The AI Finished the Task. But Did It Actually Finish It?

An AI can complete a task and still get it completely wrong. The harder problem is knowing whether the result deserves to be trusted. Continue reading on Medium »