解读丨苹果为何「死磕」自研 AI ?
TL;DR - Apple is reportedly pursuing a jointly trained China-focused LLM with Alibaba while retaining third-party models. The strategy aims to preserve control over system-level agents, privacy, user experience, and the commercial value of Apple’s platform.
- Apple could use its own model for intent understanding, task planning, permissions, and cross-app orchestration, while outsourcing general knowledge and generation.
- Mature training methods, open-source ecosystems, and efficient MoE architectures have made usable foundation models faster and less costly to build.
- A proprietary agent would keep sensitive data and privileged device operations within Apple-controlled boundaries.
- Controlling the primary AI interface is strategically important as agents increasingly bypass apps, search, advertising, and other traditional platform entry points.