Apple works in a surprising way with AI and publishes its progress, now they present a new model called MM1 that will allow you to perform different tasks
Through a body of work from Apple researchers, they have published progress on a new method for training large language models by seamlessly integrating text and information. According to the published research article, a new approach is presented to create smarter and more flexible artificial intelligence systems. The new model they called “MM1” is a new standard in AI’s ability to perform certain tasks such as image captioning, visual answers to questions, and even natural visualization. This follows a new acquisition by Apple in terms of AI itself.
It’s no longer an open secret, Apple is fully working on AI
First published by VentureBeat, this research is a “careful combination of different types of training data and model architectures” that can lead to truly cutting-edge performance in AI benchmarks.
In this work, we analyze the construction of large, high-performance multimodal language models (MLLMs). We explore the importance of various architecture components and data options.
MM1 will allow you to count objects, make references, use common sense and perform basic math functions
The MM1 model is a “family of multimodal models” of the latest generation with great attractive properties. Among them, “improved learning in context” and “multiple image reasoning” stand out. The new model family offers incredible and competitive performance in a wide range of references.
News that is difficult to digest on paper but which will soon be seen in practice
It’s difficult to fully understand what Apple researchers are working on, but we can summarize it because they combine different types of data and model architectures. The AI in question will have to generate language based on a combination of signals that can be both visual and linguistic. It can work for interpreting images or videos, answering questions about different elements visible on the screen, etc.
MM1 will follow instructions and process reasoning through pictures
Those from Cupertino do not want to be left behind in terms of competition in artificial intelligence, a trend that has been growing since last year and is materializing in competing devices. Apple has just spoken with Google to obtain Gemini licenses for new iOS 18 features that can be used on the iPhone.
