[ExI] Mass transit

[Anders Sandberg]

On 2011-01-08 03:32, spike wrote:
> This is an example of what I mentioned a few days ago about being a way
> bigger threat to society than is global warming, this bigger threat is
> feral humans
...
> It is a reason why I think most public transit notions are a dead end.
> Individual cars serve as suits of armor, providing a defensive barrier.

They also serve as propelled weapons, very good at hurting squishy 
humans. Of course, sometimes it nice to have defensive barriers against 
other mammals:
http://hn.se/nyheter/halland/1.1078155-skogens-drottning-tog-ett-skutt?articleRenderMode=image&image=bigTop 

(as Neatorama put it, an elk jumping over a Volvo, that is Sweden in a 
nutshell)

The problem is feral humans, not mass transit. After all, there are 
ferals in supermarkets, schools and offices too. If ferals are such a 
problem that they impair the utility of mass transit, then we should 
probably consider finding ways of preventing too large uncontrolled 
groups in buildings too.


> In the long run I am thinking something like the old tech ski lift
> technology is the way to move proles in the big city: moving cable, with
> the option of riding alone in an individual car or with as many as four
> riders per car. I don’t know the exact mechanism, but it should be at
> least possible to create a car that individually transfers from one
> moving cable to another, so that one need only enter the coordinates of
> the city block where one wants to end, and the rest is mechanized. This
> does away with most parking lots, or rather moves them out where there
> is plenty of room and the theoretical possibility of making the parking
> lots safe.

It could also be a light rail or automated wheeled vehicle 
infrastructure, perhaps underground. Which reminds me of a question I 
have been thinking about for a while: how much cheaper could tunneling 
become if we had either decent robotics, or decent nanotechnology? Once 
tunneling becomes cheap, a lot of systems can be buried and we can 
retrofit many current cities in cool ways.

There are interesting issues of economies of scale for mass transit. 
Depending on how the cost function looks you get completely different 
topologies. If it is expensive to make, then you get a tree structure 
(think many subways). If it is really cheap you can make it a grid 
(think roads), which in the limit approaches arbitrary point-to-point 
connectivity (think Internet).

Population density acts as local scaling for the density of branches; 
one can probably do a conformal map based on it to the plane, do a 
layout and then transform back for a first approximation of the ideal 
grid. In fact, once can likely weigh in a lot of cost functions into 
this mapping.

(The big headache is always funding - plenty of free rider problems, 
public goods problems, principal agent problems and NIMBY. But the math 
is neat!)


> It could be that I am overly focused on the feral humans thing. Damien’s
> book Transcension has Dr. Malik being slain by ferals in a most
> memorable and disturbing passage. Fortunately he was frozen and we
> eventually saved by that technology. I propose avoiding all possible
> contact with them.

Typically, bad people are a small fraction p of all people (and I think 
Pinker is right that it has decreased historically). But when you get a 
large group together the probability of having at least one grows as 
1-(1-p)^N, approaching 1 as N is large (for small N it is ~pN). On the 
other hand, if the bad person does something bad against one other 
person in the group, the chance of you being hit is 1/N. So the total 
risk to you behaves approximately as Np/N=p - the risk is constant 
whether you are in an elevator with a stranger, or at a mass rally.

Some might worry about ferals doing attacks on larger groups, and see 
that as an argument for avoiding mass rallys (beside the other reasons). 
According to what I have read on the statistics of terrorism, we should 
expect the probability of X people being hurt to scale as CX^-a where a 
is ~1.7 in developed countries (http://arxiv.org/abs/physics?0502014). C 
is of course very very small. Other forms of violence or mass accidents 
are likely follows the same curve (as observed by Richardson). So the 
approximate risk of being hurt in a size X attack is pNX/N= pX, and the 
probability of that is pCX^(1-a).

The mean for X is 22.66 (in fact, only 37% of terrorist attack injure or 
kill anybody) and the incidence is around 1.47 attacks per day anywhere 
in the world, so the empirical risk per day of being affected is 
4.76*10-9. Tails matter here, since it is a very skew distribution - the 
small probability of very large attacks (where the above approximation 
is no longer valid) make the risk weakly dependent on N if one does a 
proper Bayesian calculation (I think), so there might be a mild reason 
to avoid very big crowds.

However, the tiny value of pC makes this a very minor worry. I would be 
more worried about catching some illness from the crowd or getting hurt 
in traffic getting to it.

-- 
Anders Sandberg,
Future of Humanity Institute
Philosophy Faculty of Oxford University