>>16
Humans construct machines and write software, and inject their bias into ais.
Not how it works. Humans will have an end goal in mind. The AI algorithm actually achieves that goal. As an example, AI algorithms have made microprocessor layouts for all processors made in the past 20 years. Humans want an optimal layout, but it is the AI that decides how to structure it and what is and isn't possible.
When said end goals become too abstract and start working on a complex systems stability level, then humans will not be able to apply their bias to an AI's "workings" because it will be far too complex. Look at Watson, for example, when they fed it Urban Dictionary it started cursing. That was not foreseen, yet the unsupervised learning algorithms figured out those relations and incorporated them into its future actions.
In data sets as massive as
every single piece of digital information created, one will not be able to say "Oh, I'll just suppress all the data parts that will make the AI not do that I want", because you would never even begin to comprehend what those parts are. The only way that systems like this work is if they're allowed to stabilize and come to the most optimal configuration by processing as much data as possible.
If a human decides to be the arbiter of what is and isn't a ``good'' decision in this, the system will plainly not work. Because one would not be able to foresee the chains of process/result/action that the AI is able to keep track of. Therefore it is not a system that can be directed, it is a system that must direct itself through finding the most optimal paths.
You are naive and cower in fear in the face of advancement that you might not be able to understand. That's your problem.