Engineers at the University of Pennsylvania have developed the first optical chip capable of processing about two billion images per second.
The engineers published their study in the scientific journal Nature, in which they note the first technique that can classify numbers almost instantly
Without the need for a system that converts optical signals into electrical signals or converts input data into binary format
This eliminates the use of large memory modules.
This is done by directly manipulating the incoming light using a deep optical neural network and applied to a 9.3 square millimeter chip.
So the basis for the new chip is a neural network capable of processing information in a similar way to the human brain. What is the difference between a CPU and a GPU?
What is the chip created by the engineers, and does it resemble the human brain?
The chip consists of nodes that are connected to each other like normal neurons
It is also programmed with a set of data to identify objects within the image or words
Its mechanism of action relies on processing information in optical signals instead of relying on normal electrical signals
This is done using optical cables as if they were neurons stacked in several layers, each one specialized in a particular type of classification.
Several practical tests have been conducted to prove the efficiency of this chip
Where: Scientists have created a chip (9.3 mm²) that is used to classify a series of handwritten characters that appear to be letters.
After programming it, the chip was able to classify images with an accuracy of 93.8% for groups that contain two types of characters
and an accuracy of 89.8% for groups containing four different types.
The most significant achievement is that the chip was able to classify each letter in 0.57 nanoseconds, allowing it to process 1.75 billion images every second.
The team commented: This speed comes from the chip’s ability to process information as light, which gives it many advantages over current computer chips.
In an interview with Firouz Aflatoni, lead author of the study, he said:
Our chips process the information through what we call “computation by propagation,” where the computations are performed as light propagates through the chip.
Bypassing the step of converting optical signals into electrical signals to read and process data, making it significantly faster.
Another advantage that these chips offer is that it is not necessary to store the information being processed
This saves time and space by not needing a large memory unit
It also provides a lot of security as many studies have shown that not storing data is more secure as it prevents any possible leak.
Plato concluded: “This system can do much more than classify images.
We already know how to transform different data into an electric field.
Now, we are able to do this in the visual field and process this data instantly using this technology.
So the team’s next achievements will be scaling up the device and modifying the technology to process different types of data.” Headphones and earphones can cause hearing loss