Self-driving cars are a big no.

“If you assume out of the gate that AI is going to be wrong a significant fraction of the time, whether that's 10 or 20 or even 30%, you're already in the ballpark of a major use case,” Jonathan Rosenberg, CTO and Head of AI at Five9 said.

Why would you put a system that would mess up in charge of driving?

How about putting it in charge of something with a low inaccuracy cost instead?

We need to seek out zero cost of inaccuracy, explained Rowan Trollope, Five9 CEO. That means applications of AI where you can tolerate mistakes and still get benefit when it's not.

Not self-driving cars. Think product recommendations, call center assistants, and diagnostic aids.