Apple's ReALM performs better than GPT-4 and enhances Siri
The AI has a built-in image recognition system
3 min. read
Published on
Share this article
Read our disclosure page to find out how can you help Windows Report sustain the editorial team Read more
Apple is developing Reference Resolution As Language Modeling (ReALM), an AI model capable of outperforming OpenAI GPT-4. In addition, it should help Siri understand the context of a conversation better and process onscreen content. However, we don’t know if Apple will add its new AI with the iOS 18 update or wait for its upcoming phone lineup. Yet, they might reveal the large language model (LLM) during the Worldwide Developers Conference.
Everything we know so far about Apple’s ReALM
Up until recently, Apple kept its silence about its ReALM AI developments. However, there were rumors that the company was working on an LLM. Yet, they joined the AI race quite late, and Microsoft is already leading the market, especially since it keeps investing in various projects.
According to a research paper, the ReALM AI from Apple has four models: ReALM-80M, ReALM-250M, ReALM-1B, and ReALM-3B. On top of that, the M and B mean millions and billions of parameters. Also, the LLM can convert data into a text format that other models can understand easily.
The smallest model of Apple’s ReALM can compete in performance with GPT-4, while the larger models substantially outperform it. With this new AI technology, the company might be able to reclaim its lost territory from some of its competitors.
Even if the results of Apple’s ReALM are outstanding, it might not be able to keep up with the upcoming OpenAI model. Yet, AI technology development happens fast. So, Apple might need to focus on its characteristics rather than benchmarks.
The Apple AI has distinctive features
According to Business Insider, ReALM will not compromise the performance of your device. On top of that, with its help, Siri will understand your queries better. In addition, Apple’s LLM has an image recognition system that allows it to gather text information such as phone numbers, addresses, names, and more.
Apple researchers consider it important to train LLMs on screenshots and not only on images. This method could be handy in real-life situations.
In a nutshell, ReALM is a promising LLM and shows the progress made by the company in a short period. Additionally, its capabilities show that Apple is ready to join the AI race. Also, with it, Apple might be able to reclaim its market share from its competitors.
What are your thoughts? Are you eager to try the new Apple AI? Let us know in the comments.
