It is too hard to resist the idea of the dominant tech company “not doing anything”. They are turning around the idea of “privacy” but not doing anything about it.
Apple’s LLM gap is real. It might not last much longer. — Joan Westenberg
The force of LLMs extends beyond consumer-facing products. Through GitHub, Copilot, etc., LLMs streamline the developer ecosystem, where Apple has traditionally excelled. LLMs can assist in coding, debugging, and even designing software. If Apple’s ecosystem lacks these tools, developers will gravitate towards more AI-integrated platforms.
So, the gap is real. That’s unarguable. But Apple’s relative silence on the LLM era does not necessarily imply inactivity; behind the scenes, the company is making strides. In collaboration with Cornell University, researchers from Apple have quietly made a significant contribution to the field of Large Language Models (LLMs) with the open-source release of “Ferret” in October.
This multimodal LLM, unique in its ability to use regions of images for queries, marks a small but potentially transformative step in AI research. Despite a quiet debut on GitHub, Ferret has started to garner attention within the AI community – albeit without the typical fanfare associated with AI product drops.
Ferret, alongside Ferret-Bench, was released on October 30, with subsequent checkpoint releases unveiled on December 14. Apple Insider commented that the development initially went unnoticed in the broader tech community, but its significance was highlighted in a report by VentureBeat. Bart De Witte, who heads an AI-in-medicine non-profit, brought attention to this “missed” release through a post on X. He lauded it as a “testament to Apple’s commitment to impactful AI research,” suggesting a deeper, more strategic investment by Apple in both AI and LLMs.
Apple doesn’t have a stellar record with open source. Why Safari is based on WebKit, bulk of the API’s are restricted or stunted. It means that it is incredibly difficult to figure out a way to “bypass” the restrictions. Apple has done nothing much for BSD; despite being a stellar operating system, it remains on the sidelines. The ARM chipsets are stellar; no doubt. However, the OS itself is stunted. They have exposed only programmatic API’s for external software without innovating on the space themselves.
Therefore, the LLM “development”, like any other “innovation” will be drip-fed to the masses. Don’t hold your breath about your new shiny iPhone to run “really-smart”; it won’t be.
RadOncNotes 