Periodically, we see articles heralding the imminent death of Moore’s law and accelerating progress in the world of electronics. For the sake of your portfolio, don’t buy it. Creative destruction, using Joseph Schumpeter’s term, will continue to create breakthrough profit opportunities.
In 2007, Hewlett-Packard’s labs demonstrated the first memristor recognized as such. A portmanteau of “memory” and “resistor,” “memristance” was the theoretical fourth circuit variable first described in 1971. While HP stock will probably not yield the sort of profits we’re looking for here, it will help generate them indirectly.
Because of its unique properties, memristors will enable far more powerful circuitry. Unlike transistor-based circuits that form the core of modern electronics, memristive circuits retain their state after losing power. Theoretically, you could power on a memristor-based computer and have all the data in memory that it had when you powered off. Memristor memory could replace hard drives and transistor-based RAM.
Memristors, however, can do more than act as memory. They can replace existing processing components. This means that much more functionality can be implemented in a single component. Instead of busing data back and forth between separate memory and processing locations on a circuit board, memristors do it all. Data, then, are available for processing with shorter wait times. Memristors reduce total hardware size, cost and energy consumption. Yet memristors can multitask in other ways, opening up a whole range of exciting possibilities.
Existing circuit elements only have two states: something or nothing, one or zero. Binary arithmetic is, therefore, used to build up the mathematical functionality needed to implement the logic that gives a computer “intelligence.” Unlike binary elements, memristors can exist in more than two states. This allows the use of higher base number systems. A single memristor could, therefore, perform the work of many binary transistors. This would permit faster, more powerful processors and higher-density memory. Additional reductions in power consumption is another outcome, of course, since fewer elements would be needed.
The ability to be either logic or memory, along with the ability to assume multiple states, makes memristors more like neurons in the human brain. This similarity isn’t lost on researchers. The multistate capability, taken to its logical conclusion, means that a memristor could exist in virtually infinite states. Then, memristors could be used as analog, instead of digital elements. Such devices would be able to learn on their own, computer theorists say. In fact, simple memristive circuits have already been used to model the adaptive ability of amoebas.
Efforts are under way to fully map the 100,000 neurons in a fly’s brain. Once completed, the fly brain’s “circuitry” could possibly be reverse-engineered with memristive circuitry. This would help us to better understand animal intelligence, and hence be able to develop true artificial intelligence. If sufficiently advanced, such computing devices would be able to easily recognize and differentiate between human faces and understand natural human language.
The memristor revolution is coming. HP expects its first commercially available memristive circuits to be available in mid-2013.