In my last post on heuristics and approximation, I asked the question - *Given realistic scale and scope, how does the human brain solve such problems (like question-answering)? Does the brain map everything to combinatorial optimization? Can its modus operandi be applied to automated systems so that they solve our problems....* - which leads to further interesting exercises in intellectual masturbation.

Combinational Optimization is about finding the optimal (best/worst) candidate (based on certain acceptable parameters) among all possible options. A large category of problems can be mapped to combinatorial optimization, and most of these are NP-hard. This means that most of them do not have efficient solutions.

But we see humans, with some training and experience, solve many combinatorial optimization problems efficiently using heuristics and instinct. Chess, Go, Sudoku, etc., come into mind. In these games, the human has to search the solution space efficiently to find the correct board/grid position. And we know that humans do not do a brute force search. There is some internal process that short-circuits the search.....somehow.

Now, this somehow-situation could be explained by the following possibilities:

Anti-climax won't do. We need 'em real. That was a long session. Now for the money shot. [insert appropriate onomatopoeia here, in uppercase]

Combinational Optimization is about finding the optimal (best/worst) candidate (based on certain acceptable parameters) among all possible options. A large category of problems can be mapped to combinatorial optimization, and most of these are NP-hard. This means that most of them do not have efficient solutions.

But we see humans, with some training and experience, solve many combinatorial optimization problems efficiently using heuristics and instinct. Chess, Go, Sudoku, etc., come into mind. In these games, the human has to search the solution space efficiently to find the correct board/grid position. And we know that humans do not do a brute force search. There is some internal process that short-circuits the search.....somehow.

Now, this somehow-situation could be explained by the following possibilities:

- The human brain has a clever technique to narrow search space, and come up with exact solutions. This would mean that P==NP, and computer scientists still haven't found a proof for this. A proof of this would simply be an efficient way of solving one of the NP-hard problems (which, according to this bullet, the human brain is doing right now). We just need to know this purported technique formally to claim P==NP. This situation is highly unlikely because computer scientists have been trying since 1960s to find such a procedure in vain. Of course, their fastidious nature is also making them look for a proof that such an efficient procedure cannot exist (P!=NP). Whether the latter will be proven, or whether the question itself is beyond answer - only time will tell. But the former is very unlikely to be true.
- The human brain has come up with incredibly clever heuristics that approximate the search for the exact solution, and give us only approximate solutions. But these are so good, that they look exact almost always. But the fact that Kasparov got defeated by a computer shows that his heuristics used for chess were not infallible, and the computer's brute force search was much better (after all chess is played on a 8X8 board). Go and Bridge players still kick the machine's ass because of their games' bigger search spaces. This seems to be our Winston Wolfe (see Pulp Fiction) for the somehow-situation.
- The outlandish situation that my romantic mind wants to use to model itself is a trifle too....er....outlandish. Let me get there. It is known that Factoring a large number into its factors is a computationally hard problem. Its unlikely that a clever technique exists to find exact factors without going through brute force division. But Peter Shor, in 1994, proved that quantum computers can solve the Factoring problem efficiently! Now, quantum computers are still theoretical models which might never get built. Shor and co. coming up with algorithms that work on them motivates physicists and engineers to hopefully come up with a working model of the quantum computer in the future. Now, my outlandish conjecture is that the human brain already has a quantum computer like mechanism built into it, which enables it to solve certain combinatorial problems quickly. Modern physics accepts the reality of quantum mechanical concepts like superposition, etc. in nature. Why is it not possible that such features already physically exist in our brains, and this quantum computing brain is solving problems using quantum algorithms? Is the human brain a realistic quantum computer? Ummmm.....

Anti-climax won't do. We need 'em real. That was a long session. Now for the money shot. [insert appropriate onomatopoeia here, in uppercase]

Labels: computer science

Comments:
*Will we ever truly understand ourselves scientifically? or in the end, is it going to be just spiritual enlightenment about the Self? How anti-climactic would that be.*

Anti-climatic, just like The Matrix.

Some amusing thoughts on the quantum brain theory.

Why do some people perform better at some tasks than others. Do they know better algorithms? It might happen sometimes, but we sure dont train in quantum algorithms before solving a sudoku. So its the brain architecture tht varies. This architecture has to be transferred genetically (obviously with evolution). So if the quantum brain theory holds true, there exists a fabricator (in a human female :) tht creates such a computer based on a given genetic design. More reason to study females.

-sunny

*Will we ever truly understand ourselves scientifically?*

All I can say is that, "No system can explain its own working". You need a higher form of intelligence for that to happen. That is why the entire realm of AI has not moved much from the promises shown initially.

<< Home

Anti-climatic, just like The Matrix.

Some amusing thoughts on the quantum brain theory.

Why do some people perform better at some tasks than others. Do they know better algorithms? It might happen sometimes, but we sure dont train in quantum algorithms before solving a sudoku. So its the brain architecture tht varies. This architecture has to be transferred genetically (obviously with evolution). So if the quantum brain theory holds true, there exists a fabricator (in a human female :) tht creates such a computer based on a given genetic design. More reason to study females.

-sunny

Nice post! :)

Even someone like me who is a self-proclaimed non-techie and for whom technical questions do not present much interest, found this particular discussion rather interesting.

It has got me thinking a bit about the human brain and its internal workings...and I'm no closer to the answer than the next guy.

Even someone like me who is a self-proclaimed non-techie and for whom technical questions do not present much interest, found this particular discussion rather interesting.

It has got me thinking a bit about the human brain and its internal workings...and I'm no closer to the answer than the next guy.

all of that post has interesting stuff... but i'll pick that has kind of always troubled me. the anti-climax. and another related thought i've often had. suppose, we truly understand ourselves scientifically someday, what about newton or einstien or anybody else who wondered and wanted to know this truth, say me? :D jokes apart, i do want to know as much as they did. so what about us? we will be dead and gone, but where is our desire to know getting satisfied? will it get satisfied? may be that's also part of the truth. but it is frustrating to think about it.

Dude, can I just say THANK YOU, for introducting me to The Ladder Theory! It just all begins to make sense now....

A couple of days ago, I realised cracking Sudoku was not all that enjoyable anymore. Here's why:

Human brain seems to operate at two levels. There is a computer-ish brain that follows set logical rules. At a higher level, there is a pattern recognizer (PR) at work. The PR part of the brain passes sets the logical rules for the computer to carry out. Quite often, the PR works at an unconscious/sub-conscious level. In my case I realised that while solving a sudoku, I was actually following a set of rules without consciously coming up with them. In hindsight, I had formed these rules by a combination of pattern-recognition and brute force. But now that I've become conscious of the rules, solving a Sudoku is nothing more than running an algorithm through another set of input data. The joy of solving the Sudoku was in the using the pattern-recognizer part of the brain. But now that is gone forever - or at least until I try solving a Sudoku that is tougher than the ones I usually solve (and hence would require some more activity on the part of my PR).

Human brain seems to operate at two levels. There is a computer-ish brain that follows set logical rules. At a higher level, there is a pattern recognizer (PR) at work. The PR part of the brain passes sets the logical rules for the computer to carry out. Quite often, the PR works at an unconscious/sub-conscious level. In my case I realised that while solving a sudoku, I was actually following a set of rules without consciously coming up with them. In hindsight, I had formed these rules by a combination of pattern-recognition and brute force. But now that I've become conscious of the rules, solving a Sudoku is nothing more than running an algorithm through another set of input data. The joy of solving the Sudoku was in the using the pattern-recognizer part of the brain. But now that is gone forever - or at least until I try solving a Sudoku that is tougher than the ones I usually solve (and hence would require some more activity on the part of my PR).

quite agree with ev there. probably the only challenge left in sudoku is how fast you can solve a really nasty one or something of that sort.

All I can say is that, "No system can explain its own working". You need a higher form of intelligence for that to happen. That is why the entire realm of AI has not moved much from the promises shown initially.

Thanks for visiting my blog and also for that wonderful list of places to go.. Well I was born in mumbai and i stayed there for 20 years of my life. Right now iam trying to know Karnataka for thats where i might stay the rest of my life..That list was really interesting and helpful cos Coffee Days really dont interest me and these old places are more fun to experiment..

Post a Comment
<< Home

February 2004 July 2004 August 2004 September 2004 October 2004 November 2004 December 2004 January 2005 February 2005 March 2005 April 2005 May 2005 June 2005 July 2005 August 2005 September 2005 October 2005 November 2005 December 2005 January 2006 February 2006 March 2006 April 2006 May 2006 June 2006 July 2006 August 2006 September 2006 October 2006 November 2006 January 2007 April 2007 May 2007 June 2007 November 2007 December 2007 January 2008 March 2008 June 2008 February 2009 June 2009 February 2010 November 2010