themeatbridge

Symbols are easier than actual people. Nobody is just one thing, and complexity doesn’t travel well. Revolutionaries are either lionized or demonized, and so it becomes easier for the critics to ignore valid points while simultaneously being easier for the supporters to ignore atrocities. Rather than an objective, complete understanding of the man, you get polarized factions that don’t see each other as people.

Sure, of course it’s all fiction. LLMs didn’t exist as a concept when TNG was being written, so of course they didn’t write them to do that. Data was also an AI, and the holodeck was able to create a Moriarty so intelligent that he became self-aware and avoided deletion.

My point is that modern LLMs are an approximation of intelligence that permit contemporary computers to respond to novel stimuli in a way that seems like intelligence. The computer on the show was far more advanced because it was fictional and set in the distant future. They never bothered to explain how it works, but there was nothing approximate about the intelligence it created in the holodeck.

It doesn’t. An LLM uses modeling to approximate the shape of a question and generate a novel answer within the parameters. It’s like a shortcut to actually knowing everything and providing authoritative answers.

Sort of like the Scrabble champion who won in Spanish and French despite not knowing those languages. He simply learned patterns of words he could build with letters, but he didn’t know the meaning of any of the words.

The holodeck ostensibly has access to all of federation knowledge and history, and when it has to generate novel information (like what would Einstein say to Newton while playing poker) then it extrapolates that informtation based on the sum total of existing information. The holodeck computer has a thorough understanding of the question, the parameters, and the answers (holograms) it creates. It isn’t abstracting the question or approximating the shape of the answer.