@nntaleb nothing breeds confidence like ignorance.

@nntaleb I often want to use google or an AI assistant for things I already know but forget, e.g. "what's the song featuring artist X with lyrics like Y" or "remind me how to reset my Apple TV remote." It seems especially useful for that case, where I will know immediately if it helped me or not.

@nntaleb Bingo.

@nntaleb this gets to the heart of how these software systems will be misused: feed them with crap, then watch them spit it back out - confidently. Humans have been trained to trust the confidence of machines based on the assumption that there are well-intending humans behind this confidence; this will end badly if not checked.

@nntaleb https://m.youtube.com/watch?v=w65p_IIp6JY
Some interesting perspective on dishonest ML systems.

@nntaleb
ChatGPT is not AI, it's just a neutral network.
It has no context awareness nor capacity to reasoning about what it's been asked.
Relaying on those responses without verifying is the same as believing the first person you find on the street to give you advice on your life without knowing you.
It's useful for translation and to better understand a topic you already know (like programming functions) but people should stop using it like a source of truth.

@nntaleb Yes! I was confident it would produce a simple batch script right, and even though I can't code I could spot it forgot one of the two conditions I asked it to put in the script. When I asked it to explain how the code tested the two conditions, it said "sure" and rewrote the code to explain it. It only acknowledged the mistake when I confronted it about the rewrite.
Perhaps it could do well as an investment advisor. I mean, as well as the next guy.