OpenAI has trained a 12B-parameter AI model based on GPT-3 that can generate images from textual description. The description can specify many independent attributes, including the position of objects ...