Further proof about how much of bullshit generators these things are.

I mean, it clearly is regurgitating something, but does not understand the semantic context at all

These LLMs are so trained to spit out a confident answer that logic is irrelevant, as long as something is said

It’s kind of like how politicians very rarely answer “I don’t know” to questions, because to not give an answer is a sign of weakness

@kaspernymand if everybody knows this, then why doesn’t copilot? If we’re supposed to rely on these things to provide appropriate answers, then most people would rather expect an answer rather than a joke, wouldn’t they? If I wanted to look for jokes, I would go to a comedy club, with real humans, instead of a computer.

Do you see the limitation here? I wanted an answer to a question, and the answer was already provided in the search results below this stupid chatbox. The Copilot would not have made any of the results any better. This has negative utility because it wasted my time.