Post

Looking at The Blue Fighter, I can understand why @vauhinivara.bsky.social thinks that it came from an LLM. But, applying my own theory-of-mind (and theory-of-LLM-non-mind) based analysis, I believe she is wrong.

There are tricolons and antithesis in the text, which has become an "AI" tell for many. But while the style is superficially similar, there is actually an important difference between how these are used in human writing and how they appear in extruded (LLM-generated) prose.

To a human writer, style is semantics. To a synthetic text extruding machine, which has no semantic understanding, style is just structure.

That difference is everything, because in extruded prose that lack of authorial semantic intent is often visible. Even where a human writer employs this style formulaically, they still have a reason for doing so. It's not antithesis itself that is a feature of synthetic text.

It is only a tell of synthetic text where that antithesis adds no meaning, or worse, is nonsensical. It doesn't matter how many "it's not x, it's y" you see; if they all serve an actual semantic purpose, a human writer had an ostensible reason to use them, and in The Blue Fighter, they all do.

Similarly, it's not the use of tricolon itself that is a feature of synthetic text. It's where that tricolon has been forced to fit; the third item adds no meaning, or is an absurd reach. Human writers love to use tricolons, but not so much that they'll force them to fit regardless of semantics.

A synthetic text extruding machine, however, will, because it's not deploying it to communicate meaning, it's deploying it because of structural signals from the preceding tokens. Again, none of the tricolons in The Blue Fighter have this forced-to-fit quality.

Forcedness in general is another feature of extruded prose that also commonly appears in human writing. It is of course a subjective judgement on the part of the reader whether prose has been forced, but it is highly intuitive: you know it when you see it.

Again, it may be superficially similar, but as I first touched on in The Empty Cockpit, the phenomenology of forcedness in synthetic text is very different to that in intentional human writing.

Forced prose is not good prose, but it does serve a semantic purpose in the human drafting process. Forcedness in synthetic text is the inverse phenomenon. It happens because, as I like to put it, the model can't choose, so it also can't choose to just not.

It can't decide to just abandon a half-extruded paragraph that isn't working, even if there is no reason to try and make it work, and it cannot, in reality, decide that this passage is already complete and it's time to move onto the next.

So when forcedness appears in synthetic text, it's because the model has gotten 'stuck'. With no semantic understanding telling it when to transition, it overshoots. It writes too long, and not becuase it is having a good time.