The new AI model which Apple has been working on is dubbed as “ReaLM”, and will be an on-device AI enhancing the features of iOS, mainly Siri.
Apple, though considered to be lagging behind the AI race started by OpenAI’s ChatGPT, maintains patience in launching its first AI service or tool. Apple was seen working on AI ever since the talks of AI emerged and multiple names of the tech tools like AppleGPT had surfaced from the company sources.
After all, we can now arrive at a conclusion that Apple is on-progress to release “ReaLM” AI model, as the Apple’s AI researchers published a research paper sighting the same.
ReaLM AI by Apple
ReaLM stands for Reference Resolution as a Language Modeling (ReaLM) and is a conversational AI system with works by a novel approach of reference resolution. Reference resolution is the solution for the problem that occurs in linguistic process in natural language processing, where to know the meaning of one word, the system might refer to other words or entity as references.
Here, what it tries to achieve is that ReaLM answers to users’ queries by not just understanding the context in a conversation, but by processing onscreen content and detecting background activities. This has the potential to make voice assistants much more useful in a variety of settings, such as helping drivers navigate infotainment systems while driving or assisting users with disabilities by providing an easier and more accurate means of indirect interaction.
ReaLM is expected to improve Siri’s ability in Apple devices, personalizing it as possible as it can. The AI converts conversational, onscreen and background processes into a text format that can then be processed by large language models (LLMs), leveraging their semantic understanding capabilities.
ReaLM model comes at four sizes – ReaLM-80M, ReaLM-250M, ReaLM-1B, and ReaLM-3B. Yeah, you guessed it right, the suffix part represents the number of parameters that the model is trained with. ‘M’ means million and ‘B’ is billion. For comparison, GPT-4 boasts about 1.5 trillion parameters.
ReaLM vs other AI models
Researchers benchmarked ReaLM models against free GPT-3.5 and paid GPT-4. ReaLM’s smallest model performed comparatively to GPT-4, while their largest models did even better, according to the researchers.
“We demonstrate large improvements over an existing system with similar functionality across different types of references with our smallest modesl obtaining absolute gains of over 5% for onscreen references, the researchers explained in the paper. “We also benchmark against GPT-3.5 and GPT-4 with our smallest model achieving performance comparable to that of GPT-4 and our larger models substantially outperforming it.”
“We show that ReaLM outperforms previous approaches, and performs roughly as well as the state-of-the-art LLM today, GPT-4, despite consisting of far fewer parameters,” the paper states.
Apple is yet to confirm if iOS 18 or its latest devices will equip this ReaLM model. The firm is expected to come up with a series of AI announcements for WWDC 2024, that’s going to happen in June 2024.
Will Apple’s AI outperform its counterparts?
(For more such interesting informational, technology and innovation stuffs, keep reading The Inner Detail).
Kindly add ‘The Inner Detail’ to your Google News Feed by following us!
