According to a recent report, Apple has developed its extensive language model (LLM) AI framework, along with a chatbot service similar to OpenAI’s ChatGPT, Google’s Bard, and Microsoft’s Bing Chat, which is based on ChatGPT.
Insider sources report that Apple has internally codenamed the project “Ajax,” but some engineers within the company informally call it “Apple GPT.” The development team has constructed this language model using the
Google machine learning framework, Jax, and is currently hosting it on the Google Cloud platform.
Mark Gurman of Bloomberg Report
Mark Gurman of Bloomberg reported the information, relying on insights from “people with knowledge of the efforts.” There have been rare instances where he may have misinterpreted the information.
Apple initiated the development of Ajax “as an experiment” towards the end of 2022. Since then, Apple employees have utilized it for specific tasks, including product prototyping. However, the project has faced some challenges with intermittent progress. Initially, it was halted due to “security concerns.” Additionally, there is an internal rule at Apple preventing its use in creating consumer-facing applications or products for the time being.
Employees need to go through an application and approval process to gain access to Ajax. Once approved, they can utilize it as a web application, although it features a “stripped-down design” not intended for public use.
As of now, Apple has not reached a definitive plan for implementing Ajax-driven features for consumers. Apple executives, John Giannandrea and Craig Federighi, lead teams dedicated to resolving this matter. They are striving to devise a strategy, and there are rumours that Apple plans to announce this development in 2024.
Since he departed from Google in 2018, John Giannandrea has been spearheading Apple’s AI endeavours. However, Apple’s AI strategy has been characterized by a conservative approach, primarily concentrating on applications that leverage on-device processing to enhance the user experience or introduce additional features. This approach stands in stark contrast to the potentially more radical experimentation seen in companies like Google or OpenAI.
LLM based Chatbot
Earlier this year, a report from The Information suggested that Apple was engaged in a project involving LLMs, although the details were not explicitly disclosed. The report highlighted Apple’s cautious approach towards fully embracing the technology, as the leadership expressed concerns about potential unforeseen consequences of LLM-based chatbot and features providing inaccurate information or what AI researchers call “hallucinations” – errors in generating content. More recently, it came to light that Apple had imposed restrictions on ChatGPT usage among its employees.
Furthermore, the majority of LLMs, including Ajax, typically operate in the cloud. However, for various reasons, Apple has shown a preference for keeping AI-based features primarily running on users’ local devices.
According to The Information’s report, Apple was experiencing a loss of top AI talent to companies such as Google and Microsoft. The reason cited was that these companies held less conservative attitudes towards AI, which might have been a contributing factor in attracting skilled professionals away from Apple.
In summary, Apple actively develops and experiments with ChatGPT-like AI tools internally. However, the incorporation of these tools into Apple products faces several significant obstacles. As of now, it remains uncertain how, or even if, Apple will attempt to overcome these barriers and integrate such AI capabilities into their products.