Jun 28 2011

“Brain Like” Computing

I am fascinated by the quirkiness with which science is communicated to the public, and specifically the way the media filters and presents science. For example, there are a limited number of hooks used to sell science stories. The formula is as follows: take any scientific discovery and ask, “how will this discovery affect or interest the average person.” Then you find some holy-grail type problem or some iconic futuristic application and make any connection you can to it – no matter how tenuous.

Some of these science-journalism cliches are now familiar to all of us. Every fossil find has to be a “missing link.” If a connection can be made to human evolution, no matter how far removed from our tiny branch, then that connection will be pointed out. Every advance in virology might someday lead to a “cure for the common cold.” Similarly, any biological advance with the slightest possible connection will someday lead to a “cure for cancer.”  There is also the fairly recent addition – any advance in metamaterials will lead to the development of a “Harry Potter-like invisibility cloak.” The goal is to get the phrases in quotations into the headline of the article at all costs.

I believe there is another similar science journalism cliche in the making – relating any computing advance to the development of artificial intelligence or “brain-like” computing. The connection is not always unreasonable, but the catch phrases are key.

Media reporting on a recent study in Advanced Materials falls into this new cliche. The article itself is titled: Arithmetic and Biologically-Inspired Computing Using Phase-Change Materials. The press release – Study brings brain-like computing a step closer to reality.

“Biological inspired” or “brain-like” refers to the concept of building computers with components that both store and process information. Current desktop computers have separate components for these functions. The central processor, as the name implies, processes information, but the information itself is stored in the RAM (random access memory). For long term storage memory is placed on a hard drive or similar device, but in order to be processed information has to be loaded into RAM, which is the “working memory” of the computer. This is why increasing the RAM in your computer can make it faster, because it reduces the need to access the hard drive (which is much slower than accessing memory in RAM).

But still – information has to be shuttled back and forth between RAM and the central processor. The need to move around all that information is a significant limiting step in the speed and efficiency of computers.Vertebrate brains, on the other hand, use the same medium to both store and process information – neurons.

It has therefore been a goal of computer research to develop a material that can both store and process information – to merge RAM and the processor and thereby eliminate the need to constantly shuttle information back and forth between the two, resulting in huge gains in speed and efficiency.

Here is the abstract to the current research:

Phase-change materials offer a promising route for the practical realisation of new forms of general-purpose and ‘brain-like’ computers. An experimental proof-of-principle of such remakable capabilities is presented that includes (i) the reliable execution by a phase-change ‘processor’ of the four basic arithmetic functions of addition, subtraction, multiplication and division, (ii) the demonstration of an ‘integrate and fire’ hardware neuron using a single phase-change cell and (iii) the expostion of synaptic-like functionality via the ‘memflector’, an optical analogue of the memristor.

Sounds like the first step in developing the hardware for a computing system that combines memory and processing. That’s pretty cool – if it pans out (always the rub). This does seem like a promising direction to go in, and this type of technology may find applications in developing AI systems. It’s still a long way from being truly “brain-like”, except in the one feature described – so the connection is reasonable if a little thin. I wonder how long it will be before I have memflectors in my desktop computer or smart phone.

In any case, I may have to add “brain-like” to my list of Google alerts to track this research.

122 responses so far