Alpha Zero - The inhuman human!

"Never think you understood a work in physics if all you had done was to read someone else's derivation. The only way to really understand a theory, is, to derive it yourself..." 

- Richard Feynman

"Ramanujan's" (the great mathematician from Kumbakonam, India), E.H.Neville an English mathematician, said, "mind is free, or, shall we say, was the slave of his genius?"

.....and his genius was such that, he transcended the existing level in mathematics that his teachers and contemporaries in Kumbakonam knew, and there were no material available for him to do advanced mathematics. How was this possible!?

Ramanujan had to reinvent most of the existing mathematics, that wasn't known in his hometown, in his formative years - and many later thought this as a constraint and perhaps, he wasted time in doing so.

But, this for me is the key factor! Perhaps, this process of reinvention is what dug deep within him to reach and attain an undiminishing source which churned out incredible mathematical problems and its solutions, which he scribbled in his two notebooks, which the world is still trying to fathom nearly 100 years after his death!

Robert Kanigel, the biographer of Ramanujan writes, "Ramanujan was a man for whom, as Littlewood put it...."the clear cut idea of what is meant by proof.....he perhaps did not possess at all"; once he had become satisfied of a theorem's truth, he had scant interest in proving it to others. Ramanujan truly had nothing to prove. He was his own man. He made himself."


Well, any art.....science.....or sport is a spontaneous activity. They happen.

The great poet Bharathi wrote, ""உள்ளத்தில் உண்மையொளி யுண்டாயின் 

வாக்கினிலே ஒளி யுண்டாகும்"

He essentially means that, if true self knowledge arises within us, then whatever we do will be lustrous!

INTUITION

The key word.....in any field!


Alpha Zero.....as I understand from what I read and seen a couple of games....is playing like a human amongst the machines! Paradoxically, humans are trying to play like machine!

"It uses a novel form of reinforcement learning, in which Alpha Zero becomes its own teacher. It is no longer constrained by the limits of human knowledge."

......and human players are subjugating themselves to the machine thinking.....and constrain their thinking and delimiting their potential to create an art by overworking a position.

What Ramanujan did was this very same "reinforcement learning"..it is not new to us!

Alpha Zero presumably looked at "only" 80,000 moves per second as opposed to 70 million positions per second of Stockfish, and uses that trivial time, released from this futile search for more and more moves and variations, to evaluate a position. 

When, by reducing the calculating potential of the machine in such a drastic way (relatively!!) can make the engine play in such a way.....why don't humans realise the futility of the quest to dig deep.... calculate to near 'perfection'!?

The reason, is that a human master fears defeat......hates to lose and hence is deluding him that if he indulges in deep calculation, he will be able to ward off defeat and start winning....in vain!

This is same as man trying to defy death....in doing so reducing his quality of living drastically.

RISK

....is the key word in Chess.....and in any sports.

Risk - in simple terms is to embrace the third result.....and not try to eliminate it with overworking a position....an opening.....and try to take the opponent by surprise! 

Let us take for eg....

Topalov - Kramnik, Wijk aan zee, 2008

This novelty 12.Nf7!? was of course pre-worked and possibly to some depth.  But as a concept, this is not a novelty at all......for ages such sacrifices are common. But in this position, it is played for the first time because, human mind, starting with early 2000's, has enslaved itself to the engine designs and required an approval from it, before it could muster courage to try it out in practice! 

But, when it comes to the middlegame, Kramnik could give his Rook for four pawns against Harikrishna, even though he wasn't clear and perhaps even knew he could be losing too!  Even Anish Giri gave a piece for three pawns and prevailed in one of the games that he played this year.

Kramnik has played many such games....creating fine imbalance and prevailing.


I can show you innumerable examples from the games of Bronstein, Tal, Larsen....where they gave material embracing high risk and unmindful of losing!

The issue is not about knowing many things.....worst, ideas regurgitated by machines....and to indulge in overworking a position in chess..... It is akin to tear open the goose which is yielding golden eggs in an attempt to loot the treasure at an instant....greedly.

Polya - the Hungarian mathematician, wrote that, "there's a big difference between knowing facts and knowing how those facts fit together (while solving a problem)". Ian Stewart wrote ..... "Mathematical problems aren't uniformly impenetrable. They have their weak points, places where you can insert a probe, waggle it, exert some leverage, chip a bit off. Or, sometimes, crack the thing wide open. Mathematicians can sense things. They don't check the logic of a proof by some laborious calculation with a truth-table; they have the entire plan of campaign mapped out in their heads, and they know when the enemy is on the run. 

What we need is to sense these weak points, to know whether we are making progress or getting stuck. The subtler art is to select the right tool at the right time and to use the tools in such a manner that you get the job done."

The weak points in Chess is created in and by the mind! The art lies in selecting the right tool to tickle or trigger a weak point in the opponent's mind....then .... even the World Champion would commit.....31.Nc6?! (don't be greedy enough for 33.c5? and 36.Qc6??)....as happened in his game against Nepo in LCC, 2018.



Leonard Mlodinow was trying to find the right area for post doctoral research in Caltech in 1973. He was fortunate to befriend the great physicist Richard Feynman and had many deep conversations with him, during his tenure.

During one of his early meetings with the legend, Mlodinow, narrates the following incident where he, in an effort to draw Feynman into a conversation, started the conversation asking Feynman

"Read any good books lately?"

Feynman just shrugged.

"I've been reading about the process of discovery," I told him, trying to keep the conversation alive. I was in the midst of Arthur Koestler's "The Act of Creation".

"Learnt anything?" he asked. He was interested. That was Feynman, always interested.

"I'm having some trouble getting my research on track, and so I thought it might help."

"Yes, but did you learn anything?" asked Feynman.

He was mildly annoyed, because I hadn't answered his question. I felt rebuffed. I wasn't yet sure what I had learned, so I told him about the passage I had just finished. I tried to make it sound dramatic.

"It took place in Berlin, 1914, Imagine a cold spring morning. Outside Church bells chime. In his office at Berlin University, Einstein ponders the still-unfinished theory of relativity. In a lab not far away, in a tall steel cage, a young chimpanzee named Nueva pushes banana skins together in a heap with a stick. She is hungry, and banana skins won't do. As Nueva studies her plight, a professor name Koehler studies her. He like Nueva - and Einstein - has a hunger to satisfy, and his notes are destined to feed many books and papers.

Koehler offers banana to Nueva, only he doesn't do her the favour of placing the food inside her cage. Instead, he places it on the floor outside, beyond her reach."

"A cruel fellow," said Feynman.

"He's challenging her," I said. "To eat, Nueva will have to discover how to get the bananas. First, she does the obvious. She steps to the bars and reaches out. She strains her arms and tries to grasp at the fruit, but the bananas are just out of reach. She throws herself to the ground and rolls on her back in despair.

Not far away, Einstein is nine years into his work on the theory of relativity, and still two years from his great breakthrough."

"And probably feeling a lot like Nueva," said Feynman.

I nodded and smiled. Here we were, Feynman and I, conversing about the frustrations of research. Me and Feynman peer-to-peer!! We were connecting. I was happy.

I continued, "Seven minutes pass. Nueva suddenly stares at the stick. She stops moaning and grabs the stick. It thrusts it out of the cage just beyond the fruit, and pulls it to within arm's reach. 

She has made a discovery."

"And what did this incident teach you?" Feynman asked, not letting me off the hook. I was consciously pleased when I realised that intelligent thoughts were now actually forming in my head in response to his question.

I continued......"Nueva had two skills. One was pushing things around with a stick. The other was reaching out through the bars for things. Her discovery was that she could put the two disparate skills together. It turned her old tool, the stick, into an altogether different kind of tool. Just like Galileo did when he used the telescope, which had been invented as a toy, to look a the sky. A lot of discoveries are like that, new ways of looking at old things, or old concepts. But the raw materials for the discovery had always been  there, which is why discoveries may seem startling at the time, but are simple and obvious to later generations. So I guess I learned something about the psychology of discovery. Something I might hope to apply."

Feynman looked at me for a moment.

"You're wasting your time," he said. "You don't learn how to discover things by reading books on it. And psychology is a bunch of bullshit."

I felt as if he had slapped me. But then, after a pause, he looked me in the eye, and said gently, and with a sly grin, "What I would learn from your story is that if an ape can make a discovery, so can you."


......and when Alpha Zero can play chess that way....thinking speculatively like human, but with great accuracy.....
......human players can also play like the way "humans play".....but by embracing human limitations.....and not......