Can we drive artificial intelligence’s exponential growth sustainably?
The rapid growth of deep-learning models has raised concerns about the compute cost and environmental impact of large language models like ChatGPT. In digital processors, every arithmetic calculation involves a large number of transistor operations, hence a large power consumption. As an example, some estimates say that the data centre industry accounts for as much as 14% of electricity consumption in Ireland alone.
This energy use is only expected to increase exponentially. And so, to continue reaping the benefits of AI advancement, the industry is looking to a new paradigm shift in the underlying hardware.
Our CEO and co-founder, Tim Weil, recently authored an article for Embedded.com titled “Can we drive artificial intelligence’s exponential growth sustainably?”, outlining the unprecedented energy efficiency and computational power offered by optical computing.
By utilising the properties of light, optical computers can perform computations 1000x faster than traditional electronics, simultaneously consuming only 1/100th of the energy. Compared to the unavoidable heat generated by traditional electronics, optical computing achieves negligible power consumption.