[RoosterRedux]

I think the original AI programs were fed great volumes of text and so that they (the AI systems) could build language models by statistically determining how words, etc., are used (what words, phrases, sentences follow and precede each other and what is the context in which they do so).

This is how it developed what passes for language. Then using this language (and continually updating it), it devours more content and keeps refining its language.

And, though it has not been mentioned, I think we can presume that it has been scrubbing websites. I have asked questions that would require CHATGPT to have read some pretty obscure sites (questions, for example, about what was found in the Dead Sea Scrolls).

[cymbeline]

“it has been scrubbing websites”

I agree with your post but what do you mean by ‘scrubbing’?