The eternal debate.

Posted by Pointman on June 16, 2020 · 14 Comments 

There is this eternal debate about bringing up children to be successful, adjusted and high achievers. It’s usually framed in terms of as nurture versus nature. Is it the singer or the song? I’ve always felt it to be a silly debate no matter how heated it got. It’s one of those lightweight intellectual spats. Hand bagging each other at twenty paces, which is meaningless. Stupidity. If you’re a student of human nature, and let’s face it, we all become one in the end, you soon realise you’re looking at a spectrum, a range between those two extremes.

We’re the only creatures on the face of this planet who possess this unique thing called consciousness, which means being creative but also inherently unpredictable, because you can never 100% read what’s going on in that brain that’s being contained between those two ears.

Children have this habit of coming at you out of left field. The ones who were by all expectations expected to be so promising, turn out to be very disappointing. They fall to indolence, drugs or whatever. The one who on all his end of year grade reports would be labelled as special needs these days, screwed up all his final exams and traded down to do teacher exams instead, which he also failed, eventually found work in Switzerland as a patent clerk. In his evenings and weekends he came up with theories, the first of which was the special theory of relativity and the second was called the general theory of relativity. That dimbo at the back of the classroom playing catch up with his class mates was called Albert Einstein.

Those two papers instantly blew the accepted knowledge about all physics straight out of the water. Most change is slow, tectonic, millennial and creeping but sometimes it’s frightening fast. In March 1862 in the American civil war, two of the first ironclads, the Monitor and the Merrimack, battled each other for hours, bouncing cannon balls off each other’s armour. An English military observer of the battle sent a brief communique back to Admiralty House – “Gentlemen, our glorious wooden navy has today become obsolete.” It can be that sudden.

At a certain point in my life, I undertook some serious research into artificial intelligence or AI to use the parlance. I went the usual route, first what’s called procedural. I wrote programs that were in essence cascading sequences of if-then-else that quickly got to be massive programs. To give you an idea of that approach, let’s start designing a procedural AI to identify whether some creature is a dog or not.

Does it have four legs? Most of the time, but since we’ve all seen a three legged doggie a victim of a traumatic amputation scampering around and about, not always.

Is it big or small? It can be tiny or huge.

What colour is it? It can be red, black, mottled or pure white.

Does it bark like a dog? Mostly, but not always. Sometimes when you stroke them, they purr like a cat on your lap.

You can probably see where I’m going with the central flaw in the procedural approach to AI. There are simply too many variables, variations, possibilities, and that’s before you get into the dreaded unknown unknowns zone. It’s called the combinational explosion. You, two battalions of mathematicians, analysts, designers, a few billion dollars of budget and programmers could be working on it for the rest of your lives and get absolutely nowhere.

Point to any type of dog, no matter how big, small, speckled, coloured or different and ask the question of your average two year old. What’s that? It’s a dog they’ll reply, while looking at you like you’re some sort of retard. The human brain is the finest pattern recognition mechanism on the face of the Earth, and despite all the confident bollocks you’ll read in the literature, nobody actually has the first bloody clue about how it actually works. The brain is just another organ of the human body, but the mind it manifests is always undiscovered country you enter at your peril because here there be dragons.

That avenue of research wasn’t an intellectual indulgence. I’d seen the most vulnerable people die to some diseases we’d already banished back to the far history of the middle ages. The personal computer was just coming in, it was third world affordable and I wanted to throw something a lot better than a hand grenade into the mix. All that was needed was a basic diagnosis mechanism.

Having recognised I was going nowhere, a change of approach was obviously needed. That turned out to be neural networks. In essence, instead of designing procedural software that you pointed at things and wrote orders for them, I started creating learning machines, but the downside of that approach was I lost control. They were evolving, learning, thinking for themselves. What I lost on the neural network side was what’s termed scrutability.

Type in five or six answers and you might get a diagnosis of pancreatic cancer, but how that decision had been made was enigmatic. If there was one thing I’d learned on the procedural approach, it was that acknowledged and accomplished professionals have real problems explaining their expertise, but more importantly, their certainty. It’s intuitive, has some basis on the symptoms being presented, but at best they’re guessing but usually very accurately.

As always with these areas of blue sky research, the usual questions are the simplest answered. Who? Where? When? How much? but the big daddie question is always why? Why or what on Earth makes someone start wondering about the square on the hypotenuse being equal to the sum of the squares of the other two sides? How can red, green and blue be mixed to make any other colour? Why is there a rainbow? Why do we need rainbows at all?

I had a few more determined attempts from different avenues of approach but what I came to realise was two things. The first was that I could construct something decent, but only for a small and defined problem domain. If you wanted to know the perfect pitch angle of an attack helicopter’s blades, I was your man, but what type of flowers your paramour would love for her birthday was a whole different problem.

The second is harder to nail down. Perceptiveness, intelligence, curiosity, desire, emotions, feelings – they’re all so elusive that nailing jelly to a tree would be easier. If you can even come up with the first idea of how to program that lot, the Nobel prize is yours.

One of my sons, who has two sons of his own, was play wrestling with the older one of four year’s vintage. He’s an excellent father and was rolling around on the living room floor letting his oldest win, but making him work for it. Every decent dad has done it. I think of it as cub play while keeping it all safe.

His second son, of two years vintage, entered the living room and totally misapprehended the situation. “Get off my brother” he said and was instantly preparing to charge, all two foot nothing of him against a grown man.

My lad quickly defused the situation, but it was the perfect example of why packing the human mind into a neat AI box will always fail because we’re a lot more advanced than algorithms. For better and sometimes for worse, we’re unique. We have feelings, which is what drive us.

©Pointman

Filed under Article