Google DeepMind's EmbeddingGemma 2 Raises Questions About the Cost of Multimodal AI

At a glance

Companies are increasingly weighing efficiency as much as performance when they adopt AI. Google DeepMind's EmbeddingGemma 2 has raised questions about this trend. The main concerns are how many graphics processing units (GPUs) a task needs, how much slower responses get when images and audio are handled, and how much operating costs rise.

The original site blocks automatic reading or only shows its text in a browser, so this note covers just the headline and introduction. See the original article below for the full story.