Generating Images With Multimodal Language Models