“We are on the verge of being able to turn the sand on the beaches into thinking machines.”
I came across this quote as I was reading about artificial intelligence, a subject I am trying to learn as much about as I can while I can still be said to be compos mentis. It got me thinking about sand and all the different ways humans have worked with this common substance over the millennia.
7.5 quintillion grains. According to Gee, my Chatbot assistant. This is the approximate number of grains of sand along all of the beaches and in all of the deserts of the world. That number is 7.5 followed by 18 zeros. That’s a lot of grains.
Quartz—silicon dioxide—is the most common mineral in sand, owing to its hardness and resistance to breaking down. As hard as quartz is, though, it does break down. Sand is often the final product of the slow destruction of ancient mountains. The Appalachians once rivaled the Himalayas in size but have eroded over hundreds of millions of years. Much of the sand on East Coast beaches originates from the erosion of those ancient peaks.
Humans have been using sand since prehistoric times. Early humans likely used sand as an abrasive to grind or polish tools, weapons or pottery. Egyptians used sand as a key ingredient for use in mortar for construction of the pyramids. Romans combined it with lime and volcanic ash to create the first forms of concrete. Sand was eventually melted and glass objects, primarily beads, were created around 2500 BCE.
By the 1300s sand was being used in hourglasses to measure the passage of time. This was critical for, among other things, maritime navigation. To determine a ship’s speed sailors cast a wooden board attached to a knotted line overboard. As the ship moved, the line would unspool. Sailors counted the number of knots that passed by over a set period of time, typically using a half-minute sandglass. The number of knots let out in this time corresponded to the ship’s speed in knots per hour, or nautical miles.
Long after those ships had sailed, the first silicon-based transistors were developed, first by Morris Tanenbaum at Bell Laboratories, in New Jersey, in January 1954, and then again, independently, by Gordon Teal at Texas Instruments, in April of that same year.
The very early transistors could switch states, or perform operations, at the mind-boggling rate of 100 million times per second. Today, with changes in architecture, and the parallel processing of Graphics Processing Units (GPUs), speed on chips is measured in teraflops or trillions of operations per second.
As of June 2022, the Frontier Supercomputer achieved 1.102 exaflops, performing 1 quintillion operations per second which, if my mental calculations are still reliable, works out to be 1/7 of the number of grains of sand on all of the beaches and in all of the deserts on the planet.
To put it another way, it’s quick.
Having been born before anyone had walked on the moon and then to have watched the space age unfold in the 1960s with Mercury rockets and splashdowns and an array of heroes with all the right stuff, who had names like Shepard, Glenn and Armstrong, inching their way into space, I thought that I had already witnessed the most profound moment of my lifetime, of this era.
Now I am not so sure. It is indeed a long arc from figuring out how to get on a horse and then to use it to get from point A to point B, to strapping someone onto a large bottle rocket and using it to go from planet E to planet M, but teaching sand to think for itself ranks up there too.
Twenty years after IBM’s DeepBlue Computer defeated world chess champion Garry Kasparov in New York in 1996, Lee Sedol, a 17-time World Champion Go player, went up against AlphaGo which had been developed by DeepMind, a British AI company, in another meeting of humans versus machines in the context of a board game.
AlphaGo defeated Lee Sedol, but did it in a way that was fundamentally different from the approach taken by DeepBlue. To defeat Kasparov, the IBM program utilized a brute-force strategy based on exhaustive search and evaluation within the defined parameters of chess. In contrast, AlphaGo’s tactic involved learning from data derived from playing games against itself and thereby came to strategies that evolved beyond programming. It’s almost like it used reasoning to win.
From being used as a simple abrasive, then as a building material, to enabling machine learning, the progress that is being made now in Artificial Intelligence is as sure and steady as the sands of time being drawn by gravity from the top of that mariner’s minute glass to the bottom.
All is not lost though. I never wrote down the name of the source of the quote that prompted this piece and my faithful assistant hasn’t been able to find it either. If only I had been thinking.
Keith Krizan can be contacted at therealkbkkbk@gmail.com