DARPA to allocate $1.83 million to measure AI energy consumption — OilPrice
The Defense Advanced Research Projects Agency (DARPA) will allocate $1.83 million for the two-year E-CALC project, designed to assess the energy consumption of artificial intelligence systems at all stages of the workflow. The project will involve specialists from Michigan Engineering, the University of Illinois Urbana-Champaign, and defense contractor BAE Systems, OilPrice reports.
Assessing energy costs
The project is called E-CALC — Energy Calculus for ML Lifecycle Optimization. Its goal is to enable developers to assess energy use at every stage of AI operation: from data processing and model training to its practical application.
According to the developers' plan, the system is intended to replace numerous costly and time-consuming tests of different software and hardware combinations with energy-use assessments. Michigan Engineering noted that a mismatch between software and hardware can result in the loss of up to 30% of the electricity used to train AI models.
More current news is available on the UA.News Telegram channel Telegram.
More accurate tracking of energy costs, according to Michigan Engineering, could reduce electricity consumption, lower the cost of using AI, and extend the operating time of battery-powered devices.
Defense applications
The technology may be useful for military and defense systems, where field equipment and remote computing systems must run complex AI algorithms under strict power and battery-charge constraints.
E-CALC lead, University of Michigan Associate Professor of Computer Science and Engineering Mosharaf Chowdhury, stated that uncontrolled growth in AI energy consumption could increase costs, place pressure on energy infrastructure, and complicate the deployment of advanced systems. According to OilPrice, the International Energy Agency forecasts that by 2030, electricity consumption by data centers will more than double compared with 2024, when it amounted to about 420 terawatt-hours.
