Apple's Ferret: A New Era in Open Source AI Beyond GPT-4
Written on
Chapter 1: Introduction to Apple's Ferret
As major tech companies dive deep into the world of generative AI, Apple is making its entry with the introduction of Ferret, its inaugural open-source multimodal language model (LLM). While details regarding AI-powered features for iPhone users remain undisclosed, there's palpable anticipation for enhancements that could significantly improve Siri's capabilities.
To facilitate this, Apple is exploring the potential of a vast database sourced from news publishers, a tactic that could accelerate its AI development. This innovative model is designed to comprehend and generate not just text, but also images, sounds, and videos.
With competitors like Google Assistant and Amazon Alexa already integrating similar advancements, Apple's move into generative AI comes at a crucial time. Microsoft has incorporated GPT-4-based AI Copilot into Windows, while Samsung is preparing to introduce its own rival to ChatGPT.
To learn more about the competitive landscape of AI, explore the following video:
Chapter 2: Ferret’s Unique Features
Developed through a collaboration between Apple’s Zhe Gan, a prominent AI researcher, and his team at Columbia University, Ferret has been in development since its unveiling in October. This model stands out due to its remarkable ability to analyze and provide descriptions of small segments of images, a domain where it reportedly excels beyond OpenAI’s GPT-4.
Currently in its infancy, Apple envisions making Ferret compatible with smartphones. Given that OpenAI’s GPT-4 operates with over a trillion parameters, while smartphones typically manage around 10 billion, this goal poses a significant challenge.
To train Ferret, Apple employed eight Nvidia A100 GPUs, renowned for their impressive performance and high cost. These GPUs, equipped with 80 GB of HBM2e RAM, represent a leading choice in the generative AI technology market. The A100 is capable of achieving 312 TeraFLOPS utilizing Tensor Float 32 precision, establishing itself as a foundational element in AI computation.
In addition, Apple researchers have demonstrated the potential to utilize a smartphone’s built-in flash memory alongside RAM to operate larger models than is typically feasible. This advancement suggests that devices such as the iPhone 16 may soon come with a greatly enhanced AI assistant.
Section 2.1: Leveraging Media Partnerships
Apple's venture into generative AI could signify a transformative shift in its product offerings, particularly in enhancing Siri. Recent reports indicate that Apple is negotiating with news publishers to secure deals that would allow it to use their content for AI training, potentially positioning itself as a competitor to ChatGPT.
According to the New York Times, Apple is prepared to invest at least $50 million to establish multiyear agreements with various media outlets. This financial commitment would enable Apple to access a wealth of articles and archives, creating a rich dataset for training its generative AI models. Notable publishers like Condé Nast and IAC, which are associated with well-known titles such as Vogue and The New Yorker, have been mentioned in these discussions.
Apple's approach, as articulated by CEO Tim Cook, emphasizes responsible and significant investment, with advancements in products expected to unfold gradually. Should Apple secure these licenses, the high-quality content from these publishers could prove pivotal in helping the company catch up with its competitors in the generative AI realm while also safeguarding against allegations of unauthorized content usage.
To gain insights into Apple's strategic moves, watch this related video:
Section 2.2: The Broader AI Landscape
Apple’s entry into open-source AI with Ferret represents a crucial milestone in the tech industry. By merging formidable computational capabilities with state-of-the-art AI functionalities, Ferret not only challenges existing models like GPT-4 but also exemplifies Apple's ambitious vision within the AI domain.
These proactive steps could herald a new era in which technology giants harness AI to enrich user experiences, making advanced AI tools more accessible and seamlessly integrated into everyday life.