In 1997, IBM created a set of mainframes they called ‘Deep Blue’.
This supercomputer was capable of analyzing over 200 million possible chess moves per second to select the optimal chess strategy.
Deep Blue defeated world chess grand master Garry Kasparov.
But was this any surprise? When you compare that to Kasparov’s brain (processing an average of five
moves a second), Deep Blue logically held a slight
advantage.
But here’s the problem: Deep Blue consumed so much energy that it was a fire hazard.
During the game, IBM had a team standby with fire extinguishers to ensure the darn thing didn’t blow up in flames.
Kasparov, on the other hand, barely broke a sweat.
Calculations vs Outcome
Gerald Tesauro was a computer programmer at IBM. While the press celebrated the amazing achievement that Deep Blue made against the world’s chess grandmaster, Mr. Tesauro saw a problem.
Deep Blue was calculating millions more moves than its human counterpart, and yet it barely won the game.
The human cortex, on the other hand, was way more efficient. It consumed less energy than a
light bulb even when deep in thought.
In Tesauro’s opinion, the quantitative ability of calculations was simply too rigid.
The machine itself simply looked at moves a user made and predicted the possible outcome based on millions of other moves other grandmasters have previously made.
To put simply, it was a complex mathematical problem.
Moves vs Mistakes
Human cognition is more complicated than mathematical problems.
Kasparov may not have had the computational abilities of Deep Blue, but neurons in his brain were trained and refined through years of experience to process not only the move but everything from patterns on the chessboard to strategic alternatives.
Tesauro took a different approach. He wrote a software, which he codenamed TD-Gammon (TD for temporal difference). The program started as a tabula rasa (or blank slate- a fancy useless term I learned. Actually, I take that back. It came in useful once with a consultant neuro-radiologist who was being a smart ass and would only scan a patient’s head if I knew the answer. Fortunately for the patient, my answer saved his life).
According to Jonah Lehrer, author of How We Decide, Tesauro’s program initially made dumb mistakes, as it was entirely random. It lost every single match, but not for long. The program was designed with no moves programmed, but to learn from its own mistakes. Every night, it would play itself and patiently learn what were the most effective moves.
After a few hundred thousand games, TD-Gammon defeated the best human players in the world. The program has revolutionized the use of AI today, from ‘clever’ elevators in skyscrapers that manage efficient usage to control towers in airports that can determine the best flight schedules based on previous learnings.
Conclusion
The human brain is complex. We learn from previous mistakes, often through our brain cells determining our next moves, habits, routines, and patterns based on previous mismatches in dopamine firing rates dependent on our expectations vs outcome.
Therefore I propose that instead of approach every problem we face with the question of ‘what is the best move?’, and playing through all kinds of scenarios, the better question would be ‘based on mistakes from past experience, what is the best course of action?’
Thinking entirely based on the permutations of possible scenarios, outcomes, moves and strategies is valuable. It’s the core of game theory. However, akin to the analogy of Deep Blue, we become ‘prisoners of our mind’ and get trapped in thought loops.
I propose that this can be applied to all areas of life. How to make a business decision. How to figure out something in one’s relationship. The ability to draw from past patterns provides a useful frame of reference to future success.
If there’s no frame of reference to past experience, perhaps one should get frames of references through friends, advisors, mentors and others that they trust.
Otherwise, the outcome will probably require a fire extinguisher.
Any thoughts after reading this? Please share them in the comments below.
Most Commented Posts
No related posts.
