Artificial Intelligence Can Provide 100 Times More Energy Savings than Normal

Artificial intelligence, which is a matter of great curiosity, and the innovations it will bring to our lives are being talked about and discussed more and more every day. It is said that artificial intelligence, on which extensive studies have been carried out, will replace many professions in the near future. One of the recent rumors is about energy saving.

In this article, I will show you how artificial intelligence can save more energy with IBM’s new artificial synapses.

Artificial neural networks are indispensable for artificial intelligence studies.

Artificial neural networks can crunch data, translate text, and identify images with near-perfect accuracy, but these processes are a bit slow. That’s because the neural networks loosely embedded in the structure of the human brain are typically created in software rather than hardware, and the software runs on traditional computer chips.

IBM has shown that building the key features of a neural network directly in silicon is 100 times more efficient. Chips produced this way could turbocharge machine learning in the coming years.

The IBM chip mimics the synapses that connect individual neurons in the brain, like a neural network written in software. In order for the network to learn, the strength of these synaptic connections must be adjusted. This happens in the form of connections that grow or fade over time in a living brain, making it easy to reproduce in software but so far difficult to achieve with hardware.

IBM employees demonstrated microelectronic synapses in a research paper published in the journal Nature. Their work draws inspiration from neuroscience by using two types of synapses: a short-term one for software and a long-term one for memory. The workers also tested a neural network built from components of two simple image recognition tasks: handwriting and color image classification. They found that the system was truly a software-based deep neural network, although it consumed only about 1 percent of the energy.

This innovation could be critical in other areas as well.

These studies are not only important for artificial intelligence. If we think about it in terms of commercial production, what IBM did could be a huge commercial success. Even though the company isn’t selling computer chips these days, they may invest in efforts to reinvent computer hardware, hoping that new microelectronic components can help provide momentum for the next advances. This new technique could be the first step. It could make it more efficient and easier to implement machine learning on small devices like smartphones.