The alphago program beat Lee Sedol, world champion at the board game 'go' for the second time in a row.
What's interesting is that go can't be computed by brute force; the number of board configurations is more than the number of atoms in the universe. Instead they (Deepmind, a firm owned by google dedicated to research and application of deep neural nets.) trained a machine learning algorithm to learn the intuition that really good go players exhibit. In fact, one of the top go players that was defeated a couple of months ago said he couldn't tell from the moves alone if he was playing a human or computer.
Perhaps the next frontier would be to beat humans at trading on sports exchanges. This would be even harder, as the game changes over time.