USING TEXT-BASED AI MODELS FOR VISUAL TASKS

Summary from the TLDR AI newsletter

Researchers discovered that large language models can be powerful tools for visual tasks even though they were designed for text. The researchers achieved improved results in various visual tasks by using parts of these text-trained models to process images and videos directly.