Apple loses key AI leader to Meta amid talent exodus

Apple’s AI division is facing major upheaval after losing Ruoming Pang, the respected leader of its foundation models team, to Meta. Pang’s departure, along with several key researchers, has created what insiders describe as an “earthquake inside Apple” at a critical moment when the company is already struggling to catch up in the AI race.

What you should know: Pang was central to Apple’s efforts to build its own large language models and develop on-device AI capabilities for iPhones.
• He joined Apple in 2021 from Google DeepMind, a leading AI research lab, and led the team responsible for shrinking large language models (LLMs) to run efficiently on mobile devices.
• His hands-on technical contributions included developing a key open-source training tool for Apple’s AI models.
• Several of his top researchers are either leaving or exploring offers from OpenAI, Anthropic, and Meta.

The big picture: Internal tensions between Apple’s research-driven AI teams and product-focused leadership have been escalating, creating friction that may have contributed to the talent exodus.
• Pang’s team wanted to release some of Apple’s AI models as open source earlier this year to demonstrate progress and invite collaboration.
• Craig Federighi, Apple’s senior vice president of software engineering, reportedly shut down the open-source initiative, concerned it would expose performance compromises Apple made to run models on iPhones.
• Apple recently reorganized its AI efforts, moving the Siri team from longtime AI chief John Giannandrea to Federighi’s software division.

Why this matters: The talent drain comes as Apple faces mounting pressure to prove it can compete with rivals like Meta, OpenAI, and Google in generative AI.
• Bloomberg recently reported that Apple is testing outside models, including those from OpenAI and Google, to power Siri—a move that reportedly disheartened many on the internal AI team.
• While Apple announced Apple Intelligence in June, integrating ChatGPT into iPhones, the company’s own foundation models remain behind closed doors.
• Insiders say there’s still a lack of clear direction about whether Apple wants to compete head-to-head with models like GPT-4 or build more narrow, hardware-optimized tools.

What they’re saying: Apple executives have been clear about their strategic direction for AI development.
• In an interview with Tom’s Guide following WWDC 2024, Craig Federighi and Greg Joswiak, Apple’s senior vice president of worldwide marketing, “made it clear that Apple doesn’t want to make a chatbot.”

Looking ahead: Apple has hired Zhifeng Chen, a former Google engineer, to lead the foundation models team in Pang’s absence.
• Some fear Apple’s internal AI efforts could stagnate or become overly reliant on outside partners without Pang’s leadership.
• Others remain optimistic that Chen’s hiring will bring fresh momentum to the division.