One of the top tech companies is again back with its latest innovation that will compete against OpenAI. This claim has been made by Apple which has developed a new Artificial Intelligence system that will give direct competition to GPT-4. Apple is known for developing new features and innovations to enhance user experience. In this regard, Apple’s team developed its own AI system which claims to be better than OpenAI. The new AI system of the company named “ReALM” will beat OpenAI’s GPT-4. This news has created a buzz among Apple lovers who are looking to know more about it. A recent research paper published by the firm revealed the details of their new AI system.
The research paper reads that ReALM can comprehend ambiguous references to on-screen entities, background processes, and conversational context, enabling more natural interactions with voice assistants like Siri. The Apple researchers wrote, “Being able to understand context, including references, is essential for a conversational assistant. Enabling the user to issue queries about what they see on their screen is a crucial step in ensuring a true hands-free experience in voice assistants.”
How does Apple’s AI system ReALM work?
Apple’s research report claims that by utilising huge language models, Reference Resolution As Language Modelling (ReALM) is willing to turn the difficult issue of reference resolution into a language modelling challenge. This enables it to achieve significant performance improvements over current techniques. Using parsed on-screen elements and their locations, the system redevelops the screen to produce a textual representation that accurately captures the visual arrangement. This strategy, along with customising language models for reference resolution, enables ReALM to surpass GPT-4 in the task.
The researchers noted, “We demonstrate large improvements over an existing system with similar functionality across different types of references, with our smallest model obtaining absolute gains of over 5% for on-screen references. Our larger models substantially outperform GPT-4”.
Simply put, This AI system will help digital assistants get a better understanding of ambiguous references and context in conversations. With the help of this, the user will easily get to interact with a device that will look more natural. It is worth noting that, Reference resolution is essential to have a better understanding of language, particularly when a user is using indirect references or pronouns. The task of ReALM will be to handle this challenge by turning it into a language modelling problem. Apple’s AI system works by interpreting the visual layout of a screen and turning it into text.
The company believes that the new method developed in ReALM works much better than traditional approaches, even outperforming GPT-4 of OpenAI.
The researchers do point out that there are drawbacks to depending solely on automated screen parsing. More sophisticated visual references, such as differentiating between several images, will probably need the use of multi-modal approaches and computer versions. However, the firm has not made any public announcements, it is slowly moving forward in AI research, but it trails its competitors in the race to dominate the AI world. Tim Cook, CEO of Apple said, “We’re excited to share details of our ongoing work in AI later this year.”
It is assumed that the tech giant will share the AI announcements in the upcoming Worldwide Developers Conference. The dates of the annual conference WWDC 2024 are out and it is scheduled to be held from June 10 to June 14, 2024.