All the AI we’re building right now have proven, over and over again, to either have our biases built into them OR to have the ability to learn them quickly.

Facial recognition, already treated as this perfect, flawless tool cannot tell dark skinned people apart. It’s success rate as advertised is based on how well it could tell the Post-Grads working on the thing apart. Our systematic issues of generational poverty feed into systematic issues of education gaps, feeding back into the poverty, and NOW we’re feeding them into law enforcement AI that supposedly takes the human biases out of the equation, until it turns out that it LITERALLY cannot tell black people apart. Yeah, it’s not like the AI chose to have that flaw, it’s not actively dismissing anyone darker than khaki as “eh, you all look alike” but the failure of the developers to even consider if it worked on non-white faces lead to this.

Long story short, we are nowhere close to being able to build the AI that will remove our flaws from the equation.