I have played around with heuristic algorithms since 1970, but had not pushed myself very far. In 1970 a very simple heuristic algorithm called the Wolff machine, after its designer, was introduced to us by Dr. Bob Northcote at ICL’s Software Development Centre in Adelaide.
“But there's another, more roundabout way. You can also feed the machine labelled photos of cats and dogs, and ask it to teach itself how one species appears different to another. Instead of telling it the rules, you ask it to formulate its own set of rules through close observation.”
So quotes like the one above don’t surprise me. That has been my understanding of heuristic learning for 50+ years.
"AI is crafted by humans to do problems that humans can't cognitively achieve themselves," she says. "It allows us to access more of what we don't know, sooner."
But this quote (above) impressed me.
Consider a parallel quote about computers and simple numerical calculations. I was hooked on computers the day (literally the night) I managed to write a program that could calculate the determinant of a matrix. My limit was a 5x5 matrix, and even then (with pencil and paper) I used to calculate three times and chose the best-of-three answer.
Here was a device that used the same program over and over again. Once I had the program working, it would run any day of the year, with any size matrix I gave it.(1)
The next night I ran the program twenty times on a single matrix, held the twenty output punched cards up to the ceiling lights, and every hole aligned. Identical Results!
Today I think “of course”, but back then the concept of infallibility changed my mind.
Computers, I learned, would allow me to access more of what we don't know,
I wonder if the same thought crossed the mind of the first carpenter to use a panel saw?
(1) Well, it was FORTRAN II on an IBM 1620, so a 20x20 matrix was probably out of the question. But the principle was sound.
Cheers, Chris