Saturday, July 13, 2013

Dog and ball

Tasha's finally getting some mileage out of her Mazee puzzle ball. It's a simple 3-D maze enclosed in a transparent sphere. You load in a few dog treats and watch her roll the thing around until a treat falls out. Repeat until depleted. Earlier, we made the mistake of putting in tiny treats that fell out way too easily. No challenge. The bone shaped biscuits had more staying power and kept her occupied for way longer.

Yup, a blur of motion she is, chasing her new toy across the floor. We won't have to worry about what she'll destroy next... for a while at least.

Friday, July 12, 2013

Creating intelligence

I've been reading Visions by Michio Kaku. Published in 1997, it envisions the future century based on the work and research that was taking place at that time. About 15 years later, today, already some predictions have come to pass surprisingly quickly, others still need a slightly longer time-frame.

But what I want to share is three paragraphs from Dr Kaku's chapter on Artificial Intelligence. What I see in them is a parallel between the approach one AI branch is taking to create smart machines and the approach we in education default to in trying to create smart kids.

"Although researchers of the two schools of artificial intelligence sit side by side in the same building, the lines between them are clearly drawn. On one side of the debate are the distinguished founders of artificial intelligence who have spent a lifetime programming mammoth computers to model human intelligence. Their inspiration for a thinking machine was a powerful digital computer -- the bigger, the better. Their strategy was dubbed the top-down approach; they believed they could program in the logic and reasoning ability necessary for a machine to think. They assumed that thinking machines -- like Minerva, the Roman goddess of wisdom, who sprang out from Jupiter's forehead fully grown -- would emerge fully developed from a computer.

Their recipe for building a thinking machine was simple: First pour the complex rules and programming into a digital computer in order to reproduce logic and intelligence, then sprinkle on a few subroutines for speech and vision, attach mechanical hands, legs and eyes... Voila! You would have an intelligent robot. Inside that robot's brain would be a complete representation of the outside world, a detailed manual that described the rules for living in the real world.

Their philosophy was based on the idea that intelligence can be simulated by a "Turing machine," which forms the basis of all digital computers. But the traditionalists soon slipped into a quagmire; they profoundly underestimated the enormity of writing down the complete road map of human intelligence. Their computer-based machines turned out to be pathetic, feeble creatures. The mobile robots built on their approach consumed vast amounts of computer power, yet they were surprisingly inept: agonizingly slow and timid, they frequently got lost. They were useless in the real world." (Kaku, 1997).

The other approach is, obviously, the bottom-up approach. This approach is the learn-by-doing, trial-and-error, develop-environmental-awareness-from-your-own-immediate-surroundings method. Dr Kaku is clear that neither is the "superior" approach and that for AI to be truly successful, top and bottom must meet somewhere in the middle.

Likewise for our kids. Based on their somewhat disappointing mid-year performance, they clearly demonstrate how their top-down learning method with heavy reliance on rote gives them a lot of information to reproduce in their essays, but very little ability to string a reasoned argument together that makes much sense. The kids may have a lot memorized, but ultimately it means nothing to them. They act like conduits, passing along information but none of it is real to them. Because they have no personal stake in the info, there is also neither structure nor framework in which they can manipulate the info to any purposeful use.

Below, I'm sharing a video clip that demonstrates how a bottom-up approach does seem to boost development in the kind of discourse required of the kids for GP:

Assuming the subtitles are legit (I'm a bit dubious of the jump-cuts), this 12-year old would ace GP already. At his fingertips are issues relating to politics, religion, socio-economic inequality, law and order, and human rights which he most eloquently discusses as he explains his purpose for not being in school today.

His intelligence stems from simply being aware of his immediate environment, exercising good sense and good judgement based on a value system that respects equality, freedom and eschews exploitation and the abuse of power. He reads the newspaper, listens to people, watches TV, reads from the Internet, and most importantly, uses his own brain. He lives GP, breathes GP, thinks GP.

A kid like this is likely to view with contempt any attempt by his teacher to give him a pre-digested 'reading package'. He might even burn it for being establishment propaganda. Our kids treat reading packages like the Gospel. 'nuff said.