Episode Summary:
Artificial intelligence has historically been the most predicted development in information processing technology. Man-made intelligence has been a part of folklore since Mary Shelley wrote Frankenstein and has been a staple of science fiction for decades. While true AI remains elusive, the state-of-the-art has progressed far enough to have major disruptive effects on economies worldwide. Will the engineering professional of the future even understand how AI does what it does? Jim Anderton comments.
Access all episodes of End of the Line on Engineering TV along with all of our other series.
Transcript of this week's show:
To see any images, graphs, charts, graphics, and/or videos to which the transcript may be referring, watch the above video.
Machines that think. Artificial intelligence is one of those rare innovations that has been discussed in popular culture for over a century, long before the invention of computers. Frankenstein is about artificial intelligence, and the recurring theme of thinking machines, especially machines that run amok has been a part of the work of filmmakers like Fritz Lang and the great science fiction writers like Heinlein, Clark and Asimov.
Those writers were very optimistic about the timeline for the development of true artificial intelligence, but machines, like humans, begin by mastering simple, repetitive tasks. We learn to walk before we can read or write, and in 1961, an actual breakthrough occurred that was little reported then and is rarely remembered now: A Unimate robot began a simple machine tending job at a General Motors die casting plant in New Jersey. And simple it was. Die castings straight out of the mold are hot and dangerous to handle. The robot arm could take product out of the machine and drop it into a bin for later processing.
It was a small step, but 50 years later, 400 thousand U.S. manufacturing jobs have disappeared, both to outsourcing and automation. And that’s the sequence of job loss in manufacturing: first the work moves to low-wage jurisdictions, then frequently re-shores back to America, but with machines replacing much of the human labor. In the low-mix high-volume world of most mass production however, the general purpose pseudo-intelligent machine that we think of as advanced robotics today simply isn’t necessary. It takes a surprising amount of skill to layout a steel blank and then drill straight, concentric holes in it, but a well-designed fixture and jig turns that semiskilled process into something a chimpanzee can do. Or a machine.
And that’s the critical difference between AI driven automation and conventional factory robotics. Current shop floor automation basically uses robots to replace human hands, doing similar things. An autonomous forklift truck handles pallets in the same way a human does. But an AI driven factory operation will highly likely look at processes differently. What if the natural constraints of human physiology are pulled out of the equation? Will we still palletize products? Or is it more efficient to have hundreds or thousands of small drones handling individual parts or cartons, stacking them or even passing them from machine to machine until the truck is loaded or unloaded? Can a robot pick up a delicate part, then throw it 20 feet to another robot? Today it can. Will factories as we know them today, with linear, stepwise production processes be recognizable in the future? I doubt it.
And most importantly, will human engineers be capable of understanding almost infinitely complex product and part designs and production processes that make them? I doubt that too. I don’t understand how the microprocessor in my car engine ECU processes sensor data to decide on a fuel injection strategy. But I know that someone does. And that person is working on better hardware, firmware and software to optimize the process.
But creeping around the edges of high-tech, military and manufacturing applications, software and some hardware is now developed by machines in ways that cannot be understood and verified by human beings at a granular level. Is this bad? Not necessarily.
But if you’re wondering whether you’re going to have a job in industry 10 or 20 years from now, I suggest you look at two technologies: simulation, and generative design. AI is not yet creative, and that’s the core of the engineering art. There was a time when I said with some confidence that AI never will be. But I don’t say that anymore.