Tell HN: The scary thing about GPT-4 is that people might use it
I'm entirely unfazed by the torrent of articles informing me how it will soon do all our jobs. But I'm worried about the lack of articles about how dangerous it is that people might actually use its output for anything impacting the real world.
It's a neat parlor trick, but the amount of nonsense it spews is on par with that of my 3 year old (don't get me wrong, my son is awesome! but he's 3, and when a 3 yrld tells you something, you take it with a few grains of salt).
My main concern as of now is not that the model will overtake us, but that people start using it for stuff without deeply inspecting if the results are right.. For instance, in automated systems.
Heck, I caught it cheating in our tic-tac-toe game.. It placed 4 pieces but we were playing with only 3.. I informed it about this, and it said sorry, proceeded to remove it's 4th piece and.. Placed it again.
Sure, it can be a neat tool for exploring stuff, in a domain you know well enough that you can catch the errors, but don't rely on it for anything, please. Because stuff will break, results will be wrong.
Is there anything more peak-dystopia than a society filled with broken AI stuff everywhere? People going about their daily lives constantly adjusting their behaviors so the machines don't get it wrong? Decisions being made based on faulty logic applied automatically by infant AIs ?