Computers and brains get closer together: Brain-like computing components
The brain is like a computer; this is the fundamental metaphor at the heart of 1980s cognitive psychology. To an extent this was a useful way of thinking about the brain, it certainly stores and processes information just like a computer, and you can even (perhaps) draw some rough parallels between parts of the brain and computer components.
However, in at least one important respect, the brain appears to function very differently from a computer. A computers’ processing power is highly centralised in a single processor (or perhaps a dual/quad core processor – doesn’t matter – still centralised). The processor does all the computational work, and the hard disk stores all the data that the processor works on. This means that data is constantly being shuttled back and forth from the hard disk to the processor (using the RAM as an intermediary, to avoid the hard disk spinning up and down all the time) and this transfer of data is slow, inefficient and creates a bottleneck which restricts the maximum speed at which computers can run.
The brain, on the other hand works very differently. The brains computational power seems to be massively distributed. Specific functions are localised within regions of the cerebral cortex, but each area seems to have its’ own ability to a) make computations, and b) store the information it needs, at least temporarily. In other words, the computational substrate of the brain is the same as the storage substrate – neurons both store, and compute information at the same time. In theory this is much more efficient as it doesn’t require any moving around of data. Exactly how they do this is a bit of a mystery of course.
This post has been inspired by a paper just published in the journal Advanced Materials (not my usual kind of journal, but what the hell). The researchers in this paper describe a new type of electronic component which appears to mimic the dual-function of the brains neurons. They’ve used ‘phase-change’ materials to create a processor which can perform the four basic arithmetical functions (add, subtract, divide, multiply), and also stores data by the precise state of the materials’ crystallisation. This is fabulous stuff. Obviously it’s a very simple implementation at the moment, but it’s a step towards a ‘brain-like’ computing system, and the researchers say their next step is to link together a hundred or so of these chips and attempt to create a neural network for some simple tasks, like image recognition.
Future computers might look very different from current ones; they might make a modern computer based on Integrated Circuits look as clunky as the 1950s vacuum-tube monsters look to us now. Imagine opening a computer case and not seeing a system of discrete components, but instead seeing a big, homogeneous block of crystal, or a bag of goo, all of which contributes to massively parallel calculations and has a huge, distributed, storage capacity as well. Cool stuff.