Apple working on a contextual AI language model called ReALM

  • Apple’s research team is currently developing an innovative artificial intelligence (AI) model, named Reference Resolution As Language Model (ReALM).
  • The model is designed to understand language in context and function entirely on-device, without needing significant computational power.
  • A recent research paper suggests that this new AI model may enhance the functionality of Apple’s built-in voice assistant, Siri.

Researchers at Apple have published a new paper on an artificial intelligence (AI) model called Reference Resolution As Language Model (ReALM) that is purportedly capable of understanding contextual language.

Resolution As Language Model

The AI model is named ReALM, which is an acronym for Reference Resolution As Language model. The paper emphasises that the main focus of the model is to perform and complete tasks that use contextual linguistic cues, which are more common in the way humans speak.

ReALM is made for performing tasks on a smart device.

These tasks are divided into three segments — on-screen entities, conversational entities, and background entities.

As mentioned in the research paper, on-screen entities are tasks that appear on the device’s screen, conversational entities are based on what the user has requested, and background entities are tasks that are happening in the background, such as a song playing in an application.

Also read: MM1: Apple’s first multimodal AI model

Also read: Apple to showcase ‘visionOS advancements’ at WWDC 2024

Compared with ChatGPT

The paper also claims that despite working in such a restricted environment, the AI model “substantially” outperforms OpenAI’s GPT-3.5 and GPT-4. It uses far fewer parameters than major LLMs such as GPT-3.5 and GPT-4.

The paper further elaborates that while the model scored higher than the GPT-3.5 on the text-only benchmark, it outperformed the GPT-4 on domain-specific user discourse.

While this paper is promising, it hasn’t been peer-reviewed yet, so its validity remains uncertain. But if the paper receives positive reviews, that could push Apple to commercialise the model and even use it to make Siri smarter.

Jennifer-Yu

Jennifer Yu

Jennifer Yu is an intern reporter at BTW Media covering artificial intelligence and internet governance. She graduated from The University of Hong Kong with a Master’s degree in Journalism. Send tips to j.yu@btw.media.

Related Posts

Leave a Reply

Your email address will not be published. Required fields are marked *