Quote:
Originally Posted by cRiPpLe_rOoStEr
How could we be sure a so-called artificial "intelligence" would never reproduce the biases of its creators/programmers?
|
It would, by definition.
The AI model is defining what is "good" and what is "bad". It randomly makes many decisions and evaluates if the outcome achieved "good" or "bad". After many iterations, it develops a framework for what is "good" and what is "bad".
The programmed good and bad is all that AI is.
This is the basis for all of the dystopian fiction that imagines AI taking "good" to an extreme, to the exclusion of everything else that we failed to incorporate as good.