The human brain is a wondrous thing. It can process many different types of complex information nearly instantaneously. In a sense, your brain is the best computer out there—although today’s most powerful computers based on silicon chips can complete mathematical operations far faster than the average human, they are unable to or clumsy at performing certain tasks that our brains can do with ease. For example, researchers at Google are developing a sophisticated algorithm that can perceive patterns, which would allow it to “recognize” a dog in a picture and return these pictures when you type “dog” into Google Images. Of course, we can see a dog in a picture right away.
What if there was a computer that acted like the human brain? A bio-inspired laser system might be able to do just that. New research published in Nature Scientific Reports describes a prototype that can deliver programmed light pulses in a similar manner to the firing of neurons in the human brain and nervous system. This technology takes us another step forward toward optical computing, in which information is encoded and processed using rays of light rather than electric current. Computers powered by photons would operate faster by ten times or more than those with standard electric circuits. Not only that, but since this system has features similar to the biological nervous system, it may be able to perform operations we’ve never seen before in traditional computers.
As you probably know, modern computers make use of digital processing to perform tasks. This means that every small piece of information or ‘bit’ is assigned a value of 0 or 1. You can think of this as a sort of threshold, like a hurdle on a race track, where any jump above the hurdle “passes” the barrier and is assigned 1 while all jumps below it “fail” and are assigned 0. This allows computers to perform math operations almost instantaneously by passing electric current according to the patterns of zeroes and ones.
Of course, in the real world there are many different kinds of information, not all of which are well suited to digital processing. For example, the recognition of patterns and the process of learning something new cannot easily be coded in a digital format. These tasks are better represented with a computer scheme resembling the nervous system, in which neurons are assembled in a complex network and can pass information along paths between them. Small signals in the brain stimulate the firing of neurons, which then cause other neurons in the network to fire, in an immense variety of sequences that correspond to all the ways we humans respond to the world around us. Building computer systems that behave like neurons would open up the world of computing to a whole new set of tasks that can’t be performed with conventional binary-logic systems.
The laser system developed by researchers at Princeton University can successfully process information in a way that mimics the activity of neurons, and it all works because of the special properties of graphene. Their system consists of a piece of graphene sandwiched within a ring of optical fibers. When the system receives an “input” signal of spikes appearing at random time intervals, the laser light circulating around the ring passes through the graphene, which absorbs different amounts of light at different intensities. At a certain threshold intensity, the graphene stops absorbing light, and the light escapes as an “output” signal.
After proving the neuron-like behavior of their simple one-ring system, the researchers adapted their device to create a circuit for performing pattern recognition. This involved making a linked network of several graphene fiber ring lasers. By connecting just two of their lasers together, they showed that the system can classify different signal patterns, like our brains do when we listen to music or see a set of familiar images.
They also created a “recurrent” circuit with their graphene laser, meaning that the signal pulse travels in an infinite loop around the ring without losing any of its quality. This type of network is a common motif in our nervous systems, and is important for the formation and recollection of memories. This recurrent circuit opens up the potential to make complex networks that contain millions of these laser rings, all originating at a single source.
With the capability of mimicking the firing of neurons in the human nervous system, this laser technology could allow us to create computers that operate at the speed of light and are comparable in complexity and performance to our own bodies. Still, the system is still at the proof-of-concept phase, and the researchers have yet to produce a system that could work in practical applications. And what will happen to us humans once we create computers that can do everything humans do? It’s definitely worth thinking about as technology propels us forward at lightspeed. For now, though, rest assured that your own mind is still the most high-tech processor around.
Top image: Brain anatomy medical head skull digital (CC BY-SA 4.0)
Bhavin J. Shastri, Mitchell A. Nahmias, Alexander N. Tait, Alejandro W. Rodriguez, Ben Wu, Paul R. Prucnal, Spike processing with a graphene excitable laser. Scientific Reports 2016; 6:19126.
Neil Savage, “Graphene Flakes Make Laser Neuron Superfast.” IEEE Spectrum, http://spectrum.ieee.org/nanoclast/semiconductors/optoelectronics/graphene-flakes-make-laser-neurochip-superfastb (23 Oct. 2016).
“Inceptionism: Going Deeper into Neural Networks.” Google Research Blog, https://research.googleblog.com/2015/06/inceptionism-going-deeper-into-neural.html (23 Oct. 2016).
Elisa Franco, Kate E. Galloway, “Feedback Loops in Biological Networks.” Computational Methods in Synthetic Biology, ed. Mario Andrea Archisio, Springer New York, 2015, pp. 193-214.
“Saturable Absorbers.” RP Photonics Encyclopedia, https://www.rp-photonics.com/saturable_absorbers.html (23 Oct. 2016).
Graphene, neuromorphic, spike processing, lasers, neural network
Image: biological or artificial neural network e.g. http://www.techsteak.com/wp-content/uploads/2016/02/340.jpg