The OP applies mainly to posts consisting of text presented as written by the user. What it doesn’t explicitly address is the use of LLM-generated text in quotation blocks and attributed to the LLM.
It came up earlier, and I clarified the policy:
Since then, two people have expressed the view that we ought to make an exception for such so-called quotations. I’ll address @WeSee’s objection first:
The claim is that quoting a human author and copying LLM output are equivalent, on the grounds that neither represents the poster’s own voice. There are a few problems with this view.
First, the two cases are very different. When you quote Kant, you are referring others to the thoughts of a specific mind that produced those thoughts at a specific moment in history, for specific reasons, within a context of intellectual history that can be traced, examined, and engaged with. None of that is true of quoted LLM output, because it was generated on demand by the user—there is no authorial position behind it.
Second, the claim that neither represents the user’s voice is a misleading way of describing the situation. Quoting an author, in the context of philosophy, is part of the act of expression. It involves selection and contextualization, with a view to integrating the quotation into what the user is saying. The quotation is independent and subordinate—the user brings it in from outside to serve a point they’re making independently of the author quoted.
Using LLM output in the guise of a quotation does not work that way. The text does not exist independently. Rather, it is generated in response to a prompt that has what the user wants to say already baked into it. The user doesn’t do anything comparable to the selection that goes on when quoting an author, so what appears as an act of quotation is not one at all.
Third, quoting an author “as a source of additional information,” though it might sometimes happen, is hardly the most common use of quotations in philosophical discussion. They are more often used to present and examine ideas, insights, and arguments.
Now, I’ll address the general objection, by way of @Banno’s comment:
We are not allowing quoted AI output, even when explicitly identified as such, because it would defeat the purpose of the policy: people would use quotation blocks to circumvent the rule against posting AI-generated text. And it is not equivalent to quoting an author, as I explained above.
More generally, the forum exists for human discussion. The aim here is not to build an archive of plausible, smooth, inoffensive prose, but to facilitate discussion between people attempting to think through problems, formulate positions, and respond to one another directly. Allowing AI-generated text in any form will detract from this aim.
NOTE 1: This applies equally to cases where AI output is used to present information or evidence. Scientific data, factual claims, and references can be included by citing the original sources directly.
NOTE 2: There might be limited exceptions. For example, using AI to help create the markup for a table and then using that in a post (e.g., “Claude, kindly make me a table of synthetic/analytic vs a priori/a posteriori for posting on TPF”). This kind of labour-saving AI work is fine, I think.
I will probably post another clarifying announcement on this specific issue, once the discussion has run its course.
EDIT: Actually, I’ll probably just update the OP.