Businesses put robots to work when human workers are hard to find, argue econo-boffins Researchers have found that business adoption of robots and other forms of automation is largely driven by labor shortages. A study, authored by boffins from MIT and Boston University, will be published in a forthcoming print edition of The Review of Economic Studies. The authors, Daron Acemoglu and Pascual Restrepo, have both …
I agree with you on the whole, but sometimes you can't find low skill workers because, regardless of cost, they just aren't there.
In our fast growing area the fast food restaurants had to curtail their hours because they couldn't hire enough staff - this was before Covid.
Wages went up, but it just had the same limited pool of workers moving to whoever was paying more. The staffing problems remained because of the labor shortage.
One of the restsurants is experimenting with a robotic burger flipper, and waxed ecstatic about it. Always there for its shift, never calls in sick or takes off early, doesn't ask for a raise...
Well, a lot of folks argue Greece, Rome, etc didn't have an industrial age because they had cheap & plentiful slave labor, in the literal sense.
Supposedly the Black Death lead to a middle class and higher wages because, well, it killed off most of the serfs, so workers were in short supply.
Not a historian, but it fascinates me when something as dire as the Black Death led to an overall improvement in human daily existence. Or maybe it's late at night and I'm very drunk watching my 3D printer do its magic.
Hopefully it's the shittiest jobs that get automated first, so people don't have to do them.
Gene,
approximately correct. It was more the townies who died. That caused a concentration of wealth that was then available for investment. Meanwhile it accelerated the need for labour in country and town so the serfs who could, fled to towns post plague for better pay. This did cause a demand for better conditions in rural areas. The lord of local manor found he could not afford to really annoy the peasants as they had other options.
The parallels for us is the abolishing of blue collar jobs from 1970s had consequences as the skills were not replaced. Few skilled industrial jobs exist and the cultural preference for degrees meant even less technical hands on skilled workers got trained. Tried to get a tradie lately ? Combine this with the plummeting birth rates as the middle class expanded and you have shortages. Thats not to say there are no unemployed. West now have a rust belt problem with unskilled humans who have no skills or demand for their labour. So now what ? Historically a self appointed violent messiah complex group or figure arrives, creates massive mayhem, whether winning or losing the regional war
Organisations automate when the overall cost of employing a human becomes higher than that of the machines needed to do the job. It doesn't matter what the age of the workforce is, its all about keeping costs down......which is generally the driver most most business changes.
In California there was a major, lengthy protest to increase the minimum wage in MacDonalds "restaurants" to $15. After a long struggle the protestors achieved their goal. Very soon afterwards MacDonalds rolled out their automated ordering screens as these were now cheaper than the people that used to do the job.
https://www.forbes.com/sites/realspin/2016/11/29/thanks-to-fight-for-15-minimum-wage-mcdonalds-unveils-job-replacing-self-service-kiosks-nationwide/?sh=345fb2564fbc
"Very soon afterwards MacDonalds rolled out their automated ordering screens as these were now cheaper than the people that used to do the job."
If you read the article, it betrays its own premise. If the (location-specific) $15 minimum wage was the cause of the kiosk rollout in 2016 - then why were the screens rolled out nationwide? Why did the lower wage overseas McD's do it first? Why had McDonald's been working on self serve kiosks for a decade before then? The answer is nothing to do with labour costs in cities and states that pay $15 min wage - it's because kiosks increase spend, increase space for kitchen equipment and labour to speed up service, and reduce the amount of counter space required. The opinion piece (NB not news - in the US newspaper tradition, opinion pieces are subject to lower fact checking standards) is partisan hackery.
https://burgerlad.com/mcdonalds-touch-screen-ordering/amp/
https://eu.usatoday.com/story/money/nation-now/2018/06/07/mcdonalds-add-kiosks-citing-better-sales-over-face-face-orders/681196002/
https://www.lamasatech.com/blog/what-mcdonalds-self-order-kiosk-mean-for-fast-food/
It seems "the general public" is under the impression that automation is done for automation's sake. But the cold hard truth is that humans are VERY good at even repetitive monotonous jobs while being extremely "fault tolerant" in terms of supply and output locations, part geometry, part weight, part color, etc. Automating even simple jobs can get surprisingly tricky when you suddenly have to take into account things we always took for granted. If you need screws put into a bracket, for a human, you put down a table, chuck a box full of brackets and a box full of screws on the table (or all over the table), put a box next to the table for the assembled units, show him an example of "put this screw into that hole at least 1/4 of the way" and get on with it. If you update the part a bit and enlarge or shrink some tabs, the human won't care. If you want to automate that (and this is an extremely simple example) you need a way to feed the brackets but this feeding method might need to change if you redesign your part (making redesigns expensive). You need a way to feed your screws, but that too needs adjustment if you decide to use longer or shorter screws. It only gets more complicated from there. Using computer vision? You better hope your parts are of uniform enough color and size that you can handle them without faults. Have weird geometry with no convenient grabbing points? You'll probably need a custom gripper ($$$) that'll be finicky and require frequent adjustment. Etc, etc. Then comes depreciation. If you're automated production line costs 15 million to put in and will last 15 years before it's economically dead, then you'll need to save (or improve profits by) 1.5 million per year. That's about 5 to 10 workers (depending on wages and taxes)
If you're a manufacturer you automate because either you can't get enough workers and it's hampering your ability to deliver on contracts or you need to remove humans from certain production steps because they keep getting hurt or leaving their filthy filthy monkey prints and detritus all over your product. And even then you only automate if in the long run that is cheaper than paying off a few injured workers or just paying a few carefully selected skilled and trained workers more to do a safe, clean job.
I'd argue this is also why we've seen a slow down in the variety of products you'll see within product ranges. Cars all look alike and are made very very similar, certainly within brands. It's because they've automated to a level where doing a major redesign without taking the production automation into account would make rebuilding the production line too costly. So you get stuck with minor Form, Fit, Function iterations where everything works the same, so minimal redesign is required.
Lastly, as mentioned by Filippo above, the first (and possibly only) jobs to get automated are the repetitive, low skill ones which require setup once, then run for a long time with minimal or no supervision. The "put flowerpots on the ceramic glazing sprayer and kiln line" type jobs where you previously employed teenagers and student flex-workers under the supervision of a bored 50 y/o. But if your application requires an hour setup for every new job and jobs last only 45 minutes to an hour, it doesn't make sense to automate. Because now you need expensive equipment, a more expensive worker to set up that equipment and more time to do the same job that would otherwise have been done just fine by 2 less-skilled workers where total cost works out about the same.
"costs 15 million to put in and will last 15 years before it's economically dead, then you'll need to save (or improve profits by) 1.5 million per year"
That should of course read: " then you'll need to save (or improve profits by) 1 million per year" apparently math is hard or something....
That all sounds reasonable. But the state of art isn't going to remain stationary. I was watching some of the videos of Ocado. And computer vision does seem to to be quite good at recognising stuff and tasking robot arms to pick it up, and that stuff hasn't been designed for robots.
Given enough training data, time, fine tuning and optimized circumstances, yes, computer vision works very well. But it's not cheap nor easy to implement. Things have been improving somewhat with the advent of dedicated machine learning systems and the likes, but it's still (most of the time) a finicky process to get working right, and not one you start implementing if 6 to 10 years of hiring low/minimum wage humans works out to be about as expensive.
Computer vision is most definitely NOT a matter of pointing a camera at a conveyor belt, hooking up some robots and pressing a green button. Even the simplest project I've come across required 4 engineers and 6 months to design and implement and another 4 months before it was running smoothly. (And this was for a client that already had good training data as they already had a scanning system that took well lit photographs of their products to use as training and test data).
you mean we haven't ? So many papers are withdrawn as non-reproducible but get cited elsewhere so erroneous conclusions are built into the current soft sciences edifice. But they still get cranked out. And yes, there are (infinite regress coming up) papers about studies on non-reproducable results, outside the venerable AIR and JIR among others
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2023
