In addition to the recent news that it quietly announced theHigh-performance machine learning framework MLXIn October of this year, Apple and researchers from Columbia University in the United States actually announced an open source multi-modal large-scale natural language model called "Ferret".
Ferret can accurately identify image content and clearly describe image details and constituent elements. It differentiates between 70 billion and 130 billion parameter versions. Its inference capabilities are primarily based on the GRIT dataset, which accumulates over 110 million reference samples corresponding to various levels of spatial knowledge.
"Ferret" is only for research purposes and is not open for commercial use. However, it is uncertain whether Apple will use this model in the future, or use it as a basis to create artificial intelligence technology application services that can be used in products such as the iPhone.
In addition to "Ferret", Apple recently announced MLX, a high-performance machine learning framework tailored for its chip products, which can be used on iPhones or Mac models and enhances computing performance.
According to reports, Apple is currently training its Siri digital assistant service through a large-scale natural language model to enable it to have a more natural artificial intelligence interaction experience.During WWDC 2024 next yearPromulgation.
