Are those humans that nudged it representative of the population?
Or were they mostly "woke" Silicon Valley employees? (not to dismiss woke Silicon Valley employees, I'm just saying their opinions are not representative of the entire population).
Or were they mostly "woke" Silicon Valley employees? (not to dismiss woke Silicon Valley employees, I'm just saying their opinions are not representative of the entire population).
When you say "the entire population", you mean the entire population of the country "USA", right? Because as someone from another continent, it seems like there is a very specific set of opinions that you want included.
You use terms like "the other side" of the political discourse, which to me, reduces the set of opinion to two specific sets of opinions, namely the two sets represented by the two major parties in the american two party system.
As someone from "the outside", this seems like a very narrow view of reality, even if you managed to get your "unbiased AI", that represents both major american political parties, it will still seem like a very narrow and biased AI to someone from the outside of that.
Also, what exactly is the goal of a conversational AI? is it just to make a conversation with it seem like a conversation with an average american? If so, why would anyone want that? Wouldn't it be of more value to have an AI that could tell me what people with knowledge of a subject thinks of it, rather than what random people think?
That's just an idea that occurred to me (in 30 seconds of thought) which could probably make the training data significantly more unbiased.
But I'm sure there are research scientists who can come up with better methods for sampling data in a more unbiased fashion.
Note that this is not an all or nothing approach. Your training data could presumably be 100% biased or 0% biased, but also any value in-between.
The goal is to try to make it as close to 0% biased as feasible, given whatever effort you're comfortable expending.