I wanted to talk about the word "recursive." It's a word that's come up a lot, and because it sounds a little like a Times Table, I have glossed over it once too often. The trouble is when "they" talk about neural networks of the various gradations of complexity, from simple to complex, in the evolution of neural networks, the most complex allow for "recursion." This means that to wax lyrical and with tremendous elan on the role communication has played on the Genesis of Species I have to know what recursion means when applied to cognitive function, so I don't stick my foot down my throat.
To avoid an error of comprehension in the definitions here, my own thoughts return to the Post Structural European thinkers who began to understand the words in language as Jacques Lacan's irascible Symbolic Form notable for its fluidity that was better understood as an order of meaning rather than anything that had the nerve to claim a certainty of meaning. What did we use to say in those happier days?
We found faith in Lacan. For him language is a vast network of words, sounds, structures that shape our reality, forms the unconscious, gives structure to the world around us, it's where meaning arises, and of course in this tidal pool of interactions between something and emptiness words happen and become lodged in the memory.
Neural Networks are pathways in the brain for processing information. There's a backwards and forwards, up and down. The word recursive in a Recursive Neural Network, whether the network is artificial as in the big computer or in part of a body that belongs to you or I, puts an equal weight on the backwards and forwards, upwards and downwards of bits of information that's traveling around the network and as a result the network builds hierarchical tree like structures like sentences, not vague half backed whiffs that go around and round just thoroughly enjoying a symbol-less expression of itself for which, aggravatingly and possibly neurotically, there is no explanation.
In our minds, recursion is something that makes sense or begins to make enough sense to pass as sense. In the new world of automation, failure of "recursion" as it currently stands means insufficient data to achieve a solution. And the thing about our Recurring Neural Networks, as opposed to those other ones, as long as we are alive our constant symbolizing of the real never actually comes to a comfortable end. Instead of ending to it all as quickly as possible and switching ourselves off, we come up with things like interpretive dance and when that's not available we call forth words like curiosity, creativity, special, passionate and onwards.
No comments:
Post a Comment