17:23
10:24
09:52
11:01
10:52
01:59
17:23
10:24
09:52
11:01
10:52
01:59
17:23
10:24
09:52
11:01
10:52
01:59
17:23
10:24
09:52
11:01
10:52
01:59
Apple researchers have introduced ReALM (Reference Resolution as Language Modeling), an AI system designed to significantly improve how voice assistants understand and respond to commands.
In a research paper, Apple describes the new system as how large language models solve a reference resolution problem that involves deciphering ambiguous references to on-screen objects, as well as understanding conversational and background context.
Reference resolution, a key aspect of natural language understanding, has historically posed challenges for digital assistants. ReALM tackles this by transforming the complex process into a language modeling task, allowing it to decipher on-screen elements and integrate this knowledge seamlessly into conversations.
By reconstructing the visual layout of a screen using textual representations, ReALM surpasses traditional methods and even outperforms OpenAI's GPT-4. This breakthrough could lead to more intuitive interactions with voice assistants, particularly in scenarios like driving or assisting users with disabilities.