The Fabric of Progress: How Pattern Recognition Shaped Humanity's Machines

As part of the Promethean Action founding conference in early June, participants spent a day at the Henry Ford Museum/Greenfield Village complex in Dearborn, Michigan. There are many great museums which give some sense of the evolution of science and technology over the ages, but the main focus of this complex is to bring to life the revolutionary effects on the daily lives of people that have been caused by the mutually reinforcing developments of internal combustion, electrification, mechanization, manned flight, and mass production. For anyone spending any time there, these revolutions are starkly defined.

Today, we are going through several revolutions which will have a similar impact upon our daily lives—the development of cheap, abundant fusion power, easy access to space and the Moon, energy-dense batteries, flying cars, and humanoid home robots able to respond to voice instructions and perform complex tasks autonomously. Prototypes of such robots are already performing useful work in factories and warehouses in the United States and China. But we could also see the germs of these advancements at the Henry Ford complex.

Most people miss it, but in a small building in Greenfield Village there is the ancestor to the responsive humanoid robot. You see, the control system for such robots is commonly called Artificial Intelligence (AI), which is based in the process of pattern recognition. Where else do we commonly encounter patterns? Right! In woven cloth. The Weaving Shop contains a Jacquard Loom, which was invented by Joseph-Marie Jacquard in France in 1801. Jacquard developed a chain of cards with holes in them which determined when pull cords would lift warp threads to create patterns in the cloth being weaved. By 1890, the U.S. census was tabulated using punched cards, and punched cards remained the main way that computers were controlled until the 1970s.

As an ironical side note, the Apollo guidance computer memory was actually composed of woven wire. In this "rope memory," a wire going through a ring was treated as a binary "1" and a wire going around a ring was treated as a binary "0."

Of course, Thomas Edison invented the phonograph and the movie to record sounds and sights, and Ottmar Mergenthaler invented the Linotype machine to record text to be printed, but the recordings produced by these were not explicit step-by-step instructions used to control a machine.

In contrast, the Greenfield Village carousel of 1913 has an organ, drum, and cymbal which are controlled by a roll of paper containing punched holes. Pianos and organs controlled by paper rolls were once very common. Unlike the phonograph which would record and replay sound, the paper rolls controlled individual keys and instrument levers—they told the instrument what to do, and when to do it, in order to perform a song.

Since the Apollo project, we have made huge strides in recording memory. "Hard drives," which bear some resemblance to a digital phonograph, were created to store long-term memory, and various types of solid state short-term memory were created. Now solid state memory has become so tiny, cheap, and robust that it has begun to also be used for long-term memory. Then, the development of the internet and data centers has put nearly the entire written output of mankind into the palm of your hand. So, memory has come a long-long-long way since the days of Joseph-Marie Jacquard. But more is on the way.

What is today called "Machine Learning" (ML) is sort of Jacquard's process run in reverse. Can a machine look at reality (in the form of single images and images in succession– video) and distinguish at least some of the patterns of reality which underlie the images? If the memory (annotated data set) used to train the ML system is large enough, the ML system becomes able to match patterns with a high degree of success in limited areas. In effect, if memory becomes large enough, even a dumb machine can seem to be intelligent within certain areas of training. Of course, it is not intelligent. It is simply matching patterns, but that is very important to industrial civilization!

Already, very complex activities like driving, flying, and landing rockets on their tails, are being performed robotically. Soon, cooking, cleaning, mining, farming, assembling, etc. are going to be performed by humanoid robotic systems which can adapt to their surroundings and respond to relatively simple human voice commands. Despite ongoing sabotage of that progress, the seeds are germinating. And as we defeat the oligarchy tearing down our society, the next Trump administration will open a revolutionary bright future for everyone on and off the planet.