On: ChatGPT Outage

[MuffinFlavored]

How do you know ChatGPT wasn't trained on the same answer from the web that Google is serving you from their crawler index?

[cactusplant7374 → MuffinFlavored]

1) I don't. However, I know that if I can see the answer in a Google result I can interpret the veracity in a way that ChatGPT can't.

2) Training doesn't necessarily mean the correct output that matches the Google result.

[Terretta → cactusplant7374]

as long as you know you need to interpret veracity, and have a basis for judging, you'll be fine

and, LLM probabilities-based constructs are unreasonably effective when you suppress the creativity that you don't need in a Q&A for yourself