Sources say Apple spends millions of dollars a day developing Apple GPT

In July, murmurs surfaced that Apple had clandestinely convened a team to cultivate and deploy its own Large Language Model (LLM), whimsically dubbed “Apple GPT” by internal aficionados. This model’s aspirations encompass conversational AI, automation, and multimodal systems.

Recent disclosures by The Information reveal that Apple currently allocates millions of dollars daily toward training the LLM under its foundational model team—a division, intriguingly, established four years prior, not a nascent endeavor. At the helm stands Siri’s Chief Executive, John Giannandrea.

Although Giannandrea exhibits palpable skepticism toward chatbots, Apple has persistently experimented with its proprietary versions, such as Ajax. This prototype reportedly outperforms OPENAI’s GPT-3.5.

Sources suggest that Apple’s prodigious daily expenditure on Apple GPT is aligned with more ambitious, expansive objectives.

In the current AI renaissance, Apple might be perceived as a trailing titan. Yet, undeterred by such optics, the corporation channels its intrinsic ecosystem to innovate AI models. However, compared to behemoths like Microsoft, Apple’s foundational model team, a modest contingent of 16, seems rather diminutive.

Regarding its pursuits, Apple’s vision transcends mere conversational robots. The aspiration is to architect a Siri function capable of autonomously executing intricate user commands, eliminating reliance on pre-programmed shortcuts.

At present, to accomplish complex tasks, users must preconfigure shortcuts. Siri, in its current avatar, merely serves as a vocal executor of these shortcuts, bereft of comprehending the myriad steps embedded within. Yet, the promise of AI, exemplified by GPT’s analytical prowess, could dissect a labyrinthine task into sequential steps—a fusion with Siri would undoubtedly be a riveting evolution.

Insider whispers forecast this sophisticated Siri functionality’s debut with iOS 18, likely around this juncture next year.

In parallel, Apple’s auxiliary teams delve into the AI-driven generation of videos and images, and the understanding of interrelations amidst texts, images, and videos, denoting the burgeoning realm of multimodal systems. Contemporary exemplars in this domain include Google Bard and Bing Chat, already endowed with image recognition capabilities.

Nevertheless, amidst this AI fervor, Apple’s position seemingly lags. Previous analysis by Kuo, a well-known analyst, underscored Apple’s conspicuous lag in the AI arena compared to its rivals.