All current automation is rote repetition of actions, database handling, or positional prediction. These get very incredibly complex, yes, but that complexity is wrangling the input data to something the machine can make sense of. Maintenance requires adaptive, very active, problem solving, because you do not go in knowing what is wrong. The machine has to independently make the diagnostic, and independently come up with a solution.
Given we still run into situations like million-dollar heavy machinery failing due to ambient vibrations and expensive trips get taken to find the billion-dollar company forgot to turn on the power, this is a long way off. This requires extremely strong general artificial intelligence. Maintenance is one of the very last jobs you can automate, because you have to reliably automate all of design and construction before you can even start on being confident it'll work.
Reason for that being the simple matter that to fix it, it has to understand how to make it, so that it understands what it is that needs fixed. And you're adding quite a few steps between "make it" and "fix it".
What machine learning is increasingly teaching us is that complex tasks are actually a series of simple tasks and machines are gradually breaking them down and automating them, thus increasingly complex tasks are being removed. And the thing is, we're already heavily automating design and construction. Manufacturing was literally the
first thing to start getting automated. As for design, AI is literally starting to do that
right now. We have numerous devices and tools and even
entire cars that were created entirely by AI.
Computer algorithms are painting pictures, composing music, and designing cars. This represents a revolution in the worlds of art and design. Here you can find out all about the creative applications of artificial intelligence.
www.bmw.com
Heck, we've got a thread here on the Sietch by a member who's having an AI produce art. AI Creativity is a thing that's happening, design is not the terrible hurdle it's being put forward as.
Imagine, already on the front of maintenance, if you take your car to a mechanic one of the first things they'll do is plug a diagnostic computer into it to get codes telling them what's wrong. It's not much of a step from there to the codes automatically downloading to a maintenance robot. This then lets it determine which part is malfunctioning.
From there it's possible to automate more maintenance. A simple decision tree determines if the problem is software (automatically debugged by the diagnostic computer) or hardware (Modular part to be replaced by machine arms). Most of the tasks done by a human can be gradually automated away as tech advances, until you have a self-driving car that drives itself to the automated garage where it gets an automatic oil-change and the automated garage uses the diagnostic computer to determine if any parts are beginning to fail, using visual recognition to determine if the tires are worn, etc. The garage orders parts automatically, schedules the self-driving car to come back in a week, bills the customer's credit card, and replaces any parts as needed next week This entire process takes place at night while the car's owner is asleep and won't need the car.
At first, this will just supplement the human mechanic and can only handle routine issues. Then each year, the number of issues the automated garage can handle will increase as the mechanic's tasks are increasingly broken down. In a few years, one mechanic handles all the tasks that previously took an entire team and really only gets called in for ridiculously complex problems the machine couldn't handle. 3D Printing technology is eventually implemented into the garage so it doesn't order parts any more, just raw materials and manufactures whatever it needs on-the-spot for repairs. A few years after that even more complex machine learning teaches the computer how to modify existing part schematics to print customized parts for specific situations, allowing it to remove the one mechanic whose job hinged on handling edge cases where normal parts wouldn't work. A few years after that all the tasks are fully automated and the only mechanics left are hobbyists.
www.cnbc.com
