Slashdot Mirror
Will Humanoid Robots Take All the Jobs by 2050?
Anonymous writes "Marshall Brain (the guy who started HowStuffWorks) has published an article claiming that robots will take half the jobs in the U.S. by 2050. Some of his predictions: real computer vision systems by 2020, computers with the CPU power and memory of the human brain by 2040, completely robotic fast food restaurants in 2030 (which then unemploy 3.5 million people), etc. It's a pretty astounding article. My question: How many people on /. think he is right (or even close - let's say he's off by 10 or 20 years)? Or is he full of it?"
Any economist will tell you that this guy has no clue what he's talking about. Maybe robots will be around and maybe not. The fact is that there is an infinate amount of work to be done, not some limited supply that is portioned out. This is basic, basic, economics you'll discover in any book on the subject.
"To lead the people, you must walk behind them"
In a fly-by-wire aircraft (which is a lot of recent large passenger planes) you already bet it on the computational prowess of machine. It might be (is) several machines with different software comparing/contrasting/voting and monitoring each other, but machine it is - and if it decides the engines won't throttle up, then they won't, no matter how hard the pilot pushes the stick.
Better not fly on an Airbus.
They are already using computers to limit what the pilot can do.
Ok lets look at a number of problems
When can we expect good computer vision? There are lots of progresses in the field. New statistical techniques. Faster algorithms for supervised learning. But still. I guess if you had asked 30 years ago when perceptrons were quite fashionable how long it would take to have real good computer vision you would have gotten the same estimate of 20 years. Doing some work in computer vision I must say that to my knowledge we are still very far from building anything thats real. We are rather at the stage where we discover 2 new problems for each problem solved. Problems are for example: Attention, efficient learning, efficient inference, symbol grounding, categorization. So I guess it will take many more years. Or forever.
What about self repair ? One of the really cool things about humans is that they mostly repair themselves. Our bodys endure constant abuse. Our bodies constantly repair the damage at least over approximately 100 years. A large number of robots would demand constant repairing.
Are robots really cheap? Lets face it people are there. We already have a very high rate of jobless people. Given the right taxation systems these people should be a lot less cheaper than any robot could ever be.
Dont get all of this wrong. Computer Vision and Robotics will improve. But it will improve the same way that tools improved throughout the history of mankind. They slowly get better and more useful. While we find novel ways of using them. And spend our time doing more interesting stuff. Like reading slashdot.
Googlefight "Slashdot Troll" against "BSD is dying" 303:229. BSD thus cant die.
Such systems would have to be built to inherently limit the ammount of actual human interaction. But if that could be done, and each robot could be kept at a cost of, say a modern luxury automobile, then even with replacements, maintenence and repairs, then it wouldn't be inconceivable for one "manager" to be the only human at a popular urban resturant.
The problem would be that said resturant would act like a giant vending machine, with a hole for money, and a hole food appears in, and you have to find a (busy) manager if something goes wrong. This is definetly fine for McDonalds-style food distribution, but not a place you'd take business clients, relatives, or dates to. It's a niche, though a popular one.
On the subject of McDonalds, I've tried the new automated ordering kiosks. They work well. They do not reduce the need for human labor, they increase it slightly - someone still has to make the food, put it together on a tray, and even find the correct customer to give it to, then exchange money. Then there has to be another employee ready to help people with the kiosk itself. The kiosk is merely a tool to keep lines shorter, and people happier. It works rather well that way, and since labor is cheap, it ends up efficient for McDonalds even though it requires more people on average to run it. But that's just my observation.
Ryan Fenton
Not just Moore's law, it also assumes that we can actually program these robots.
It takes at least 5 years for a human brain to be "programmed" to do most simple, coordinated tasks. To program a robot would take at least as long. First we have to develop "learning" algorithms" so that the robot can be taught to receive instruction. A neural net might work, but then it
would have be trained on examples for, maybe, several years. Or we could write very long procedural/OO/functional (take your pick) programs to handle every possible contingency the robot might encounter.
My prediction is we will never simulate the human brain because
"what if it is more complicated that it is smart?"
(I cannot remember where I first heard this, but it is by a neuroscientist in England, I think)
A very well attended symposium was held at Stanford in 1999 that covered this very topic (in even more optimistic depth, in the case of the majority of the speakers). Entitled, Will Spiritual Robots Replace Humanity By 2100?, the symposium was organized by Doug Hofstadter and was themed around two books that expoused very similar views and were written independently of each other around that time: Ray Kurzweil's The Age of Spiritual Machines and Hans Moravec's Robot.
Kurzweil has actually been preaching about this for quite a while now, and the details of Marshall Brain's article are eerily reminiscent of both of the above mentioned books.
How can anyone talk about robots taking over the economy without mentioning Hans Moravec? After all, he's only been doing work in robotic vision and navigaton for the past thirty years or so, and has been on record predicting human-equivalent intelligent machines by 2050 since the mid-1980's.
He's even got a start-up company that wants to manufacture control heads - basketball-sized sensor+computer units that could be used to run forklifts in warehouses.
My personal prediction is that within ten years, we'll see the first automated tractor-trailer truck. It'll have a Moravec-like brain that will run the truck for the 95% of the time the truck is rolling cross-country, and a satellite link for a driver to help direct it for the last 5%.
To a Lisp hacker, XML is S-expressions in drag.
What he wants is to get people to think NOW about this before it happens. The government needs to have a plan to transition from our current mostly capitalist system to a more, well, socialistic one. Why? With machines able to do all the work, you won't be able to sell labor any more. Instead government will be able to give people about whatever they want. As was said in the article, it could mean permanent vacation for everyone with everyone getting at least the basics of survival if not a bit more and will live somewhat comfortably. However the problem is that we are not aiming that way right now. When the robots do all the work, we will have tens of millions of angry and repressed people who will have nothing to do but breed like rabbit, commit crimes and start riots. It will be ugly when that happens too because not all the robots or money in the world will stop several billion people from revolting in anger if they are left to suffer while the rich live off of the robots doing what was once their jobs. Several billion you say? Well obviously this isn't just going to be a problem in the US. This will start to happen worldwide as well and places like China and India had better enjoy our jobs they are getting from us while they can because they will be next if and when robots arrive.
--Won't that be grand? Computers and the programs will start thinking and the people will stop. - Dr. Walter Gibbs
The cars weren't crap because of the workers. The cars were crap because of the people in charge. The guy who practically invented quality control (Deming) went to Detroit first, to the heads of American car companies. They laughed at him. So he went to Japan (where's he's a hero.)
The average worker on the line didn't have anything to do with that decision. If management had decided to implement quality control they would have gone along with it. The CEOs of the big three automakers were asleep at the switch. It was their screw up that cost the US all those jobs. Deming practically begged them to implement quality control, he was an American, and he wanted American companies to use it. It's one of the big ironies of the whole thing that the resurgence of Japanese manufacturing is largely due to an American. And most Americans have never even heard of him.
When I was much younger there was a Taco Bell in my hometown that had a similiar system, I guess as a test/trial. You would go to the display, touch the screen and it would give you all your options, also asking (try to audiolize this) in a contralto "Would you like to add:" then in an enthusiatic baritone "Sour Cream?!"
They were fun to use, if you liked computers, but most people bypassed them and went to the counter, I guess because they would rather blame the employee for screwing up their order than themselves. Taco Bell had the system for a year or two, then removed it, but if you go there today you can still see the holes in the floor where they were. I miss them every time I get a messed up seven layer burrito...
What I would like to know is why no one has set up an IR or wireless interface so people could arrange orders on their PDA and then just send them when they drive-thru. IANAP, but it would seem to me that if you could download fast food restaurant menus to your PDA, decide what you want before you get there, send the order to the restaurant's network through the IR port on the drive-thru sign or wirelessly, you could speed up the ordering process and reduce the errors of communication that crop up as you try to yell your desires into the microphone.
Maybe McD's can do this with their WiFi service.
Ive seen spooky video of Industrial-Robot manufacturing facilities in Japan with THOUSANDS of robots assembling their own.
eerie.
It doesn't say, "hummmm...this pilot doesn't know what he is doing, I'll
put the engine at 0 instead."
Ahh, yes it does; at least in the case of Airbus, it will refuse to execute certain combinations of settings, that indicate to the computer that "this pilot doesn't know what he is doing".
Right now I work at a local Safeway, and at least half of the people who work there after 3pm go to my school. And right now we are going through labor negotiations (read: strike on Wednesday) with a new CBA that hopes to prevent the company from testing out new automated checkout systems here in Canada (BC)
No, don't look at history. We are quickly approaching a point in time where the old rules really will stop applying. This isn't the same as any of the other fancy tools we've come up with in the last few million years.
i d= 6445724
Remember the technology growth curves which point to a singularity sometime in the next 50 years. Its driven by positive feedback. We're going to see massive changes coming at a pace with which we can't keep up. The system is entering a non-linear region and its anybody's guess how it will play out, but it is definitely not the same as factory robots or the printing press.
Sure, there will probably be a need for human creativity and oversight for the next 50 years, but like other posters have said you will start to see the uneducated and unskilled increasingly out of opportunities for employment. How will we handle it?
An interesting counterpoint is China, where labor is so cheap that humans replace machines. Teams of people populate circuit boards where the same job might be automated in Malaysia. I think this is a temporary phenomenon, though, and once we see the next generation of machines the price point will shift and never come back.
http://ask.slashdot.org/comments.pl?sid=71191&c
http://www.masturbateforpeace.com/
It's a problem of numbers, though. Yes, you'll always need a few people to monitor, upgrade, repair, etc, but you'll never need anywhere near as many. So a huge factory full of employees gets reduced to maybe five guys taking turns with the pager, getting paged to go in if anything goes wrong. You'll never eliminate people entirely, but you'll come way too close enough. The heart of the problem is that although there are tasks that can never be done as well by machines, there are also people who cannot do anything (profitable) cheaper/faster/better than machines. Right now the uneducated or just plain not-too-bright can work in McDonald's, but what happens when anything easy/simple enough for them to handle can be done cheaper by an automated system? Unemployment rises not when there are no jobs at all for humans, but when there are fewer jobs than there are people seeking them.
Visit me on #weirdness on the Galaxynet.
but it's been completely human-run for the last two years. The touch screens are still on but they have no signal coming in; they just flicker and give you a headache as you deliver your order verbally. It's kind of sad.
That Arby's has never done too well, though, so I'm not sure its reversion to traditional methods is reflective of the technology. (Roast beef sandwiches, in yuppie California? Not exactly a recipe for success!)
To all Slashdotters who live in the SF Bay Area, this can already be seen, and has been seen by probably thousands of people by now. It's been a failure!
At the Arby's fast food restaurant in San Jose, on Stevens Creek Blvd. (just west of Valley Fair), you can see the remains of a prototype automated ordering system. This must have been a prototype, because I've never seen it (or even heard of it) at any other Arby's.
It runs on IBM CGA displays, in pure text mode (80x25, colored). It uses touchscreens. It looks to have been installed around 1985 or so (I remember seeing a copyright notice somewhere that said that).
The idea was that you would touch items that you wanted to order. It worked fairly well. There's lots of combinations of various screens to press, but not so many that it would be confusing. At the end of your order, you could see the total amount of money you had to pay.
Then, the human interaction comes in. The touchscreen displays are on a countertop, angled towards the customer, but over the countertop are conventional fast-food ordering cash registers. After getting to the final screen, you just kind of awkwardly stood around, and a clerk would come over and eventually take your money. Then you get an order number and wait for your food, as always.
It seems strange to have this hybrid system. If a person is going to end up confirming your order and taking your money anyway, the computer doesn't save much time at all, or really make it any easier. Some people were confused by the computers. Getting a custom order, such as getting lettuce and tomato put on a Big Montana (which disappointingly comes bare by default), was impossible using just the computer systems. Many people simply ignored the computers and gave their orders directly to the clerk! They didn't mind this at all, and in fact preferred it over having to go through the computer.
The system is somewhat in ruins now. After 18 years, many of the screens have worn out, and in fact are turned off. Sometimes they flash odd colors. The last I remember the system fully working was over a year ago. Needless to say, all order taking at this Arby's has been returned to being done in the old fashioned way!
So attempts to automate fast food are nothing new... maybe someone older than me can post about the "Automat" systems of the 1960's?
Dr. Demento On The 'Net!