Sunday
Sep162012

What "Plausible" might be able to teach a stuffy scientist

So, yea, you probably know my story... I wrote a computer program that generates statistically plausible models even though relevant models were grown out of nonsense.  And I've been working with those models for some time now.  When you build enough models this way - even though they are nonsense, there seems to be very real insight.

But I happened upon something a bit less obvious.  Most of human science happened in much the same way, truths grown out of mysteries.  And plausible models are easy to find in a big multivariate space, even if they are really rare.  Give me an experimental space with 10^300 combinations and one in a trillion chance of finding a model and I can tell you there still are a insanely huge number of plausible models in that space.  

So what I'm saying is that most of the sciences are probably "best known truth" and not "absolute truth".  It occurs to me that there might be trillions of almost perfect ways to put the universe together into a grand unified theory.  

Some might be very similar (say the same idea but expressed differently via both information network theory and say 11 dimensional holographic string theory.)   Other plausible theories might be quite different structurally but give essentially the same results (say a fated deterministic model vs. a choice and free will based model.)  

Clearly humanity isn't there yet with a grand unified theory, gravity doesn't quite seem to fit into the quantum world, dark matter and dark energy appear to be very real things that make up about 95 percent of the universe and both still have foggy edged definitions.  Scientists talk about the quantum world as a sea of crazy discrete possibilities without seeming to consider that probability might well be a dimension just like time or space, and the observed reality is commonality of sorts built out of unions and intersections of entangled parallel probabilities.  

There are physical realities that grow out of pure mathematics - Heisenberg uncertainty for example.  But I wonder about the theories that grow out of data and observation.  Models of perfection probably rest in a candidate pool of giga-billions, among whom even the ugliest roughest pretender might still look nearly perfect.

Be kind to skeptic's, that might serve us well in the long run...

Thursday
Aug232012

Mars Curiosity 360 Link

This link has a high-res beautiful and desolate view of mars that you can pan 360 style.

http://www.360cities.net/image/curiosity-rover-martian-solar-day-2#112.23,25.13,59.0

Its hard for me not to react emotionally to the thought of actually being there... such isolation.  I wonder what sounds the thin martian air would convey from the crunchy rocks as the rover motored along... I imagine it sounds bit like a hike across the pumice pebble pile at sunset crater volcano in Arizona.

 

Tuesday
Jul032012

A Missing Camera on Bald Mountain (813) 469-1396

I found a digital camera that had weathered the elements for a few years near Cashiers NC.  The camera was destroyed, but the memory card photos tell a story of dogs, a skinny dipping couples, a man that climbs the mountain with a six pack of glass bottles and his sun eating girl.   His girl probably left her camera behind under the broken glass bottles by accident.

Maybe someone can figure out who these people are... I'll mail the memory card to the owner if they can describe a memorable picture and I still have the memory card. The photo of the sun-eater was taken in almost exactly the same place I found the camera.  Calling the number on the dog tag brings up the Pirates of the Caribean theme song.  I could pay for a reverse lookup, but if... well... I wasn't that thrilled to hike for a couple hours with small kids and happen across leftover broken glass in an otherwise pristene locale... and I suspect the dog doesn't belong to the camera's owner but rather friends?family.     And the awkwardness of sending off a few moderately unusual photos to Grandmama by accident.       

Friday
Jun012012

Nonsensebox hacks itself

 

I'm still amazed what this odd set of algorithms acomplishes.  Thursday I fed it market data and the predictive model showed an r^2 of 1.00.  "Huh, wow!" I thought, "its good, but predicting the rise and fall of 70 odd stocks perfectly to the cent over the course of a day seems a bit far fetched."

Looking into it, the math functions for the model NB built was nesty & loopy - it looked valid, but was strangely indeterminate in ways that tripped up my code (or less likely the built in fuctions or math kernel).  NB actually found a perfect cheat rather than solved the problem at hand!  Diabolically brilliant - almost like something a human hacker would do.

I added a extra line of code and NB shouldn't be able to do exactly that same trick again.  Back to 2^353 combination space and find me some models!

Wednesday
May162012

Thunderbolts and Lighting?

A rough pattern for a sleeveless quilted shirt for a Galileo costume for Q's school: