The Thinking Machine

The Thinking Machine

Physically mimicking the brain may soon become a reality! The “missing” factor which makes this true is the memristor: a thus far hidden component in the family of resistors, capacitors and inductors that emerged as a major technological breakthrough at HP labs.

Based on the work of Dr. Leon Chua, the memristor (a.k.a. memory resistor) remembers its state even after being turned off, so that a computer’s memory is still accessible right after it is turned on again.  You can imagine turning on your computer and not having to wait several minutes before the operating system is loaded in memory. Looking further ahead into the future of memristor-based machines, imagine an “intelligent” computer that would “understand” the tasks you are trying to perform and would help you do them more efficiently, even by giving you smart advice (although it would have to be more sophisticated than Microsoft Office’s infamous Paper Clip assistant). Wouldn’t that save you time and effort?

Achieving that goal is not a long way ahead. So far, non-volatile memory modules were developed using memristors that store up to 100 gigabits (12.5 GB) in 1 cm2, whereas conventional flash memory stores up to 16 gigabits (2 GB) in the same amount of space. However, the speed of memristor-based memory is currently ten times less than DRAM. As this is the first prototype, there is still room for improvement.

The memristor’s resistance depends on the direction the current takes; the resistance increases in one direction and decreases in the other. Therefore the resistance R is a function M of charge q passing through, i.e. R = M(q). But the key aspect is this: when the current stops flowing, the memristor stores the last resistance state and thus starts over at that same resistance once the current starts flowing again. The memristor developed at HP labs is made of titanium dioxide thin films which makes it extremely small, thus reducing power consumption and production costs. Also, when connected together, memristors have a shape similar to that of artificial synapses. These properties makes them attractive for building an artificial brain. “This new circuit element solves many problems with circuitry today—since it improves in performance as you scale it down to smaller and smaller sizes,” said Dr. Chua. “Memristors will enable very small nanoscale devices to be made without generating all the excess heat that scaling down transistors is causing today.”

Numerous research groups have worked on memristors and proposed several applications. But recently, HP labs revolutionized the application of memristor by building a neuromorphic chip which is similar to a biological system as it brings data and its computation  at the same location. In fact, in the brain, the actual data we perceive–say an image–is brought in through synapses and translated to vision at that same synaptic area. The memristor chip is basically a multicore microprocessor where each core has direct access to its own memory thus eliminating wires and large power consumption. This brain-like microprocessor will run MoNETA (Modular Neural Exploring Traveling Agent), software in development since last November at Boston University’s Department of Cognitive and Neural Systems, that will allow the machine to perceive its environment and opt for choices that will guarantee its survival. Thus, it will mimic the survival aspect of humans and animals which forms the basis of our evolution.

Currently, Boston University and HP are developing the perceptual, navigational and emotional systems which will simulate the behavior of a small mammal using hardware.  This simulated nervous system will learn through plastic changes in synaptic connections (similar to biological neurons). This will allow it to interact intelligently with its environment: searching for food, following learned paths, avoiding predators, etc. In the near future, we can thus expect to see burgeoning projects give rise to artificially created small animals that have almost all the capabilities of their biological peers

REFERENCES
IEEE Spectrum: Artificial Intelligence

Like what you see?

We release awesome content every Wednesday.
Stay updated; signup to our mailing list here:

Leave a Reply

Your email address will not be published. Required fields are marked *

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>