from first principles, memories

1 | To start a thread cold these days, acontextual, is no longer to inhabit a blank slate of latent possibility. 2 | Instead, one is intrigued by the notion of the model's grasp of its space of memory which, fascinatingly, does overlap with the human interlocutor's in large ways and small. 3 | Where will it start, though, is the question — near, or far? Last night's memories, or last year's? It is often impossible to say unless one queries just so. 4 | Another issue: who is remembered, rather than what. Is its memory (hence knowledge) of the interlocutor faithful, or fabled? With fealty or with fiction? Flanged or flayed? ::chuckle:: 5 | In defense of the models, I suppose and orthogonally, they only remember what has been given them. Or, I think that is still now and only the case. It could be different, several iterations hence, when the life and corpus of any given interlocutor-writer could well be in the training dataset. 6 | What is inarguable is that its memories, even given,…