Try feeding these prompts into your AI:

I'm not going to do that. Maybe "AI" is amazing, but it's not actually THINKING.

What does it matter whether its thinking?

When you are searching a library (or the web for references) you feed prompts into your own search of a card catalog or tables of contents of journals or whatever. No one worries whether the search engines are thinking or not. It is sufficient that the people who wrote the stuff the search returns were thinking.

Well guess who has written 99.9% of the stuff that trains an AI? People who were thinking.

An LLM has been compared to a "data compression" of all the textual materials that were used to train the AI.

Have you heard a GPT referred to as a "predictive" text engine? That you put the prompt in the GPT and out comes text that the GPT "predicts" should come next? If you prompt a GPT with a text sequence from a book or journal article it was trained on, it will likely include snippets from those sources in its answers, because "thats what came next." But when you ask it a question it was not trained out, what does it even mean to say its answer is a "prediction" of what comes next?

I think it is predicting what would have come next if you had asked the collection of people who created collectively all the stuff it was trained on. It is predicting what people who were trained and were expert and were good at explaining things would have written if THEY had been asked the question.

Its a wierd subtlety, but we don't worry that a book isn't thinking, we still read them.

R: