A new study from MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL) demonstrates a novel method for training AI models to understand and generate images based on textual descriptions. The approach, called ‘MAGE’ (Masked Generative Encoder), treats image generation as a unified process of tokenization and prediction, similar to how large language models work. This method allows a single model to perform both tasks—creating images from text and identifying the content of images—more efficiently than previous systems that required separate models. The researchers report that MAGE achieves state-of-the-art performance on several benchmark tests, generating high-fidelity and diverse images while also excelling at image classification tasks. The technique could lead to more versatile and efficient AI systems for visual content creation and analysis. Read the full article at https://technologyreview.com/2023/06/01/XXXXXXX/.