I am reading a book titled “AI Super Powers China, Silicon Valley, and the New World Order” by Kai-Fu Lee. It has captured my imagination as a good story about something technical. I am nearing the end, and the author is presenting how AI will begin to take over in the very near future (like 5 years). He extols the many benefits of automation and the sadness of lost jobs. I thought about it and must tell you that automation doesn’t come easy. I spent 40 years at a company that produced what the world likes to call “zip ties.” At first thought, one can believe that a zip tie is not complicated, so why wouldn’t it be easy to automate? The tie is merely a plastic strip with a molded tip and a locking head on the opposite end. The product’s geometry comprises thin sections and thick sections, sharp edges and soft edges, massive sections like the tie head and body, and tiny delicate sections like the locking mechanism inside the head. Next is the material used to mold the tie. All plastics are not the same. Some are easy to melt and mold but don’t stand up to the rigors of application; the plastic must be tough, flexible, and strong. The material we finally chose to use was Nylon 6-6, but it came with its own list of problems. To mold, the material had to be very dry, but in actual use, the molded product must contain water to make it tough, flexible, and easy to apply. The product will be stiff and brittle if left dry as molded.
Raw nylon came to us as a pellet in boxes or bags. We use a special machine to melt the pellets and then to send the melted plastic into a steel mold with the product’s shape cut into it. This requires a channel cut into the mold that extends to the cavity. To make money, we required the number of cavities in the mold to be more than one. When I first began at the company, our typical mold had sixteen cavities with a binary runner system designed to make each runner extend from the nozzle to the cavity to be the same distance. We melted and molded more nylon in the runner system than in the product. What that meant was a gross imbalance in the cost. We didn’t make money molding runners. A human manually removed each shot of sixteen parts and its runner from the mold. This operator was incentivized to process the maximum number of shots per minute.
The molded product had to be removed from the runner, moisture-conditioned, and packaged before being sent to a customer. Initially, another department did the moisturizing, packaging, and boxing. When I left the company forty years later, we had the entire process automated. As many as two hundred parts were molded at one time. The parts were degated, counted, and packaged into plastic bags of one hundred or one thousand ties. A single person performed quality checks and put ten packages of a thousand ties into a box. We stopped automating at that point because paying back the machinery was more expensive than allowing the existing QC person who tended to four machines to take the final step. What I am getting at here is that AI makes sense, but automating every process may not.
Another aspect was in mold, and molding machine maintenance. When the mold is subject to being squeezed together by a machine capable of applying a million pounds of force to keep it closed during injection, the tiny parts inside the cavity are stressed beyond imagination. The result is that when a tiny part breaks, the product from that cavity is junk, and the process has to shut down to fix the flat. We then turn to quick mold change and maintenance procedures to replace broken parts. All of this is the result of thousands of man-hours of development.
Even the author agrees that automating the human hand motion is not possible at this time nor in the foreseeable future. AI may be great at analyzing orders and finding trends, or it might even be great at finding trends in the molding process, but only with scads of data. It took an entire team of electrical and process engineers years to determine how and what to measure to predict or even see trends. Eventually, we measured the process and improved our product’s quality and consistency.
In the end, we learned that automation comes at a great cost and that the cost of maintaining the equipment continues as long as the process goes on. Changing the process becomes unthinkable once it is solid and running smoothly. When that finally happened, the powers to be decided there was an advantage to sending the whole kit and caboodle to places like Singapore, Costa Rica, and Mexico, where the labor costs are lower. Just to let you know, I left out China. That is because the wisdom of our owner was that he paid for developing his process, and he believed the Chinese should pay their way, too. We had a security system in place that rivaled the NSA and CIA to keep our competitors from stealing our technology.
In conclusion, I say this, bring it on AI we are ready for you, but are you smart enough to take on the challenge in front of you?
Filed under: Biography, Book Review, Manufacturing | Tagged: Artificial Intelligence, Cable ties, Injection Molding |
AI ain’t gonna be used to make useful products. Oh, maybe in the beginning. But already, I think the power monsters are designing ways to give them every possible advantage over their rivals and that will include weapons and ways to shut down any measurable protective ability by the victim nations and populations from these aggressors.
You are right, but I don’t believe they will use AI initially, either. The mechanics are much too complicated to be designed and implemented by computers. My experience with designing by computer is equal to using the computer instead of a pencil, compass, and straight edge.
What is lacking in thought here is the answer to the question of just who will pay money for AI generated products when only 1% of the population has jobs. They forget that the masses need to eat, and when left hungry they rebel, and when they rebel AI will not have the algorithm to defeat them.