Odd mistakes, like answering "What is Toronto?" for a category "U.S. Cities", but Watson kicked butt overall.
I wonder how much of that is due to how quickly Watson could trigger the buzzer. There was an overlay showing Watson's top 3 answers with a confidence percentage and a cut-off line, if Watson's confidence was below the cut-off it wouldn't attempt to answer and those were the questions the humans were able to buzz in on.
http://www.geekcultu...rchives/1505.html