A newly developed education method can accelerate the learning process of artificial intelligence by 100 times while significantly reducing energy consumption.

Today, artificial intelligence technologies have become integrated into every aspect of our lives. However, the training of large language models (LLMs) and other artificial intelligence systems is causing concerns about sustainability due to the increasing energy consumption. Data centers in Germany consumed 16 billion kilowatt hours (kWh) of electricity in 2020, and this figure is expected to rise to 22 billion kWh by 2025. However, scientists are aiming to solve this problem with a new method for artificial intelligence education.

According to Scitech Daily, traditional artificial intelligence education is a process that requires a large amount of computing power. Neural networks increase the accuracy level by adjusting parameters over many iterations while processing data. However, this process is both time-consuming and requires high energy consumption. Physics-based Machine Learning professor Felix Dietrich and his team have developed an education method that can fundamentally change this process. The new method accelerates learning by using a probability-based approach instead of training neural networks through iterations. Thus, artificial intelligence can learn 100 times faster compared to traditional methods while maintaining the same accuracy level.

HOW DOES IT WORK?
In traditional artificial intelligence training, the parameters inside the network are randomly selected, and the model is optimized by adjusting these parameters over thousands of iterations. In the new method, probabilities are used to determine parameters at critical points. This significantly reduces the computational load and speeds up the learning process. Researchers indicate that this method can lead to significant advancements not only in artificial intelligence education but also in dynamic systems such as climate models and financial markets.

A MORE SUSTAINABLE FUTURE
Dietrich states, “Our method enables artificial intelligence models to be trained with much less energy, thus reducing costs and environmental impact.” He emphasizes that this innovation could be used more extensively in the future. This new approach could transform artificial intelligence technologies into sustainable technology by increasing energy efficiency. If applicable in large-scale systems, it could open the doors to a whole new era in artificial intelligence education.