Hmm… not music related, just a book I recently finished.
It’ll be geeky so be warned. It’s a book about brains mostly, introducing (or arguing) a general framework on how our brains work. If you are interested in Artificial Intelligence, or just how our brains work this is the book to read, it’s not very technical, and it’s easy to read, you’re not required to have any pre-knowledge of neuroscience or A.I.
I’m always fascinated with all the science fiction intelligent robot. Not so much on how brains work, but I just find it amazing that a human made object could behave like a human. When I first took Artificial Intelligence in my third year of University, I thought it was going to be the best subject ever, well it did turn out fine, I still enjoyed the subject, but I was disappointed when I discovered A.I. is all but a large, exhaustive search . There’s no intelligence at all, not in the A.I. that beat the human chess champion, it didn’t beat a human by playing smarter and understanding the chess game more, it just tried searching harder, more steps ahead.
What defines “Intelligence”? Without going into any philosophical non-sense, the common definition of intelligence is intelligent behaviour. That is if you behave like an intelligent human, then you have intelligence. The book argues that behaviour is only a side effect. If you are intelligent, you can read a book, understand the content, sit there silent but you do know the story, even if you don’t tell anyone. It is because we focus on intelligence being evaluated as some sort of output, we try to build “intelligent” machines that will achieve the goal, and when it does so, it is deemed intelligent. Ie. if a piece of software translates English to Chinese perfectly, it is intelligent, even though at the end of the day it understands no words, no grammar, nothing about English or Chinese.
The book defines intelligence as building a model based on perception, and then predicting or anticipating events using learnt patterns or memories. It describes the brain as a memory-prediction system, where we are constantly learning patterns so we can recall learnt patterns and anticipate future events. It then goes further into explaining how our memory stores pattern through hierarchical invariant representation. Invariant representation means a representation of the learnt patterns in a non-rigid way. For example, you don’t remember a person’s face by remembering every pixel on the face, if you see it in a different angle, different lighting, you’d recognize the face either way. Hierarchy can be described as abstraction, a concept any computer scientists would understand. From electrical spikes to bits to assembly subroutines to procedures to modules to system, through hierarchical abstraction we are able to hide minute details from the higher levels, and so complex things can be built – if we were to program things in bits (0’s or 1’s), we would probably get no further than making a calculator.
Same goes with the brain, through layers of neocortex it argues that the brain uses the same method of hierarchy and abstraction to store memories. It goes on further suggesting how the brain does it with axons, synapses, etc. Not in an extremely gory detailed way that would make a research paper, but good enough to back up his speculation.
All in all, a really thought provoking book and very intellectual stimulating. If you are interested in Brains or Artificial Intelligence or just feeling bored, give it a try.