Dear MEL
Topic Readers,
Google
DeepMind: AI becomes more alien
Is it an evolutionary
improvement or revolutionary change? Last year and this past May, AI Go machine,
AlphaGo, beat two top go champions by learning how humans
had played the complex game, which was developed several thousand years ago in
China. This itself could mean an AI algorithm, and deep learning combined, surpassed
human ingenuity of thousand years in just a few years.
Now, Google’s
AI hand, Deep Mind, says their new model, called AlphaGo Zero, beat the
original AlphaGo by 100 games to zero only in 72 hours of study. The new model
had had no knowledge as to how the game had been played by humans but the
rules. It played itself without any knowledge or experience and developed
techniques that humans had ever thought of.
It seems that
AlphaGo Zero is more like a self-thinking AI rather than a learn-to-judge style
AlphaGo or other “conventional” AI applications, which have been used to take
humans’ roles. It sounds like the difference between practice and theory, or application
and science. The question is how the new revolutionary AI is going to be used,
especially under already-too-influential Google’s hand.
Enjoy reading
and thinking how you will work with AI in the coming few years.
(because it’s getting too difficult to predict
what the world will be like beyond a few years)
No comments:
Post a Comment