Mutant chips swap Moore's law for Darwin's

Moore's Law says that processor power will double roughly every 18 months, but scientists at Sussex University are developing techniques which owe more to Darwin's Theory of Evolution than to the co-founder of Intel's powers of prediction.

By applying evolutionary algorithms to electronic system design, the scientists have found that a computer may be able to redesign itself to work better and more efficiently than before.

At Sussex University's Centre for Computational Neuroscience and Robotics, Adrian Thompson has been developing chips that mutate by themselves, reforming their circuit structures over and over again to find the configurations that work best.

The chips, known as Field Programmable Gate Arrays, are covered in transistors called logic gates, which can be disconnected and reconnected dynamically to any other cell on the chip.

Reconfiguring the chip in this way allows the processor to be fine-tuned to suit the task in hand, and could allow computers to actually undergo hardware upgrades simply by downloading a new set of code or instructions.

"Evolutionary algorithms capture the bare essentials of Darwinian evolution - selection acting repeatedly upon heritable variation - but are in other ways very different from evolution in nature," said Thompson.

He explained that there are many potential applications for evolutionary algorithms in electronics, such as the optimisation of component placements in the design process itself.

"Evolved circuits can have a richer spatial structure and internal dynamics than normally envisaged, and can extract unusual leverage from the physics of their medium of implementation ... be that physical silicon reconfigurable chips, or even proposed future technologies for nano-scale systems," he said.

Although the application of evolutionary algorithms would go against the grain of currently accepted design rules, because even the scientists admit they can't understand how it works, the technology may be laying the foundations for developments in nano-technology.

These microscopically small computers are so far only talked up by cyber-technology writers like William Gibson, the author of Neuromancer, but with the development of evolutionary chips they may well become tangible visions of the future.