Published on : Jul 30, 2019
Currently, we are experiencing a highly accelerated growth and adoption of AI that the ability to influence a broad range of changes in society. They include better lifestyle conditions, growth of the economy, ease of access to entertainment, well-being, and education. However, there are many questions about the sustainability of artificial intelligence with respect to factors such as accountability, explainability, and privacy among others. Thus, it is important for us to find a solution and that too quickly so we can tackle the problems created due to the current unsustainability of the AI.
Optical Processing Units May be the Future for AI
One solution to these recurring issues is based on the optical transistor and its functionalities. Recently Google announced the development of their Tensor Processing Unit or TPU. These TPUs consume much lesser power as compared to the CPUs and GPUs per unit of computation.
However, the question remains, can this transistor-based technology can be made faster and consumer even lesser power for computation? Well, the answer is a big yes! In recent years, there have been multiple attempts to leverage light for low-power and faster computations. Even though the current solutions are a bit rigid in terms of hardware and are compatible with certain machine learning modules, the research studies are trying to expand its reach.
In this field of study, France is leading in terms of developing new hardware. This is due to the availability of large amounts of both national as well as private funding to continue the research. LightOn, a French company, recently announced the development of a new device based on optics. The company is calling the device as Optical Processing Unit (OPU). OPUs have a very specific function that is transforming the linear input vectors followed by the transformation of non-linear vectors.