Mechanics without wrenches

byPhilGoetz10y15th Apr 200978 comments


Say you're taking your car to an auto mechanic for repairs.  You've been told he's the best mechanic in town.  The mechanic rolls up the steel garage door before driving the car into the garage, and you look inside and notice something funny.  There are no tools.  The garage is bare - just an empty concrete space with four bay doors and three other cars.

You point this out to the mechanic.  He shrugs it off, saying, "This is how I've always worked.  I'm just that good.  You were lucky I had an opening; I'm usually booked."  And you believe him, having seen the parking lot full of cars waiting to be repaired.

You take your car to another mechanic in the same town.  He, too, has no tools in his garage.  You visit all the mechanics in town, and find a few that have some wrenches, and others with a jack or an air compressor, but no one with a full set of tools.

You notice the streets are nearly empty besides your car.  Most of the cars in town seem to be in for repairs.  You talk to the townsfolk, and they tell you how they take their cars from one shop to another, hoping to someday find the mechanic who is brilliant and gifted enough to fix their car.

I sometimes tell people how I believe that governments should not be documents, but semi-autonomous computer programs.  I have a story that I'm not going to tell now, about incorporating inequalities into laws, then incorporating functions into them, then feedback loops, then statistical measures, then learning mechanisms, on up to the point where voters and/or legislatures set only the values that control the system, and the system produces the low-level laws and policy decisions (in a way that balances exploration and exploitation).  (Robin's futarchy in which you "vote on values, bet on beliefs" describes a similar, though less-automated system of government.)

And one reaction - actually, one of the most intelligent reactions - is, "But then... legislators would have to understand something about math."  As if that were a bug, and not a feature.

We have 535 Congressmen in the United States.  Over the past half a year, they've decided how to spend several trillion of our dollars on interventions to vitalize our economy.  But after listening to them for 20 years, I have the feeling that few of them could explain the concepts of opportunity cost, diminishing returns, or the law of supply and demand.  You could probably count on one hand the number who could solve an ordinary differential equation.

This isn't the fault of the congressmen.  This is the fault of the voters.  Why do we regularly elect representatives who are mechanics without wrenches?

We like to praise the man who achieves great things through vision, genius, and force of personality.  If you tell people that he had great tools, people think you're trying to diminish his accomplishments.  People love Einstein above all scientists because they have the idea that he just sat in a chair and conducted thought-experiments.  They like to believe that he did poorly in math at school (he didn't).  Maybe this is because they feel math is a crutch that a true genius wouldn't need.  Maybe it's because they would like to think that they could also come up with general relativity if they just had enough time alone.  They love scientists who say they work by visualization or intuition, and who talk about seeing the solution to a problem in a dream.  It's not evident that Einstein was smarter than John von Neumann or Alan Turing, yet most Americans have never heard their names.

I think that what America needs most, in terms of rationality, is not training in rationalist techniques - although that's of value.  What America needs most is awareness of how much of a difference intelligence and education and rationality can make.  And what America needs second-most is for people to recognize the toolkits of rationality and appreciate their power.

Most people don't realize that there are small bodies of knowledge that radically amplify your intelligence.  Even a general understanding of evolution, or thermodynamics, or information theory, gives you a grasp on all sorts of other topics that would have otherwise remained mysterious.  Understanding how to rephrase a real-world problem as a function maximization problem lets you think quantitatively about something that before you would have had to address with gut feelings.

One reason for this may be that, in the mind of the public, the prototypical smart person is a physicist.  And particle physics, quantum mechanics, and relativity just aren't very useful toolkits.  People hardly ever get an insight into anything in their ordinary lives from quantum mechanics or relativity (and when they do, they're wrong).  You don't have to know that stuff.  And, as 20th-century physics is thought of as the pinnacle of science, it taints all the other sciences with its own narrowness of applicability.

With the exception of math, I can't recall any teacher ever trying to show me that something we were studying was a toolkit applicable beyond the subject being studied.  The way we try to teach our students to think is like the (failed) way we tried to teach AIs to think in the 1970s (and, in Austin, through the present day) - by giving them a lot of specialized knowledge about a lot of different subjects.  This is a time-tested way to spend a lot of time and money without instilling much intelligence into either a computer or a student.  A better approach would be to look for abstractions that can be applied to as many domains as possible, and to have one class for each of these abstractions.


(PS - When I speak specifically about America, it's not because I think the rest of the world is unimportant.  I just don't know as much about the rest of the world.)