Wait you mean if I train a guessing machine on how to guess based on answers anglos give it ends up having equally bad takes I don't understand how could this be
The datasets are curated specifically to give these kinds of responses. You can't have an AI that is not curated or else it will just turn into a 4channer, which was a big problem with early days of AI, but the moment you decide to curate, you are also intentionally injecting bias into how it is trained, which of course that bias will reflect whatever institution produced the AI.
They didn't. They were describing to those who read that it isn't just an unlimited guessing machine that is taking answers from Anglos (not wrong either); but more so that it's been deliberately biased.
Wait you mean if I train a guessing machine on how to guess based on answers anglos give it ends up having equally bad takes I don't understand how could this be
The datasets are curated specifically to give these kinds of responses. You can't have an AI that is not curated or else it will just turn into a 4channer, which was a big problem with early days of AI, but the moment you decide to curate, you are also intentionally injecting bias into how it is trained, which of course that bias will reflect whatever institution produced the AI.
I feel like you said what I said with more words
garbage in, garbage out
They didn't. They were describing to those who read that it isn't just an unlimited guessing machine that is taking answers from Anglos (not wrong either); but more so that it's been deliberately biased.