Most digital computers are built on a simple, albeit revolutionary, principle suggested by Alan Turing in 1936 whereby data and instruction sets (the “programs”) are stored together (in your “Hard Disk”) whilst information processing takes place separately (in your “RAM memory”). This is what a “universal Turing machine” does; most modern digital computers today are universal Turing machines.
Alas, this is not how brains work. Brains function by performing processing and storage together. This realization has obsessed researchers in AI since the beginning, and various approaches have been tried to overcome this disparity, including parallel processing and neural networks.
(Picture left: The fundamental components of electronics)
In 2008 a team at HP labs invented the memristor, the “fourth” fundamental component in electronics (the other three are the resistor, the capacitor and the inductor). Memristors are like mutant resistors with a memory. Their resistance increases when the current flows one way and decreases when it flows in the opposite way. And they “remember” what current flowed through them. As a fundamental component in electronics, the application of the memristor is potentially boundless.
The exciting news for AI research is that memristors store and process information at the same time. With memristors one can build computers other than universal Turing machines that can do very interesting things. Researchers have already demonstrated that by using memristors they could model the adaptive behavior of unicellar organisms. Taking this finding a step further one could use memristors to mimic certain single cells called neurons, which in turn opens up the possibility of neuromorphic architectures performing massive pattern recognition based on memristive properties. Such architectures could learn and adapt their behavior accordingly, ushering a new era of “brain-like” computers.
(Picture left: A memristor under a miscroscope)
The research interest in this area is enormous and recently a research team at Exeter University led by David Wright announced a method of building “brain-like” computers by using phase-change materials to perform general purpose operations, such as the four basic ones of addition, subtraction, multiplication and division. Moreover, they have demonstrated that they can mimic neurons and synapses. Synaptic functionality was demonstrated by the “memlflector”, an optical analogue of the memristor.
Memristors are not only a major technological breakthrough but a theoretical one too, particularly in building intelligent machines. Mimicking brains with memristors suggests that we could develop machines that will program themselves, become thus independent of their programmers. We may imagine such machines exhibiting intuitive-like behavior by inventing novel solutions to unexpected problems. This may bring us closer to seriously begin to discuss technological methods for deciding whether such machines would be conscious or not.
References:
Turing, A.M. (1936), “On Computable Numbers, with an Application to the Entscheidungsproblem”, Proceedings of the London Mathematical Society, 2 42: 230–65, 1937
C. David Wright, Yanwei Liu, KrisztianI.Kohary, Mustafa M. Aziz, Robert J. Hicken. Arithmetic and Biologically-Inspired Computing Using Phase-Change Materials.Advanced Materials, 2011; DOI: 10.1002/adma.201101060
