Slashdot Mirror
Hydraulic Analog Computer From 1949
mbone writes "In the New York Times, there is an interesting story about a hydraulic analog computer from 1949 used to model the feedback loops in the economy. According to the article, 'copies of the 'Moniac,' as it became known in the United States, were built and sold to Harvard, Cambridge, Oxford, Ford Motor Company and the Central Bank of Guatemala, among others.' There is a cool video of the computer in operation at Cambridge University. I remember that the Instrumentation Lab at MIT still had an analog computer in its computer center in the mid-1970s. Even then, it seemed archaic, and now this form of computation is largely forgotten. With 14 machines built, it must have been one of the more successful analog computers — a supercomputer of its day. Of course, you have to wonder if it could have been used to predict our current economic difficulties."
No, you don't have to wonder that. The current economic difficulties were easily predicted by many.
The problem was that the people with any kind of ability to stop the conditions that led to the current situation were those who profited most from those conditions. Not a good recipe for prevention.
"Trolls they were, but filled with the evil will of their master: a fell race..." -- J.R.R. Tolkien on Olog-hai
If lots of people are extracting money from the system and not contributing real wealth, then there will be problems. Money has complicated dynamics, but its not magic. People who make a living shuffling numbers around in spreadsheets are providing a useful service that makes the system more efficient. But only up to a point. Few people believe greed is a vice anymore, hence certain results follow.
There is a serious flaw in thinking that computers can accurately model macroeconomics, or predict systematic collapses, any better than commonsense and basic logic can
It might not be an accurate descriptive model (one that describes reality), but discriminative models are very useful -- not to predict systematic collapses, but to discriminate irrational behavior. Computers step in when your response time has to beat efficient markets. You could be a common sense and logic genius, but you can only apply those rules and react in so much time, and computers will always beat you for speed at applying your own rules. So it's a different kind of crutch, and just like a crutch, it can help you walk better, not walk for you.
I might have digressed from TFA, but this is /.
An old-timer with old-timey ideas.
I think it's not so much that it can't in theory be modeled, as that in practice it's extremely hard---much harder than modeling the weather, which we can barely do accurately out to 5 days. There's somewhat of a gap in expectations when you go from high-level qualitative descriptions of phenomena, like hurricane experts discussing trends in intensity and formation basins, to an implemented, detailed computational model that purports to simulate what "really" happens. The 2nd one usually diverges extremely quickly from what actually does happen, because these sorts of nonlinear systems are very sensitive to slight errors in the model or its initial conditions.
That's fine, as long as people understand its use and limitations; but there's a tendency, especially among the only-sort-of-technical folks who are involved in a lot of areas of business and economic policy, to trust these computer models as more than they are, as if the fact that a "computer simulation" told you it makes it some sort of neutral third-party truth that represents the state-of-the-art in guessing what will happen.
10 PRINT CHR$(205.5+RND(1)); : GOTO 10
It doesn't take a computer to predict what happens. You get a boom, people spend money, they overspend, you get a crash. Repeat. The details vary, but the basic pattern seems pretty clear.
Indeed, when it comes to anything safety crtical KISS is a good principle to follow as it will make it much easier to ensure safety.
note: i'm known as plugwash most places but i screwd up registering that here somehow in the past and now can't register
All you need is the Micawber Principle:
"Annual income twenty pounds, annual expenditure nineteen pounds nineteen and six, result happiness. Annual income twenty pounds, annual expenditure twenty pounds ought and six, result misery."
From David Copperfield, by Charles Dickens
Prove anything by multiplying Huge Number times Tiny Number
As we say in Economics circles: "Yes, the man is a genius: he predicted 9 of the last 3 recessions!"
FATMOUSE + YOU = FATMOUSE
It's worse than that, actually. Weather prediction is at least based on physical processes, which, allowing for some minor concessions to quantum mechanics, are based on fundamental particles which follow deterministic rules. Given all the initial conditions and sufficient computing power you could accurately simulate what the weather will be at any point in the future. Moreover, you can use simulations to predict how that future weather will change in response to deliberate artificial influences; weather doesn't have goals of its own, and won't actively resist attempts to control it.
Economics isn't like that. The "fundamental particle" of economics is people, and people are adaptive. Under many conditions is it possible to predict how they will behave--assuming rational self-interest (i.e. sanity) and decent psychological models of their personal value scales--but all that breaks down when someone attempts to use the models to control the outcome. At that point you have a competition between the people being studied, who seek to achieve their original goals and thwart any attempt at outside control, and those seeking to do the controlling. As with any long-running competition between creative individuals, the outcome is impossible to predict with any degree of accuracy.
"The state is that great fiction by which everyone tries to live at the expense of everyone else." - Bastiat
Often, digital computing is clumsy, awkward, and overly precise.
A good example of this is in Aviation. Pilots are still trained to use the E6B Flight computer which is really a glorified slide rule.
When you are estimating fuel consumption, and figure you'll probably use about 11.5 GPH for 3 1/3 hours, it's not important to know any more accuracy than perhaps to a gallon or so, since reality will always be a bit different than your calculations, anyway. You don't need to be exactly precise on your degrees of heading, and when you are computing weight & balance, it's stupid to calculate your moment to the 4th decimal place.
Knowing how to use an E6B, you can get calculations in a second or two with a single hand that are "good enough" - and that's important when you're flying an airplane in turbulence while trying to stay on top of busy ATC calls in a heavily trafficked area. You can't even enter the first of 3 or 4 digits into a digital calculator in that time, and you'd have to use two hands. (Frequently, when flying, your hands are both occupied and you're steering with just your feet)
Digital isn't always better.
I have no problem with your religion until you decide it's reason to deprive others of the truth.