Life & Death > Machine Logic

Machine Logic

By GREG MAFFETT
Published: January 1, 2011

I had a beer with my Dad tonight.

I know that is not logical as he had the grace to pass away a few years ago. But that does solve a big problem for me up front. Since he's gone, logic tells me I'm writing fiction.

So if I may take a moment, thanks Dad. You are still solving problems years after you left.

He was a lot of things, many things that no one could be proud of. But he could solve problems. He could build a car from the ground up. He could build nearly any machine you needed from the ground up. As long as it was made of metal, wire and hydraulics.

Not many people could spend time with my Dad. That was probably the one thing we had in common. Well, perhaps there was another.

He walked away from math early on. Stopped going to school in 10th grade. Just gave up, he did. You had to spend time with my Dad to have any hope of puzzling him out. He could talk for three solid hours and say nothing. He often did. Getting though that was an test of endurance that few could pass.

But eventually if you logged the hours, you could get to the core. On one such day, he told me that if he had taken the time to learn math, he would have been much better at his job. He knew what he didn't know. He knew he couldn't calculate the answer to a machine design problem. So he just over engineered all his designs. The people he worked for didn't know any better. It was a life of fooling small timers. He knew he was selling something that could only sell in a small market.

But tonight I am watching his mind at work. Or how it would have worked, had he stuck with math. I am watching his mind on math while sipping a strong ale.

I was doing the rather usual act of sitting on a barstool. My grandfather used to sit on barstools and cry incessantly. I know what he cried about, but it isn't relevant. What matters is that he passed on his crying on barstool genes to me. I'd give them back if I could. Doesn't work that way though.

Oh, yes, the mind that I was watching. That mind is currently owned by an Isreali Mathemetician. I spent my evening drinking ale and reading his paper. A paper that made great strides in recommendation science.

What this man was doing with math is exactly what my Dad did with metal. This man built a machine to solve a problem. In this case the problem may have seemed trivial. He was working on the improving the Netflix recommendation software.

I"d come to this paper because I'd noticed that in the last year, Netflix reccomendations had improved dramatically. In fact they just recommended three films in a row that I rated 5 of 5 stars. I only rate maybe 25% of my films that high. But they were on a run. I wondered why. Had they improved their software?

You bet. They ran a competition from 2006-2009. They used real data, so some of that was certainly data I provided. Netflix turned the raw data over to the geeks, the logic guys. Then the geeks did what they do. They tore into in.

I watched the winning team buld the machine the same way my Dad built his machines. There would be a problem he would see. A black box would appear to solve the problem. The next problem would then be attacked. When all the problems were addressed, he would turn it on and see if it worked. It usually worked a little. Then the shims and the tweaks would start. He was tuning the machine. Making up for the parts that sort of fit, resolving tolerance issues.

I watched these math geeks do the same thing with their machine. They built it, they ran it, they improved it every year.

At the end it was a time crunch. It was a contest and they had 30 days to get the most out of their machine to claim the prize. Here it was a pure Return On Investment play. Where do you get the most for your effort?

My Dad knew. If it was a go-kart about to race, he knew if it was the tires or the carb that needed work. If it was an assembly machine, he knew if it was the strength or frequency of the vibrations that moved the parts down the line. He just knew.

These guys doing the prediction model knew too. They made the right call with 30 days to go. Rather than looking for more variables, they knew to look for the blend. This solution required a correct blend of variables, not more variables. In the last moments before the deadline, the found the blend.

And I found I was repeating my grandfathers tears. I was on a barstool 2,500 miles away from the one he favored. I used to think he was crying about his childhood, the life he missed. I've not walked in his shoes, but I think I just sat in his barstool. I think I get it now. We were both crying about the same thing.

For him it, the beauty that could have been his Son. For me, it was the genius that could have been my Dad.

For you?

Any Comments?


More...

» Doing more with less

Saturday

By GREG MAFFETT
Published: September 23, 2012

Saturday, its the equinox, but this has nothing to do with that.

The Pale King

By GREG MAFFETT
Published: December 28, 2011

A few thoughts on how math and fiction relate.