[no_multitudes]

I am not sure if you are serious? It really has nothing to do with being grammatically correct. The article is chock-full of all of the most common LLM tells. The aggressive use of negative parallelism, filler adjectives like "deeply", and unvaryingly breathless tone all feel like decisions an actual thinking person would not make.

[Sophira]

On the one hand, people say that AI models (LLMs, image generators, etc) are just stealing from people and that they cannot be original.

On the other hand, people say that AI models have tells that no actual person would do.

Which is it? You can't have it both ways.

[noitpmeder]

Imo those are not nearly as mutually exclusive as you're presenting, as would be evidenced by anyone seriously reviewing the output of these systems, to the point it feels like a bad faith argument or an uninformed one

[no_multitudes]

I said one of those things (sort of) but not the other. I am not sure why you are speaking as if "some people say x, and others say not x" is a gotcha. I am aware that many people say many different things.

[nottorp]

No actual person would do when writing.

It looks like it was the prevailing accepted style by the human team who was initially training the "AI"s. Who were not writing.