Going Back to the Source

The first article in the "Stories from the Trenches" series included the observation that "Successful biomimetic innovation often benefits from repeatedly returning to the biological inspiration to validate assumptions and resolve challenges."  I received one comment that questioned this statement - as a biom* project develops and needs to integrate into the human domain, it may be more valuable to be open to a wider range of potential solutions from technology, basic engineering or physics.  

As always, the context is important.  "Going back to the source" can prove valuable if we lack a good understanding of the biological principles, or if the innovation process runs into into challenges that are not easily solved using current solutions.  An example is the Technology Review post Why even a moth’s brain is smarter than an AI.  We often assume that neural networks are inspired by our brains.  We have come a long way since the early attempts at artificial intelligence, but in many ways, our attempts at emulating nature fall far short of the mark.  

​The study described in the post looked at one aspect of learning - the speed at which moths can identify odors compared to the extensive training sets required by most of our artificial neural networks.  The research had two benefits - it increased our understanding of moth brains, and led to a new way of designing neural networks.  The key was to ask a deceptively simple question: how well do our attempts to emulate nature compare to the organism or phenomenon we are emulating?

Growing Up with Lucy: How to Build an Android in Twenty Easy Steps by Steve Grand has a similar message.  Grand argues that much of what we think we know about our brains is colored by our technology.  Trying to build an android using the best knowledge available to us, and then comparing how well it works compared to our (or a moth's) brain, can advance our understanding of biology and bootstrap our ability to emulate it.

0
Your rating: None
None
Login or register to tag items