At the same time, AI models that are trained primarily with text to detect hate speech, struggle to identify hateful ... model to detect hateful memes is a multimodal problem.