Apple's Big Model Update: ReALM Released, Visual Element Parsing Better Than GPT-4

Apple unveils ReALM, a new AI system that outperforms GPT-4 in understanding on-screen visual element references; ReALM enhances reference parsing for conversational assistants by transforming visual layouts into textual representations; Despite progress, ReALM has limitations in handling complex visual references and needs to be combined with multimodal techniques. (AI Tech Bazaar)