[Computer-go] 50k-100k patterns

Jim O'Flaherty, Jr. jim_oflaherty_jr at yahoo.com
Wed Aug 8 08:15:11 PDT 2012

Here's a more accurate article about the synaptic pruning:

 From: "Jim O'Flaherty, Jr." <jim_oflaherty_jr at yahoo.com>
To: "computer-go at dvandva.org" <computer-go at dvandva.org> 
Sent: Wednesday, August 8, 2012 10:10 AM
Subject: Re: [Computer-go] 50k-100k patterns


My memory is a bit foggy on every detail and timing, but what follows is my own "thoughts" about learning Go and language acquisition.

The term in biology and around neuroscience is "plasticity". It's a measurement as to the degree of "changeability" the brain has at any given time. What has been observed across many different domains is that this plasticity is extremely high in the brain until around pubescence. And then overall it begins to degrade (i.e. moving towards the old adage "it's hard to teach an old dog new tricks") throughout the lifetime of the individual. Of course, the brain has different modules which have different rates of degrade, some faster than others.

Another method of referring to brain plasticity has to do with "development windows". There is substantial empirical evidence suggesting that the degree of plasticity for a child between birth and 6 years old is considerably higher than that of a 6 to 12 (pre-pubescent) year old. And that it drops again around 18-20 year for females and around 22-25 for males. After that, it plateaus for about a decade and then begins a slow fall off (again, this is for the overall brain, not specifically for each particular module/area).

Consider your own empirical evidence about how Go is taught in Japan and China. If an adult brain could learn as quickly and easily as a 6 yo brain, then the 8-10 years it takes to teach a 6yo to become a professional Go player ought to be roughly equivalent to an 20yo achieving a similar thing by 30yo. This just does not happen as a general rule, even for those who become consumed by the game and immerse themselves in deep study with teachers. And notice the age distribution of the title holders throughout the years. While there are statistical outliers, the bulk of the winners fall within a pretty narrow age range. This is the emergent effect of how plasticity works with the brain. And it's reasonable evidence to imply attempts to work outside the plasticity model generally fail (with the statistical outlier exceptions).

The bottom line is that the brain has an aging and adaptation model that strongly follows the same biological models as the physical body. The young body is more pliable and more energetic. If you doubt this, notice how the Olympians for each sport typically fall into a very narrow biological age range. It's just the nature of organisms in general, and of homo-sapien in particular. So, teaching a young child's mind patterns while they are more impressionable will occur more rapidly and with less resistance (consider there are less things for which those patterns to compete with in terms of utilizing neural connectivity).

So, while there is truth that an adult can do work to acquire an accent a child picks up unconsciously. The adult must always be consciously choosing the accent where the child never does. This points to where in the brain the bulk of the connections had to be made. And it also points out that the adult must expends the more scarce "concsious focus will poower" resources whereas for the child, it's leveraging much more abundantly available sub-conscious/intuitive aspect of their brain.

One other thought that might help in seeing that a child's mind and learning is different at a fundamentally biological level - the great pruning. From fertilization until around the age of 6, a child's brain is continuing to grow new neurons and connections. The activity can be described as frenetic. However, around the age of 6, the neuron growth expansion slows down. And between the age of 6 and 10, the child's mind engages in a slowly accelerating process of neuron and dendritic pruning. By the age of 10, more than half of the neurons have disappeared and a large percentage of dendritic connectivity has dissipated. It's the first "wisdom over direction perception" cycle; i.e. the brain learning the abstract pattern and then letting go of much of the empirical facts that helped originate the pattern. This particular notion typically makes the OCD types quite nervous and uncomfortable, a category into which many Go players would fall {smirk}).

So, if this "great pruning" has any real truth to it, it would likely be a large influencer in why both acquiring a language, an accent and learning intricate Go patterns prior to 6 are then "hard coded" at a very deep and abstract level into the resulting pruned network by the time the child has reached 10 years old. In effect, they have been imprinted deeply and subconsciously/intuitively in a way no adult can possibly match.

Fascinating tangent! Totally related to may own AI research related to Go.

Thank you,


 From: Darren Cook <darren at dcook.org>
To: computer-go at dvandva.org 
Sent: Wednesday, August 8, 2012 7:12 AM
Subject: Re: [Computer-go] 50k-100k patterns
> Well, it might not be an irrefutable theory, but it's a lot more than
> just a myth. Linguists and neuroscientists have been studying the
> issue for decades. Check out
> http://en.wikipedia.org/wiki/Critical_period_hypothesis

Thanks for the interesting link.

OK, "myth" was overstating it. But "I feel this is a theory that people
should not assume is true, because there is a large body of
counter-evidence" doesn't roll off the tongue so well :-)

Something I suspect a lot of people on this list might have tried is
learning Japanese kanji. I meet a lot of Japanese learners (even as
young as in their early twenties) who throw their hands up and say they
are too old to learn kanji. They've heard this myth (or whatever, see
above) that only children, with their sponge-like brains, can do it.

In fact it appears to be the opposite: my children go to the local
Japanese school, so I've
 been able to see firsthand that Japanese school
children spend at least one hour/day, 5 days/week, plus considerable
homework/juku, for the first 6 years, dedicated to learning the first
1000 kanji. That is 2000 to 3000 hours. Plus using those kanji in all
their other lessons.

Adult learners, who actually try, can learn the same amount of kanji in
far fewer total hours.

There is a very under-rated theory, "premature literacy" (perhaps it is
more well-known under another name??), that says children shouldn't
learn to read/write their mother tongue until they are 10 or 11, as the
brain isn't ready for it. (Or, I've even heard claims it damages the
brain, so it cannot learn other stuff so well.)

And, taking that further, so far in fact that I end up back on topic
(!), it has always seemed to me that go patterns are more like written
characters than sounds.

>> More evidence that adult
 learners can learn new accents: the existence
>> of impressionists.
> Impressionists like Debussy or like Degas? Or perhaps you mean
> "impersonators"? :)

Meaning two here!


Darren Cook, Software Researcher/Developer

http://dcook.org/work/ (About me and my work)
http://dcook.org/blogs.html (My blogs and articles)
Computer-go mailing list
Computer-go at dvandva.org

Computer-go mailing list
Computer-go at dvandva.org
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://computer-go.org/pipermail/computer-go/attachments/20120808/46d8ab3f/attachment.html>

More information about the Computer-go mailing list