[pavish]

It is true that AI can carry bias and often produce results that are unexpected and offensive. It was clearly put into simpler terms by a ted talk I got to watch recently:

The danger of AI is weirder than you think (https://www.youtube.com/watch?v=OhCzX0iLnOc)

In terms of maturity, the AI we have now is much closer to a statistical analytics engine than to the all knowing AI governments shown in sci-fi, which is to say that it is in its very early stages.

I can't wait for the good stuff, but I'm also concerned that there's going to be multiple unexpected ripple effects in the path towards that goal.

[johnnycab]

The linked TED talk is by Janelle Shane, an optics research scientist and an AI researcher. She maintains a blog, which is quite funny and has also written a book drawn from her experiences with NN's, particularly GPT-2.

https://aiweirdness.com/