Smartest Machine on Earth

The video players only show one chapter of an episode at a time. For a better viewing experience use the hotlink beneath each player to open up the full episode on the PBS website. Otherwise you'll have to click through five or six times during the viewing to see the whole episode.

Watch Smartest Machine on Earth on PBS. See more from NOVA.

What is "machine learning"?

How a machine learns behaviors and information based on empirical data and examples. To teach what an "A" is, developers must give the machine millions of examples so the machine can find patterns and determine the letter "A", even if it has never been show that specific example.

How did the IBM team employ that concept in the development of Watson's AI? What advantage did that provide over previous attempts at "intelligence"?

The team developers showed Watson a bunch of old Jeopardy! questions with a cheat sheet of the answers. With the thousands of example questions and answers, Watson could look for similarities between previous questions and the current question being asked. The advantage to this attempt was the Watson could actually analyze examples for simililarities and decide, by itself, what the possible answer is/would be.

I've often mentioned the term "Empirical Scepticism". What does that mean? How does that relate to the concept of Machine Learning? How does this relate to your life?

Empirical Scepticism is the sceptisicism between different answers. For Watson, when he is given the word "shoot", he is sceptical to whether that word is the shooting of a gun or the shooting of a camera. This relates to our life because, unlike machines, humans can determine which definition of the word to use based on the wording of the statement or question.