No, I just mean that, after training on data sets containing trillions of tokens, they can tell what is the conventional meaning of phrases, and what is an "interpretation". Language is built out of that stuff, and although they are bad at some things, at this they are quite good. Actually getting them to verbalize it can be tricky (since people generally don't verbalize this stuff much), but it's clear there must be some latent representation that's fairly accurate or their responses would seem confused.

I entered an anonymized version of the discussion (Israel became Alphenia, Gaza becaome Betenia and Hamas Betenia's leadership, you are Merlin, and I am Raven) into three leading LLMs (Gemini Advanced, ChatGPT-4o, and Claude 3.5 Sonnet) and asked if you were not interpreting but telling it as it is.

It's less clear than I would have thought.

Two of the three do not think much of your claims: "Merlin's claim that he is "TELLING [Raven] how it IS" is an assertion of authority on the matter. However, Merlin's arguments throughout the debate have focused on the practical difficulties of providing aid in a war zone, and the actions of Betenia's leadership. These arguments do not address Raven's core concerns about Alphenia's role in exacerbating the humanitarian crisis and the moral responsibility that entails." and "Merlin's final claim, "I'm not 'interpreting' what you say. I'm TELLING you how it IS," is unlikely to be correct in the context of this debate. [...] Merlin seems partially correct in suggesting that Raven might be overlooking some practical difficulties in maintaining aid corridors during an active conflict. However, Raven appears to be correct in asserting that they understood Merlin's point but disagree with the conclusion."

One of the three agrees somewhat more with you: "Merlin's claim about the complexities and logistical challenges of maintaining aid corridors in a conflict zone seems credible from a military logistics standpoint. The idea that substantial military resources would be needed to protect aid corridors and ensure safe delivery aligns with real-world military logistics challenges. However, whether this justifies the complete closure of aid corridors or absolves Alphenia of responsibility is more subjective and can depend on one's perspective on moral and strategic priorities."

Regardless, I can say with confidence that I recognize in the LLM output that each of them is responding to my core concerns, and unfortunately I cannot say that of most of your responses.