A potential deal would let users choose the AI model they want to power their Apple Intelligence experiences
Apple is reportedly in talks with Meta about integrating its generative AI models into Apple Intelligence software, after recently striking a deal with OpenAI.
Announced at the recent Worldwide Developer Conference, Apple Intelligence is the iPhone maker’s software providing generative AI features to Apple devices. Leveraging OpenAI technology, Apple Intelligence allows users to create and rewrite content across Apple and third-party apps.
According to the Wall Street Journal, the two companies held talks that would potentially see Meta models like Llama 3 added to Apple Intelligence.
A potential deal would let users choose the AI model they want to power their Apple Intelligence experiences, similar to how the AI search app Perplexity lets users switch up the underlying model.
Lian Jye Su, chief analyst of applied intelligence at Omdia, said the partnership could provide a “richer user experience for Apple users when using Meta’s apps on Apple products.”
Reports suggest the discussions have focused solely on Apple Intelligence. However, Omdia's Su questioned whether the partnership would extend to mixed reality devices, given that the two companies are direct rivals in that space.
“Meta has been investing in spatial computing and embodied AI, which can also benefit Apple’s premium XR products,” Su said.
The first wave of Apple Intelligence will not extend to Apple’s Vision Pro headset despite it housing the powerful M2 chip.
Potential candidates for Meta’s reported Apple integration include Llama 3 for language processing and Emu for image creation. Meta is also working on Chameleon, a model that processes text and visuals, and JASCO for music generation but these models remain in research stages, however.
The Llama series, Meta's open-source foundation models, are the most likely contenders. However, the smallest current version, Llama 3 8B, may exceed the processing capacity of some Apple devices.
Unlike traditional generative AI experiences which are run in the cloud, Apple wants to run them on devices.
To support this, Meta could develop a smaller version for on-device use. Its researchers are currently working on a Llama model on the other end of the size scale — a giant 400 billion parameter version.
Alexander Harrowell, Omdia's principal analyst for advanced computing, said if Apple were to use any of the Llama models, they would need to pay as hyperscaler-level companies like Microsoft, Amazon and Google are required to pay for usage.
Harrowell said that while Apple is not a hyperscaler, it would likely need to seek permission if charging for the use of the models.
In addition to Meta, Apple held talks with Google, as well as Claude developer Anthropic and Perplexity about integrating their generative AI technologies into Apple Intelligence, according to the Journal.
Apple's partnership with OpenAI is expected to increase demands on OpenAI's infrastructure. This week, the Microsoft-backed company acquired the database startup Rockset to improve ChatGPT response times through real-time indexing.
By partnering with multiple AI providers, Apple could avoid relying on a single service that might struggle with high user demand. ChatGPT has experienced several outages in recent months. OpenAI CEO Sam Altman has repeatedly cited the need for more infrastructure to support growing workloads.
Read more about:
ChatGPT / Generative AIYou May Also Like
