Artificial Intelligence, Machine Learning and Deep learning
Being an ardent lover of the game of Chess, the choice of this subject is easy to explain. In this blog, I would like to highlight the dramatic events that unfolded recently in the game of chess and derive some useful learnings from the same, especially with regards to Artificial Intelligence and Machine Learning.
First of all, the dramatic event I was referring to is about Google’s DeepMind AlphaZero. DeepMind is the world leader in artificial intelligence research acquired by Google in 2014. AlphaZero is a software program, developed by a team at DeepMind, that learned chess in four hours (all by playing with itself), and subsequently when put to test against StockFish (an open source chess engine, rated much higher than any Grand Master), it won with absolute comfort, 28-0 in a 100 game series.
I got enthused by this development and spent some time going through the videos. Based on my analysis, I would like to present few interesting learnings and messages. These messages, I feel, are equally important from our professional and personal life stand point.
- 1. Learn from machines – We, human beings, are comfortable learning from fellow human beings and other living beings. While learning from machines is not a new concept to us, it is not very apparent. My strong opinion is that we should seriously consider this as an additional source of learning. The reason for making this as first and foremost message is substantiated by the next set of messages that follow.
- 2. Unconventional – Barring the exceptional people and/or exceptional situations, most of us prefer conventional methods in our professional or personal life.However, after watching the chess videos and looking at the outcome, I got the impression that unconventional is the way to go. Most of the moves made by AlphaZero defy the convention, and so much so that the chess Grand Masters kept saying “we don’t know why Alphazero made this move; and even if we have to validate this move, against whom will we validate? Stockfish?” For chess lovers, I would like to cite few examples here.
- It’s an unwritten rule that major powers, particularly the Queen, doesn’t take the center stage until the initial development of pieces is complete. However, in many games that AlphaZero played, I noticed the Queen taking aggressive center stage role from the word go.
- Castling is a special move in chess, where king and one of the two rooks move simultaneously with the purpose, that the king is moved away from the center to a more secured place. All this happens in the rank (row), where the king is initially positioned. However, in one match I noticed that the king moves to the third row and secures himself through other major pieces (other than Rook). Should this at all be called “kind of castling”? Also, a King taking active role in the end game is common and understandable. With AlphaZero, I get a feeling that the King is part and parcel of the attack and should be active from the beginning. Something difficult for the chess players to digest and follow.
- 3. Zugzwang positions – In chess, a player is said to be “in zugzwang” when any possible move will worsen their position.For all the chess players, getting the opponent into a Zugzwang is kind of once in a lifetime. In case of AlphaZero, it’s the order of the day. Most of the matches I watched are based on this underlying objective.
- 4. Cramping opposition for space – When you walk upto a chess board where two players are playing the match, the first instinct is to count the pieces on both sides to check the game balance. This assessment is based on material advantage. The one who has more material is considered to be in a better position. Then you would apply the next logic, the positional advantage, and that could give you a different result. In simple terms, having more material does not necessarily yield a win. Sometimes the positional advantage could get you the win. In Alpha Zero case, you would find it playing more for position than piling up the material. Piece sacrifice and cramping opposition for space to achieve this, is a common norm.
- 5. Clear thinking – A machine-defeating human is nothing new. Gary Kasparov (erstwhile world chess champion) lost to IBMs Deep Blue in 1996.The comparison between earlier matches and the current ones is – earlier ones used to be more hardware capability-centric. Those IBM machines were enabled with huge processing power to churn the data and come out with the best move. In the present case, it is more software-centric. With reference to Alpha Zero, there is a significant additional variation to the implementation. AlphaZero analyzes 80k moves per second (MPS), where as Stock fish analyzes 70 Million MPS (1:875 ratio). This point, when read with points 2, 3&4, leads us to the clear thinking ability the AlphaZero possesses. Similarly, in our professional and personal life, if we demonstrate clear thinking, the results could be better.
To conclude, I was fascinated by this development and probably the single most subject on which I spent lot of time over weekends recently. But, when I started converting this into “what is in it for us”? is when I realized the messages we could derive from this.
Hope you enjoyed reading this one.
Latest Blogs
The business world is moving quickly and the only way to make informed decisions is to leverage…
As businesses turn to cloud services to meet their growing technology needs, the promise of…
Clinical trials are at the heart of drug development, producing vast, complex datasets that…
The rise of machine customers introduces essential questions that stretch our technological…