Creating an AI that behaves like the human brain is one thing, but an AI that also performs as well as the brain is something else. Conventional circuitry can be a way more powerful than standard computers, but it can operate so fast only as a part of a neural network. However, it looks like the barrier has been cracked thanks to Alexander Tait and pals at Princeton University in New Jersey. They have created the world’s first photonic neural network, showing that it calculates at ultrafast speed.
The photonic neural network imitates the brain with neurons, which are some really light waveguides cut into silicon substrates. Each of those nodes works in a particular wavelength, so you can sum up the total power of the light to make calculations as it’s fed into a laser. The laser then sends the light back to the nodes and completes the circuit. A photonic neural network has the ability to calculate a differential math equation 1,960 times quicker, compared to a traditional processor.
The Princeton prototype consists of 49 synthetic neurons. It’s quite simple and not even close to replacing the processor in your phone. However, the photonic neural network could have remarkable implications for artificial intelligence. As the researchers claim, they could be specifically useful for fast processing info in control, radio, as well as scientific computing. It’s also important for the learning systems to achieve more human like accuracy and response time while identifying objects and actions.