I've read several dozen articles about the current spate of generative AI,
in a context for a client.

My thought it, as previous posters have suggested: If asked a LMM (like
ChatGPT) to ingest the complete works of SLC, including the Speeches,
Letters and Autobiography, in theory you could have something that would
respond to queries in a simulacrum of SLC's public voice. ,

I would definitely add dates to the algorithm in there. 1865 to 1909 is a
long span.

Assuming you had done this, what would you have? A high-tech Twaint.

On the other hand-- suppose you applied your new Twainbot to filling in the
blanks on an unfinished MS like Huck & Tom Among the Indians, or #44?

You wouldn't have Twain, but you would have something derived from Twain.

DDD