Bob McGinley::J2Evolution

Evolutionary algorithms now surpass human designers

I read this news article a while back, and I decided to save it till I had some more time to enjoy it.

Evolutionary algorithms now surpass human designers

Instead of pitting God against science, the emerging spat centres on evolutionary algorithms (EAs), which mimic the processes of natural selection and random mutation by "breeding", selecting and re-breeding possible designs to produce the fittest ones.

snip

Proponents of EAs say they could replace traditional methods in many fields from designing exotic new types of optical fibre and USB memory sticks to more aesthetic computer-generated art. Critics argue that the technique may lead to designs that can't be properly evaluated since no human understands which trade-offs were made and therefore where failure is likely.

A couple points about the article.  First off, I don't understand the constant reference to 'God' when discussing evolutionary evolution.  Evolution exists, just like gravity, just like air (ether).  You can argue about where it all started till your blue in the face.  It started, and it exists.  Get over it. 

Simulation and Bridges

As someone who studies and obsesses over traffic and roadways, I was really struck by the bridge collapse in Minnesota.  The bridge failed in a big way. 

We can see in this video that the bridge fails in the section over the river first.  Then the sections further away collapse, seemingly in a chain reaction.  I couldn't help but wonder why the sections weren't isolated, so that failure in one section would not collapse the others.  I'm not a bridge engineer or anything, it was just my first thought.  If I was a bridge engineer, I would design the bridge so that sections could fail independently, like the Bay Bridge did in the San Francisco earthquake.  Notice, the whole bridge didn't fall into the bay.

Then I thought about how simulations could have possibly assisted in detecting the structural failure.  Simulating structural fatigue on a scale like this is difficult to pinpoint accurately.  There are just too many variables to model to predict this kind of failure.  The only thing todays simulations can do is give a wide probability.  Then I read this:

Could Tiny Sensors Detect Bridge Crisis?

Hey, now I have some accurate, real-time date to work with.  This could be interesting.  What happens before a bridge collapses?  What are the signs?  Could we have known about the Minnesota bridge a week before it occurred?  Should every bridge have sensors?