I just love receiving challenging questions and, often, being given an ‘answer’ is more of an anti-climax than informative. When at school in the mid 60s I remember our physics teacher, Mr Whitworth-Smith, had a teaching style based on posing the question and giving you tiny clues as to where you could find the answer or, better, how you could go about working it out. He was one of my formative role models and, although I suspect that he isn’t around anymore as he would be close to 100 by now, wherever he is I send him my thanks.

Anyway, one of my colleagues at Leeds, Dr Jon Summers, reminds me of dear ol’ Smiffy and certainly has the same teaching style – I always learn something after spending just an hour with him and invariably come away with reading material that expands my horizons. Our last soirée ended in an Indian restaurant (where else, always good in Leeds) dribbling into a Chicken Naga and chewing over Moore’s Law. The ‘scaling’ process of CMOS (complementary metal-oxide-semiconductor) is nearing the end with the production of 14nm microprocessors although the overarching Law, ‘doubling the number of transistors every two years’, will probably continue for a few rounds yet - albeit by architecture methods other than scaling, e.g. 3D.

We desperately need to find the paradigm shift from CMOS quickly or the avalanche of data traffic will swamp the current generation of ICT hardware and we will suffer unprecedented growth in power consumption or the Internet will have to be throttled-back or taxed (Jevons style).

The problem is that, today, silicon can have six million transistors in the area of one full-stop on this page but doubling that presents real problems of ‘sculpture’, the CMOS cell size being too big to make switches (transistors) that work efficiently at smaller scale. The perfect switch is one that can turn off fully, turn on fully, has a trigger (gate) that consumes almost no energy and, if that wasn’t enough, doesn’t leak in either direction. Oh, and does so really (really) fast. Now with a commodity chip from Intel having 1.48 billion switches on it, the enormity of how far we have come has to be marvelled at - after all, the first single transistor was big enough to be assembled by hand in 1947 by Bell Labs…

It could have been between beers and after the onion bhaji that Jon passed me my latest reading matter, but in fact, it was less romantically shoved under my nose in his office with the challenge: ‘read that!’. The title alone made me salivate, ‘Science and Engineering Beyond Moore’s Law’. A full-blown treatise on all (those know so far) optional technologies, architectures and materials that ‘could’ replace CMOS/silicon and maintain Moore’s Law for another 50 years. However it wasn’t the discussion on graphene or arsenide’s that fired up the old grey matter but the possibility of using one of the family of E. coli (Escherichia coli) - yes, a natural bacteria, as a logic switch or programmable memory.

The point is that if, for CMOS, you plot ‘instructions per second’ against ‘bits per second’ the curve follows an exponential trajectory (the same principle as Moore’s Law) BUT clearly will miss the human brain by several tens of orders of magnitude. Our efforts with CMOS/silicon are nothing (almost literally) compared to what nature achieved hundreds of thousands of years ago.

Nature (God, if you prefer) appears to be on a parallel, but far advanced, curve, not just further on, and it/he/she may even have achieved perfection already. We are probably on the ‘wrong’ technology curve and maybe should jump to a computer model based on natural bio-cells not semi-conductors? I was reminded of one of the key methods that inventors use to dream up ‘new’ ideas – using nature as an exemplar, such as in the structural engineering of plant-life etc.

Now I am no biologist so had to bone up (pun intended) on cells etc and then, as I re-read, the learned paper it hit me. I got it. What is a bacterium/virus’s function? To reproduce itself, and it does so by running software code (DNA) through fine grained logic switches (akin to a programmed transistor array) with consummate efficiency, zero leakage and on a scale that is one millionth the power demand of our best transistor.

E.coli is just one, of many, suggested platforms that could be a good candidate for a replacement of CMOS – so would this be the first ‘real’ computer virus? God knows… and has done for millennia.

Read the paper: ‘Science and Engineering Beyond Moore’s Law’, Cavin, Lugli and Zhirnov published May 13^th, 2012, in Vol.100 of Proceedings of the IEEE.