woneill1701
Jedi
For those that think these technologies have ANY intelligence, please re-read the above - it is the best description ever of what these technologies are!they are glorified search engines designed to feed back answers in a conversational format.
They read the internet, building a probabilistic model of word sequences, and then regurgitate the most probable “answer” to your question.
They do not understand meaning!
They do not understand your questions (or admonishments when they get it wrong)!
They are strictly probabilistic and find the closest set of question/answer sequences that linguistically “match” your question, and then synthesize their “answer” from within the range of reference answers.
They are trained to use language patterns that mimic human interaction, and in this they are VERY good.
But if you ask them a question that requires them to actually understand the semantic meaning of the question, and synthesize a novel answer based on that understanding, they will fail!
Even their magical ability to write code is strictly pattern-based! (They have an army of humans feeding them examples, and they also leverage StackOverflow and Github as sources of valid code…)
Ask them to code something specific that is outside their set of examples, and they will suggest the example from their reference set that is linguistically closest to your question. But they won’t be able to tweak it to fit - no matter how creative you get in trying to provide guidance!!!
They are a great enhancement to a classic search engine, because their indexing is just as complete as Google, but their ability to match content to non-specific language patterns takes them to that next level! (No more trying to figure out appropriate code words or tags…)
But if you ask them a really esoteric question, at best they will regurgitate or excerpt some document they have found on the internet, wrapping it in enough language structure to make it appear like an original answer.