Is AI eating all the energy?

The article discusses whether AI technologies contribute significantly to energy consumption and compares their energy usage to other common activities. It covers key issues of AI efficiency, training and inference costs, and their impact on data center energy use.

• AI energy use is in focus due to its impact on the environment.
• AI's energy consumption is not inherently wasteful.
• Training AI models incurs a high, one-time energy cost.
• Enduring AI tasks consume less energy - about 3 Wh per task.
• AI energy use in data centers is roughly 4% of the total consumption.