Apple claims that its ReALM system could potentially enhance Siri’s efficiency, surpassing OpenAI’s GPT-4 in this endeavor. The ReALM aims to enhance Siri’s contextual understanding in conversations, improve content processing on screen, and detect background activities more effectively.
Researchers compared the ReALM models with OpenAI’s GPT-3.5 and GPT-4 that power ChatGPT and ChatGPT Plus. In a recent paper, they noted significant advancements with ReALM across various reference types. They mentioned absolute gains of over 5% for on-screen references and substantial performance boosts compared to GPT-4 with larger ReALM models.
Despite having fewer parameters, ReALM is said to outperform existing systems and perform similarly to GPT-4, a leading language model today. With Apple’s WWDC approaching, the focus is anticipated to be on AI advancements, particularly in Siri’s functionality.
This development could pave the way for Siri to become more efficient than ever, marking a potential milestone in voice assistance technology.
