HP makes progress with memristor memory technology

Scientists from HP and the University of California have announced advances in the understanding of how memristor memory technology works, bringing the creation of artificial intelligence one step closer to reality.

The scientists published findings in Nanotechnology Journal about the physical and chemical properties of memristors, a type of non-volatile random access memory which does not 'forget' data even when power is turned off.

HP had already built working samples of memristors its labs, but the way in which the technology works was not fully understood until now owing to its minuscule size.

Highly-focused X-rays were used to map the the 100nm-wide channel where the switching of resistance takes place. This data can now be fed into a mathematical model of memristors and used to improve performance.

The technology has the potential to be a strong competitor to flash memory in about five years, according to Stan Williams, HP senior fellow, in an HP Memristor FAQ.

"The ability to 'remember' the total electronic charge that passes through [memristors] will be of greatest benefit when they can act like synapses within electronic circuits, mimicking the complex network of neurons in the brain, enabling our own ability to perceive, think and remember," he said.

Mimicking the brain's synapses could lead to the "creation of semi-autonomous robots" if complex networks of neurons can be reproduced in an artificial system, Williams added.

HP is not the only company trying to bring fictional Skynet-like technology to the real world. IBM has been working on enhanced artificial intelligence in the form of its Watson project.

The supercomputer beat human champions of the US trivia game Jeopardy in February, and is able to understand questions, figure out the context, associate the words with possible answers and present an answer.