Umair Haque / Bubblegeneration
umair haque  

 
 


Design principles for 21st century companies, markets, and economies. Foreword by Gary Hamel. Coming January 4th. Pre-order at Amazon.


 
Tuesday, February 01, 2005


Alife

A very nice article about Alife and Avida, which is a kind of Tierra successor. Highly, highly recommended.

And if you're feeling truly geeky, read this paper about norgev, which is evolves simple neural networks based on a very interesting biosimulation (genome/protein/neuron/environmental interaction + regulation) of, essentially, the brain. I think norgev has massive potential and is one the coolest sims...uhh...ever:

"...Within Norgev, we have shown that such structural biocomplexity can arise in silico, with dendritic connection patterns surprisingly similar to the seemingly random patterns seen in C. elegans."

(via Smartmobs - Dave, you should really check out norgev).

-- umair // 1:57 PM // 1 comments


Comments:

Thanks for the note. That's insanely cool.

The problem has always been design, and interestingly enough I think it was based on a subconscious bias that believed the brain is actually a 'designed system'. With neural networks we've been trying to emulate at too high a level, and thus do too much work. A neural network should be designed at the most basic interaction level, with elements for Hebbian fortification or loss, feedback, and feedforward. In fact, no attempt should be made to make a neural network do anything other than respond to random (and at times repeated) stimuli. IF you artificially accelerate this 'evolutionary process', you might just spawn the first functioning silicon brain. Or you might just get a pile of junk, who knows.
// Blogger dhd // 10:39 PM
 
 

Recent Tweets







    input
    portfolio
    contact

    mail.
    uhaque (dot) mba2003 (at) london (dot) edu

    skype.
    umair.haque

    atom feed

    technorati profile

    blog archives