Apple researchers have advanced AI by developing an efficient method to run large language models on devices with limited memory.

This method involves storing parameters on flash memory and transferring them to DRAM as needed, thus improving efficiency.

The approach also uses sparsity in FFN layers and a hardware-inspired cost model for optimization, increasing inference speed.

Such advancements could enable more immersive visual experiences and make AI systems more accessible on consumer devices.

Apple has also developed a method for generating animated 3D avatars, which is up to 100 times faster than previous methods.

These advancements could enhance performance and personalization for iPhone users across various applications.

Apple has not publicly acknowledged the implementation of generative AI in its products or confirmed its work with Apple GPT, an internal AI chatbot.