Multimodal models are fast becoming a reality -- consequences be damned

AI systems that can reason across audio, text, and other forms of data are fast becoming a reality. But they have major flaws, like bias.