At University College London (UCL) there is a new computer system which is able to repair itself to keep critical systems in working order. This new technology works by linking data with instructions for how to handle it. For example, it would link the temperature outside with what to do when it's too hot, and then divide the results into a pool of 'systems'. Each system has a memory with context-sensitive data (meaning it can other interact with other similar systems). Instead of using a counter, systems are executed based on a pseudorandom number generator which attempts to introduce the randomness of nature into the system. Systems then carry out their instructions simultaneously rather than using a queue.

This system contains multiple copies of its instructions, spread across many subsystems so that if any one of them fails or becomes corrupted, the computer can access a clean copy to repair its own code. Instead of crashing like a typical computer system upon encountering an error, this system continues on because each subsystem has its own memory.

This technology has the potential to be used by combat drones to regenerate or to create more realistic models of the human brain. Modern day computers are inadequate at modelling processes such as neuron activity or bees swarming. This is because nature is "distributed, decentralized, and probabilistic" as well as being capable of self-correction.

The researchers who developed this new computer system plan to present their findings to a conference on evolvable systems in Singapore in April.

Sources:
New Scientist