Rumors are circulating with increasing insistence that speak of a major enhancement of iOS 18, the next version of Apple’s mobile operating system, in terms of artificial intelligence. By now we know very well that the official presentation should take place in June during WWDC 2024, the developer conference, while the final distribution is scheduled for the fall, coinciding with the launch of iPhone 16. Thanks to a new research report from Apple, we can get an idea of how artificial intelligence could change the game, perhaps as early as the company’s next mobile operating system.
The documentation talks about a new artificial intelligence system called ReALM (Reference Resolution as Language Modeling), whose main goal is to help Siri or other voice assistants not only listen, but to understand requests in depth.
One of the goals of ReALM should be the one long pursued by all the other competitors, that is, to be able to make the language more conversational, fluid, contextualized and consequently “intelligent”. But here the approach may be different from that of traditional methods that focus on the conversational context. ReALM converts conversations, screens, and background tasks into a text format that can be processed by large language models (LLMs), leveraging their semantic understanding capability.
“We demonstrated significant improvements over an existing system with similar functionality across different types of references, with our smallest model achieving absolute gains of more than 5% for on-screen references. We also benchmark GPT-3.5 and GPT-4, with our smallest model achieving comparable performance to GPT-4, and our largest models outperforming it substantially.”
The document describes four dimensions of the ReALM model: ReALM-80M, ReALM-250M, ReALM-1B, and ReALM-3B. The “M” and “B” indicate the number of parameters in millions and billions, respectively.
Apple has not yet confirmed whether this research will play a role in iOS 18 or its newer devices, however, the promising results of ReALM could lead to a significant improvement in Siri and other AI systems.
