A new study from MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL) demonstrates a novel method for training AI models that significantly reduces computational costs and energy consumption. The research, published in Science, introduces a technique called ‘modular training,’ which breaks down complex neural networks into smaller, specialized sub-networks that can be trained independently and then combined. This approach cuts training time by up to 50% and slashes the associated carbon footprint, addressing growing concerns about the environmental impact of large-scale AI development. The method also shows promise for improving model adaptability and efficiency in specialized tasks, potentially lowering the barrier to entry for AI research. Read the full article for a detailed breakdown of the methodology and its implications for the future of sustainable AI development at https://technologyreview.com/2024/05/15/modular-ai-training.