Read the May-June issue of MIT's Technology Review magazine at www.techreview.com. "Beyond Silicon" is about computers after Microsoft breaks up and Moore's Law poops out. In 1965, Gordon Moore promised that silicon device densities would double every 18 months.
With Intel's help, he's been keeping that promise for 35 years. But some say limits loom. Even Moore says his law might be repealed after 2017.
So, some scientists are now into molecular computing. Molecule switches, at a few nanometers, are much smaller than - and might be connected by carbon nanotubes - to replace silicon transistors.
Scientists hope soon to build a 16-bit molecular memory that is 100 nanometers on a side. Defect-tolerant software might run reliably on molecules despite statistical fluctuations in their chemical manufacturing.
So, other scientists are now into quantum computing. Their approach is not to overcome statistical fluctuations, but to exploit them.
Quantum bits, or qubits, can be used, for example, to factor huge numbers for our National Security Agency - if I told you why that's important, I would have to kill you. Physicists envision, but can't yet come close to building, quantum coprocessors.
So other scientists are into biological computing. They set bacteria loose in beakers to perform computations. One scientist has a genetic flip-flop, a basic computer building block. Another has a bacterial clock that ticks every 150 minutes. Yet another is figuring out how to combine what he calls genetic applets.
Still other scientists are into DNA computing, which is also done in beakers. A simple version of the travelling salesman problem has been solved with strands of DNA. A single test tube of DNA might perform 10 trillion additions per second, a million times faster than an electronic computer.
Of course, the doublings we enjoy under Moore's Law might just continue with silicon. Might the doublings continue by pushing silicon chip fabrication from two dimensions into three?Then there's gallium arsenide (GaAs), a semiconductor often touted to outperform silicon. Might GaAs, or some other exotic semiconductor, extend Moore's doublings?A blind spot in "Beyond Silicon" is optical computing. Optical transmissions are doubling way faster than Moore's Law. Optical switches are shipping. Let's use photons instead of electrons in computing circuits.
Parallel computing is another likely source of doublings in the growth of computing power. But this brings us perilously close to Metcalfe's Law, so let's drop it right there.
Which gets us to Wirth's Law. Nicklaus Wirth is at the Swiss Federal Institute of Technology in Zurich. He says software gets slower faster than hardware gets faster. Or, referring to the recently chairmanised CEOs of Intel and Microsoft, Grove giveth and Gates taketh away.
We might also rely on software to get future doublings of computer power. Sun chief scientist Bill Joy estimates that in our next factor of a trillion in improved machine intelligence, one factor of a million will come from hardware and the other from software.
The impending breakup of Microsoft will surely hurry that up. Now, from wherever future doublings of computing are to come, there is one thing for sure: the kinds of computing in the future will be different.
Early computing was about weaving and artillery ballistics. After several doublings, computing became debits and credits, then inventory management. Next came the four major personal computer applications: word processing, spreadsheets, word processing, and spreadsheets. Today, with Pentiums running Windows, computing is mostly about Napster.
Perhaps, with some combination of silicon, molecular, quantum, biological, DNA, three-dimensional, plutonium cyclamate, optical, parallel, and post-Microsoft computing, we'll finally be able to compute world peace.
Technology pundit Bob Metcalfe welcomes comments on the Pay-AS-We-Go Internet at firstname.lastname@example.org