October 24, 2002
Where Are the Robots?

While the rate of advance of computer and communications technologies have vastly outstripped anything I imagined when I was a child, in one crucial and key field we still lag behind what I thought then would be commonplace now: robots.

Where are all the robots that I expected to see populating the twenty-first century? Why are factories still filled with assembly-line workers, rather than robots? Why is it that the tasks that we don't think of as involving a huge cognitive load--figuring out which bolt will fit into which nut and how firmly to grasp something--in fact turn out to be computationally difficult beyond all measure?

Nevertheless, the industrial robots are coming--very slowly, but they are coming...


Economist.com: :Robots: From The Economist print edition


The number of new industrial robots installed in Japan dropped from 47,000 in 2000 to 28,300 last year. This dragged the global total down from 98,900 to 78,100. Although Japan's stock of robots also shrank, from 389,000 to 361,000, it is still almost half the world total. The United Nations World Robotics report estimates that global robot installations will rise by an average annual rate of 7.5% over the next four years.

Posted by DeLong at October 24, 2002 01:52 PM | Trackback

Email this entry
Email a link to this entry to:


Your email address:


Message (optional):


Comments

The robots are retarded (in time and capability) because we tried to build them as rational (logical) entities. The statistical approaches now being used in robotics work pretty well (in vision, navigation, speech recognition, etc.) but are still not nearly as developed as the logical tradition. However development is snowballing nicely now that most of the resistance from the logical tradition has been overcome.

There's a curious parallel with economics, where the neo-classical synthesis depends on a very fragile rational/ logical process that produces unrealistic equilibira, while a statistical process can produce much more robust equilibria with much weaker assumptions.

Posted by: Jed Harris on October 24, 2002 02:42 PM

I know services are where it's at -- but looking at Germany and Japan's installed capacity of robots one must have a little less pessimism about their economic futures.

Posted by: MJ Turner on October 24, 2002 03:04 PM

Industrial robots haven't caught on because people fear giant robot attacks. Fortunately, Old Glory Insurance has made considerable headway in providing affordable insurance to the elderly in case of robot attacks. See for yourself.

Warning: Persons denying the existence of robots may be robots themselves.

Posted by: FMguru on October 24, 2002 04:55 PM

I disagree with the premise "While the rate of advance of computer and communications technologies have vastly outstripped anything I imagined as a child".

We haven't even got workable voice control, let alone a computer comparable to HAL in 2001. As regards communications, video telephony (a staple of imagined futures since Dick Tracy) is still off in the distance somewhere.

And while the Internet is great, personal computers in other respects haven't made more than incremental advances since the 1980s (or, in the case of Windows, since 1995).

Paul Krugman did a piece a while back comparing the predictions of people like the Hudson Institute. He found that, like economists with recessions, they forecasted all the main innovations we actually got and a bunch that we didn't.

Posted by: John Quiggin on October 24, 2002 05:00 PM

This is my field, so I think I'll throw in my two cents' worth.

First, the distinction of what a robot is in most of these surveys (an automated manipulator with a -- theoretical at least -- reconfigurability for different tasks), while it may have some technical merit, is a useless and misleading one for business/economic analysis reasons. The real action over the last 20 years has been in more dedicated-purpose machines, which share most of the underlying technologies, but do not get counted in these surveys (and are not weighted nearly so much toward Japan).

As an example, take the very common "chip shooting" machines, which are used for automated installation of electronic components onto circuit boards. (They are so fast these days that they might just as well be literally shooting the chips out. The designer of one of these machines recently told me that every millisecond he can shave off the pick-and-place cycle time for a component adds $1000 to the market value of the machine. The majority of their basic technologies are the same as "robots", their economic impact is the same, they may outnumber robotic arms in the world, but they don't get counted.

These types of machines are reprogrammable to the extent of being able to handle different types of components and board locations, but not to the extent that they could be turned into, say, welding machines.

It turns out that the dream of the truly (software-) reconfigurable manipulator is still largely a dream. Even most robots are purchased configured for a single type of use, such as welding, even if the robot manufacturer has the capability of configuring them for other uses as well.

Why is this? Well, as Brad's original post suggested, using computers for interaction with the messy real physical world is a far, far, more difficult task than for purely abstract symbol manipulation. Even what we consider "unskilled labor" requires the use of a lot of the kind of judgment that computers are really bad at, and that 50 years of AI research have not made much of a dent in.

This is why robots see virtually no use outside of the rigorously structured environment of the factory floor. Being able to handle the case where a bolt is presented to the robot backwards is still a difficult problem.

Why do the Japanese dominate in the use of robots? They are the best in the world at the drudge work of process refinement, which requires a detailed understanding of every minute element of the process. If you try to automate a process without fully understanding it, you will produce junk. (This killed off the first wave of enthusiasm for robots in the US.)

Second, investment capital was basically free in Japan for a long time (which I think was the key factor that almost destroyed the Japanese economy), which skewed the economic decision in favor of automating. What killed off the second wave of enthusiasm for robots in the U.S. was the realization that the required preliminary step of fully understanding your manufacturing process solved by itself most of your problems, so it wasn't then worth spending expensive capital on the robots.

Posted by: Curt Wilson on October 24, 2002 05:26 PM

To put it another way logical processes are easy to automate because we have to think about it - it's unnatural to think logically, so those few who can do it on occassion pretty much know how they do it. Since we know how we do it, step by step, we can tell the computer how to do it.

Motor and spatial skill, on the other hand, are largely unconcious. We don't really know how we do such tasks and therefore they are very hard to program.

Posted by: Ian Welsh on October 24, 2002 06:58 PM

Curt said pretty much what I was going to say (but better). Essentially, we HAVE robots, all over the place, they just aren't particularly anthropomorphic.

A microwave with a revolving platter and timer may not be the Jetson's Rosie, but it helps you around the kitchen.

Posted by: Dave Romm on October 24, 2002 07:21 PM

great post, Curt. Some things that have been swirling around in my thoughts just a bit out of focus are now clearer, thanks to you.

Posted by: kit on October 24, 2002 07:52 PM

Curt's post is pretty much on the mark, but misses a couple of things. Japanese robotics has, in general, been much more market driven than US robotic work. In the US the focus has been on displacing unionized skilled labor. In Japan it has been on increasing productivity. If you compare former GM chairman's Smith's agenda for robots to the great Taishi Ohno's comments on automation, you will see the difference between a finance and "management theory" based approach to technology and one rooted in the actual workings of production.

Posted by: yodaiken on October 25, 2002 07:20 AM

Is it also possible that assembly-line workers in Guatemala or Indonesia are cheaper than the maintenance of robots here in the US? That seems likely to explain a lot from my POV.

Posted by: Chris Quinones on October 26, 2002 03:18 AM

Chris writes: "Is it also possible that assembly-line workers in Guatemala or Indonesia are cheaper than the maintenance of robots here in the US? That seems likely to explain a lot from my POV."

Price isn't the only consideration - robots also don't talk back, work 24/7 without much complaint, and so on.

Posted by: Ian Welsh on October 26, 2002 10:14 AM

True, but given the overproduction we're experiencing, I suspect continual ability to produce is not such a large consideration right now, whereas bottom-line issues are of paramount importance. And I wonder how much the cheap labor overseas sees fit to talk back?

Posted by: Chris Quinones on October 26, 2002 12:07 PM

And I suspect we're spiralling into a demand problem as well, but that's another matter. Henry Ford famously paid his employees enough to buy his cars. Modern capitalists have forgotten that simple wisdom.

Posted by: Ian Welsh on October 28, 2002 08:05 AM

Ian:

I think you've bought into an old myth regarding Henry Ford. This is the same Henry Ford who in a few years was hiring goon squads to stop labor organizers.

In reality, Ford found that even when he paid twice the prevailing wage for labor of this skill level, he generally could keep people on the line for only a few months, because they hated the assembly-line work so much. So he dramatically doubled the wages in one fell swoop and found that at four times the prevailing wage, he could keep people. He could afford this because the productivity of the assembly line was so high, but other justifications for the move were just PR.

I've long thought this one of the most pernicious myths of economic history. It just can't make economic sense when you are employing only a small fraction of the potential market to raise the cost to the rest of the potential market.

Loved your comment on automating logical vs. motor processes, though. Mind if I borrow it in the future?

Posted by: Curt Wilson on October 28, 2002 02:49 PM

Curt. Fair enough. Thank you for correcting me. Nonetheless while it's not true on a micro scale (ie. individual businesses should keep their costs down) it's the sort of behaviour which if everyone engages in it can cause problems.

By all means feel free to use my comment on logical vs. motor processes.

Another interesting little factoid - even most majors in physics instinctively think in terms of Aristotlean physics. I mean, it just makes sense that something keeps going until it runs out of energy, right? :)

Brad inveighles about how few people seem to think properly about problems - but the reason is that it is deceptively hard and our brains, in fact, are set up not to do so in most situations. Thinking logically is profoundly unnatural in most circumstances.

"Descartes Error" by by Antonio R. Damasio is a good book on this subject (though he says little about motor skills).

Cheers,
Ian

Posted by: Ian Welsh on October 28, 2002 05:56 PM

Curt. Fair enough. Thank you for correcting me. Nonetheless while it's not true on a micro scale (ie. individual businesses should keep their costs down) it's the sort of behaviour which if everyone engages in it can cause problems.

By all means feel free to use my comment on logical vs. motor processes.

Another interesting little factoid - even most majors in physics instinctively think in terms of Aristotlean physics. I mean, it just makes sense that something keeps going until it runs out of energy, right? :)

Brad inveighles about how few people seem to think properly about problems - but the reason is that it is deceptively hard and our brains, in fact, are set up not to do so in most situations. Thinking logically is profoundly unnatural in most circumstances.

"Descarte's Error" by by Antonio R. Damasio is a good book on this subject (though he says little about motor skills).

Cheers,
Ian

Posted by: Ian Welsh on October 28, 2002 05:56 PM
Post a comment
Name:


Email Address:


URL:


Comments:


Remember info?