The technology is centred on neuromorphic chips which are made up of networks of transistors that interact the way human neurons do. This means that they can process analog input, such as visual information, quicker and more accurately than traditional chips.
Bhavin Shastri, a postdoctoral fellow in electrical engineering at Princeton University said that one way of building such transistors is to construct them of lasers that rely on an encoding approach called “spiking.”
Depending on the input, the laser can provide a brief spike in its output of photons or not respond at all. Instead of using the on or off state of the transistor to represent the 1s and 0s of digital data, these neural transistors rely on the time intervals between spikes.
Shastri said: “We’re essentially using time as a way of encoding information. Computation is based on the spatial and temporal positions of the pulses. This is sort of the fundamental way neurons communicate with other neurons.”
Shastris work with Lawrence Chen, a professor of electrical and computer engineering at McGill University, is trying to get the laser to spike at picosecond time scales which are one trillionth of a second.
They managed to do this by putting a tiny piece of graphene inside a semiconductor laser. The graphene acts as a “saturable absorber,” soaking up photons and then emitting them in a quick burst.
Graphene is a good saturable absorber because it can take up and release a lot of photons extremely fast, and it works at any wavelength.It also stands up very well to all the energy produced inside a laser.