Just to punctuate my love for all things Google (or Alphabet, it will take awhile before I get used to that), Nature magazine just published a paper describing Alphabet-owned artificial intelligence company Deep Mind’s* system which used deep learning to beat the top player in the game of Go. Beating a human in the 2,500 year-old game of Go has long been considered an important milestone in artificial intelligence.
Unlike other strategy games such as chess and checkers, Go has never been beaten by a computer – until now. We all know the famous match between Deep Blue and Kasparov in 1996. Less well known (at least to me anyway) is Chinook, which mapped every possible move in Checkers – 500 billion moves in all – making sure that it can never be beaten by a human.
The game of Go however cannot be beaten by brute force alone. There is simply not enough computing power that exists that can map out every possible move.
In chess, at any given turn, there are an average 35 possible moves. With Go—in which two players compete with polished stones on 19-by-19 grid—there are 250. And each of those 250 has another 250, and so on. As Hassabis points out, there are more possible positions on a Go board than atoms in the universe. (Wired)
The very best human players also rely on intuition, looking at the board and just “knowing” that it “looks” good.
“It’s something subconscious, that you train through years and years of playing. I’ll see a move and be sure it’s the right one, but won’t be able to tell you exactly how I know. I just see it.” (Redmond in Wired)
And that’s sort of what Deep Mind did. They trained the system to learn the game by looking at 30 million Go moves from expert players. And in order to beat those expert players, they then matched the system against itself, coming up with even more winning moves.
So anyway, exciting times to be had. And maybe scary too. Because even if it’s a baby step for AI, who’s to say that the progress from here on out will be limited to baby steps? For more on AI in general, here’s a lengthy, but wonderful read on it, from my favorite blog: The AI revolution: The Road to Superintelligence. And if you want to understand more on deep learning, here’s a great read (haven’t finished it but as a “poet” and non-hacker, I find it very readable): Neural Networks and Deep Learning.
Here are this week’s most relevant news in tech and retail:
- Scientists make a ‘true’ neural network using brain-like chips
- Apple Reports Slowing Growth in iPhone Sales
- Walgreens won’t use “unsafe” Theranos lab for tests
- SpaceX: Nickelodeon is sponsoring a team in Elon Musk’s Hyperloop competition; SpaceX unfurls its astronaut parachutes for the first time
- Oracle’s finally killing its terrible Java browser plugin
- Amazon: Amazon stock plunging after earnings but analysts predict 27% gain; Your impulse buys are costing Amazon a fortune; Amazon to delivery companies: Yes, we’re building our own service but don’t worry
- Under Armour shares jump 17%
- Coach: Upbeat 2Q earnings but revenues disappoint
- Uber partners with Nordstrom, Google, T-Mobile, SAP for UberRush package delivery service
- Chanel CEO Maureen Chiquet leaving company
*Acquired by Google in 2014 and is now part of Alphabet