A newly developed educational method can accelerate the learning process of artificial intelligence by 100 times while significantly reducing energy consumption.
Artificial intelligence technologies are now 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. Nevertheless, scientists aim to solve this issue with a new method focusing on artificial intelligence education.
According to Scitech Daily, traditional artificial intelligence training is a process that requires a significant amount of computational power. Neural networks increase accuracy levels by adjusting parameters over many iterations while processing data. However, this process is both time-consuming and requires high energy consumption. Professor Felix Dietrich and his team in Physics-Based Machine Learning have developed an educational method that could revolutionize this process. The new method accelerates learning by using a probability-based approach instead of training neural networks with iterations. This way, artificial intelligence can learn 100 times faster compared to traditional methods while maintaining the same level of accuracy.
HOW DOES IT WORK? In traditional artificial intelligence education, parameters within the network are randomly determined, 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 burden and accelerates the learning process. Researchers point out that this method could lead to significant progress not only in artificial intelligence education but also in dynamic systems such as climate models and financial markets.
A MORE SUSTAINABLE FUTURE “Our method enables artificial intelligence models to be trained with much less energy, thereby reducing costs and environmental impact,” says Dietrich, emphasizing that this innovation could be used more widely in the future. This new approach could transform artificial intelligence technologies into a sustainable technology by increasing energy efficiency. If applicable in large-scale systems, it could open up a whole new era in artificial intelligence education.