Uncovering the Expense of Producing AI-Generated Images

Researchers from Carnegie Mellon University and AI startup Hugging Face collaborated to research the carbon footprint of AI-generated images. They’ve found that creating an image using AI requires the same energy as charging a smartphone. Text generation, on the other hand, uses much less energy. The researchers urge machine learning scientists to be transparent about the nature and impacts of their models to understand their environmental impacts. With large-scale AI systems like OpenAI’s ChatGPT, which has millions of daily and monthly users, the energy consumption adds up.

For example, one writer’s photo taken with different arm positions was likely stitched together from multiple images by an algorithm, similar to what smartphones do. Another story discusses the controversy surrounding whether a game produced by a large studio can be considered “independent.” The Analogue Pocket portable console’s restock announcement is detailed, and Google’s delayed GPT-4 rival launch event due to struggling with non-English queries is also reported.