Sunday, June 19, 2005

Biology is not magic

My poor, neglected blog. Lots of travelling the last two months. During the last trip I actually wrote a couple of prospective blog entries on paper, but all of that may not make it here. For me, blogging is mostly about the inspiration of the moment - not so much about planning.

Anyway, todays theme is: Biology is not magic. I might consider printing a T-Shirt with that slogan: recently I've been quite fed up with people - mostly rational, well-educated people - that have little or no knowledge about biology and biological processes going all "Oooh! Shiny!" and using "this will be done similarly to biological processes" or "this is biomimetic (similar to biology)" as a motivation for how to do something. I'm thinking mostly of industry and technology now, but this might apply in some other areas as well.

(and then of course we have the infamous "in the stone age men were hunters and women took care of kids" which seems to be endlessly useful as a "proof" for whatever loose statement someone fell like making. Bleh. That is almost unprovable due to lack of scientific data, and there are other issues as well)

Really, people. Why is it that so many intellectuals in economy, physics, computer science and other well-established areas seem to totally lack an ablility for logical, critical thinking when it comes to seeking inspiration from biology? Is it some kind of wish-fullfillment; "I secretly wanted to go into biology, but since it pays less/my parents wanted me to do X, I didn't - and look, now I can solve biological problems anyway! Or is it a "I recently became bored with my subject of research/I desperatly need a new research profile to get funding" sort of thing?

You know, the worst case I've seen yet was: "I use a 10-node neural network. Neural networks are brain-inspired, so this is a biology-like analysis, which means it is a good analysis." And this from a senior scientist. Shudder. What about input data, for starters? I can't imagine that explicitly putting non-biological data into a "biological" analysis method would do any good. It's like trying to type your C++ program in chinese - wouldn't be likely to work. And even a rather clueless person should be able to guess that humans have at least 1 million neuron in the brain (the real figure is much much larger) - and if your method is 100,000 times less complex than the thing you compare it to, do you relly think it will work in the same way?

We, as a human species, need to get our minds around the fact that humans are not the epitome of perfection (neither is the rest of creation, for that matter). I suppose this notion comes from the heritage of "created in the image of God", or maybe it is simply cognitive bias. In some cases this leads to pure sexism (the idea that women are a less perfect variety of men has been prominent through much of history), in other cases it leads to racism, or ageism, or ...

Neither is evoultion some kind of engineering process. It is a slow, noise-ridden random optimisation process with a huge amount of dimensions. Sure, if you run your optimisation algorithm many times you may get a good result. But you'll never know if you get stuck in some kind of local minimum along the way - and the higher the number of dimensions, the slower is the process. As long as biology works better than what you have right now - by all mean, use it for inspiration for your technological improvements. But be critical. Think about what your assumptions mean and whether they are valid. Consult a biologist who is an expert on the system you want to mimic. And remember that there may be much better solutions that nature simply hasn't come round to yet.

0 Comments:

Post a Comment

<< Home