Researchers at George Washington University have recently achieved an exciting new breakthrough in the field of Artificial Intelligence; discovering that using light instead of electricity for performing computations can improve machine learning neural networks in terms of both speed and efficiency.
Machine learning neural networks are a type of AI that strives to replicate the same functions as performed by our brains, enabling it to learn tasks unsupervised.
The processors currently used for machine learning are restricted by the amount of power that can be feasibly provided to them. In addition to this, the limitations of these networks are also caused by the slow speed at which electronic data is transmitted between said processors and the memory.
What George Washington University researchers have discovered is that these limitations could be overcome with the use of a photon-based TPUs (tensor processing units), essentially using light instead of electricity, to create more efficient, and more powerful, AI technology.
The research on these photon-based TPUs revealed exciting potential, showing its performance could be 2 to 3 orders higher in comparison to electrical TPUs. This could yield a range of different commercial applications, including both 5G and 6G networks, in addition to operations that process large amounts of data.
This research was published in Applied Physics Reviews, titled “Photonic tensor cores for machine learning”, it’s abstract stating that:
“With an ongoing trend in computing hardware toward increased heterogeneity, domain-specific coprocessors are emerging as alternatives to centralized paradigms.”
“This work shows that photonic specialized processors have the potential to augment electronic systems and may perform exceptionally well in network-edge devices in the looming 5G networks and beyond.”
An author on the paper, Mario Miscuglio, commented the following on this breakthrough:
“We found that integrated photonic platforms that integrate efficient optical memory can obtain the same operations as a tensor processing unit, but they consume a fraction of the power and have higher throughput”
“When opportunely trained, [the platforms] can be used for performing interference at the speed of light.”
“Photonic specialised processors can save a tremendous amount of energy, improve response time and reduce data centre traffic.”
This innovative new research could potentially have transformative effects on the way these neural networks run, accelerating this technology’s capacity, and thereby its applications throughout the world. Whilst this research was only just released by the Applied Physics Review very recently, who knows the extent to the potential this breakthrough holds.
Dani’s actuarial experience and passion are key. He is a strong advocate of innovation, optimism and communication, both within the team and for the clients. Dani’s ability and experience with data ensure that we always maximise value and efficiency for every project, enabling us to unlock hidden value for the clients business.
