Apple researchers have published a study on MM1, which pertains to the process of training data in a blended manner, examining how training data of different formats and model architectures affect the efficiency of testing AI. One aspect of this research revealed that the choice of image encoding method and the image resolution used for training greatly impact the model’s efficiency more than designing various data connection components. Moreover, a model with 30 million parameters, one of MM1’s capabilities, excels in learning from the best context data, continuously supporting consecutive prompts.
The research highlights Apple’s development of an AI model that stands out in continuous prompt support. It is anticipated that Apple will disclose more details at the WWDC developer conference later this year.
Source: VentureBeat
TLDR: Apple researchers share findings on MM1 research, emphasizing the impact of image encoding methods and resolution on AI model efficiency, leading to the development of a standout AI model with continuous prompt support. Anticipated further details at WWDC conference.
Leave a Comment