AI Generated Image
Apple looks to shrink AI models for iPhones
Baton Rouge Business Report • • technology
Key Points:
- Apple is in early talks with Silicon Valley startup PrismML about technology that significantly compresses large AI models to run directly on iPhones, enhancing on-device AI capabilities.
- PrismML’s compression method reduced Alibaba’s 27 billion parameter Qwen model from 54 GB to under 4 GB, improving speed, energy efficiency, and memory use with minimal performance loss.
- Apple is assessing the technology to improve Siri and enable faster, more private AI processing without relying heavily on cloud computing or internet connectivity.
- Experts highlight potential applications in computational photography, video generation, and health tools, but emphasize the need for extensive real-world testing to confirm battery life and reliability.
- If successful, this innovation could advance Apple’s AI strategy by shifting more AI workloads to devices while maintaining privacy, though it may not decrease overall demand for processing hardware.
