[kenty]

> it's common for people to be bad at a second language

Non-native speaker here: huh, is "you are absolutely right" wrong somehow? I.e., are you a bad english speaker for using it? Fully agree (I guess "fully agree" is the common one?) with this criticism of the article, to me that colloquialism does not sound fishy at all.

There might also be two effects at play:

  1. Speech "bubbles" where your preferred language is heavily influenced by where you grew up. What sounds common to you might sound uncommon in Canada.
  2. People have been using LLMs for years at this point so what is common for them might be influenced by what they read from LLM output. So while initially it was an LLM colloquialism it could have been popularized by LLM usage.

[popopo73]

>is "you are absolutely right" wrong somehow?

It makes sense in English, however:

a) "you are" vs "you're". "you are" sounds too formal/authoritative in informal speech, and depending on tone, patronising.

b) one could say "you're absolutely right", but the "absolutely" is too dramatic/stressed for simple corrections (an example of sycophancy in LLMs)

If the prompt was something like "You did not include $VAR in func()", then a response like "You're right! Let me fix that.." would be more natural.

[kenty]

Thanks for the thorough explanation, that, indeed, is a level of nuance that's hard for me to spot.

Interestingly, "absolutely right" is very common in German: "du hast natürlich absolut Recht" is something which I can easily imagine a friend's voice (or my voice) say at a dinner table. It's "du hast Recht" that sounds a little bit too formal and strong x[.

Agreed on the sycophancy point, in Gemini I even have a preamble that basically says "don't be a sycophant". It still doesn't always work.

[d-lisp]

> in Gemini I even have a preamble that basically says "don't be a sycophant". It still doesn't always work.

Using this kind of strategy eventually leads to the LLM recurrently advertising what it just produced as «straight to the point, no fluff, no bullshit». («Here is the blunt truth»).

Of course no matter how the LLM advertise its production, it is too often non devoid of sycophancy.

[popopo73]

There's likely a cultural element to it as well regarding how we admit mistakes and correct ourselves.

With how blame-avoidant western individualist culture can be, seeing something "admit" doing wrong so quickly, and so emphatically, could be uncanny valley-level jarring.

[account42]

It's a valid English phrase but it's also not unlikely that someone states something as a fact and then goes immediately to "you are absolutely right" when told it's wrong - but AI does that all the time.

[Ekaros]

It fails the basic human behaviour. In general humans are not ready to admit fault. At least when there is no social pressure. They might apologize and admit mistake. Or they might ask for clarification. But very rarely "You are absolute right" and go on entirely new tangent...

[fedeb95]

I'm also a non native speaker. The point is, you tend to make grammatically correct phrases (mostly), but in a way that's not very common between native speakers. You're right that there are many factors at play. I think dismissing something as AI generated just because it uses common AI-generated strings is not correct. I'm speaking in general and not about the specific case.

[cannonpalms]

Just don't open a comment with

> You're absolutely right!

And you're good