‘SyNAPSE’ – Computers Working Like a Human Brain

Modern computers are really good at doing things that humans struggle with; they can do vast numbers of repetitive tasks, without tiring and consistently – the last time is as good as the first. This is because of how computers operate; they perform functions in a linear fashion, so one function at a time, rather than being able to perform multiple functions simultaneously. They can do these complicated repetitive tasks far quicker than the human brain can, just use the calculator on your computer to see what I mean.



An IBM Watson supercomputer. The company is now working on the SyNAPSE chip, which is structured to mimic the neural networks of a biological brain.

Computers improve by increasing the number of functions they can do per second; We talk about modern computers getting ‘faster’, so they are not fundamentally different to what has come before, they just do it more quickly. If the same was true of cars, then all we would do is increase the engine size, and see that as being the improvement. But there is a limit to this, the faster the chips run, the more heat it generates, and this needs to be dealt with; and that is one of the major limiting factors to producing ever faster processors in PC’s.

The long term answer is to fundamentally change how computers actually operate, and get away from the current linear approach. Computer companies have worked for a long time on creating software programs that emulate the human brain, you just need to look at speech & handwriting recognition, language translation and face recognition to see practical examples of software that work like this. But this is all done using existing linear style computers; the next frontier is building the hardware that also works like the human brain, so multiple processes are undertaken at the same and not in a linear ‘step by step’ fashion. These computers would naturally be better at doing human style tasks (that the software mentioned above is trying to emulate). This development has the umbrella title of ‘SyNAPSE’ (Systems of Neuromorphic Adaptive Plastic Scalable Electronics). IBM have so far managed to build a computer on this basis that has the power of a Bee’s brain, but a lot of money and a lot of manpower is being invested in this. If Moores Law is correct (where computer speed doubles every 18 months), we are not that far from a computer that has some genuine practical applications.

This type of computer uses a lot less energy, is naturally smaller & improvements are not reliant on the processor just being ‘faster’, as it is much easier to scale this type of PC up. And because they are better at doing the same tasks as humans, the way we interact with them will be a lot more intuitive, and dare I say ‘human’ and feel less like ‘programming’, which is essentially what we currently do with existing computers.

So computers moving forward will be a mix of a linear computer (and its ability to do repetitive, highly focused tasks) and these new computers that work more like a human brain (so is able to anticipate, and make more human style decisions). This is the same way that the human brain works, where the left and the right sphere work differently to each other, but both are needed and complement each other.

If you want to learn more about this;

The original Guardian article that inspired me to write this :- http://www.theguardian.com/technology/2014/dec/09/synapse-ibm-neural-computing-chip

A good video on IBM’s work on this :- https://www.youtube.com/watch?v=gQ3HEVelBFY

TED Talk – Cognitive computing :- https://www.youtube.com/watch?v=8zYp4yH4PoQ