A new study from MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL) demonstrates a significant advancement in making AI systems more energy-efficient. The research focuses on reducing the computational power required for large language models (LLMs) during the inference phase, when the model generates responses. By implementing a novel method that selectively activates only the necessary parts of the neural network for a given query, the team achieved a 50% reduction in energy consumption with minimal impact on performance. This approach, termed “conditional computation,” could make deploying powerful AI models more sustainable and cost-effective, especially on devices with limited resources. The findings highlight a growing priority in AI research: balancing capability with environmental and practical constraints. Read the full article at: https://technologyreview.com/2024/05/15/energy-efficient-ai-breakthrough-mit