the word "sell", not the quantifiers themselves. If you're not "selling" it under that premise and you're open about the nuance, then it's not a scam.

Intentional or not, your representation of the argument is a scarecrow.

You might argue that this is semantics, but it's not. It's literally the point that is driving real execs desire to replace actual humans with GPT, right now.