Yet, as humans, we assume that when facial recognition fails, the whole process is inherently flawed. But was it really?

According to Josh, that is the most fundamental thing to understand when it comes to machines. Not meeting our human expectations, doesn’t automatically make the technology itself a failure. These things were, by definition, built on logic, which begs the question: Can a robot's solution actually be wrong?

The point of introducing machine learning into our products was never to have them do all the work. Instead, algorithms and logic-based solutions ought only provide humans with better insight so as to empower us to arrive at better solutions, faster.

This fundamental understanding our users that really helps us make better products. This might be a simple example, but if a computer can figure out how to walk on it's own, maybe it's time to start investigating why and how these solutions were formed.