Apple researchers have released a preprint paper on their ReALM large language model, which claims to outperform OpenAI's GPT-4 in specific benchmarks. ReALM can understand and handle different contexts, allowing users to point to something on the screen or running in the background and query the language model about it. This ability to understand exactly what is being referred to is crucial for chatbots, as it would create a hands-free screen experience. The researchers aim to use ReALM to understand and identify three types of entities: onscreen entities, conversational entities, and background entities. They demonstrated large improvements over an existing system with similar functionality across different types of references, with the smallest model achieving absolute gains of over 5% for on-screen references. But, Apple's plans to integrate ReALM into its iOS and macOS software offerings are yet to be revealed.