Apple Announces MM1: The Next Family of Multimodal LLMs
Apple has unveiled MM1, a family of multimodal models capable of handling images and text.
According to the research paper, Apple’s MM1 is a family of multimodal models with up to 30B parameters. The company states that the models are also competitive with regards to performance after supervised fine-tuning on a range of benchmarks. Some of the benefits to MM1 include in-context learning, multi-image reasoning and chain-of-thought prompting.
The news comes in the midst of Apple aiming to ramp up its AI investments, with Bloomberg highlighting that the company will spend roughly US$1bn per year on the technology. The company is also reportedly in talks with Google to to bring Gemini AI to the iPhone.
MM1 could improve existing Apple products
Significantly, researchers at Apple believe they have made a breakthrough when it comes to training models with both images and text. They have suggested that the MM1 model is competitive with the initial versions of Google Gemini and are able to outperform it.
It states that these findings could help those trying to scale the power of AI by offering better performance and reliability with the MM1 models.
In addition to being able to follow instructions and reason across images, MM1 is also reportedly able to make predictions. This allows the AI to count objects, demonstrate common-sense and word knowledge and perform maths functions.
With such multimodal understanding capabilities, it is suggested that Apple could use MM1 to improve Siri, its voice assistant function. It could do this by helping it to answer questions based on images. Additionally, MM1 could work to help understand shared images and texts within iMessage and offer users more relevant response suggestions.
Having also not revealed much about how it is developing and using AI in other areas of the business, Apple has not yet released MM1 or stated what it will be used for. However the research team has revealed optimised data handling strategies following studies that examined the impact of different data types on model performance.
Harnessing AI to create real digital transformation
This development highlights that Apple has refined its focus towards developing and improving its AI efforts. It has been undergoing consistent research into its generative AI (Gen AI) capabilities, suggesting that the company’s strategy in 2024 may be evolving towards AI digital transformation.
Likewise, the tech giant reportedly cancelled its long-term electric vehicle (EV) project, recently, redistributing the efforts towards its AI division in order to better facilitate growth.
In line with these developments, during a post-earnings conference call in February 2024, Apple CEO Tim Cook said: “We view AI and machine learning as fundamental technologies, and they’re integral to virtually every product that we ship. We’re excited to share the details of our ongoing work in that space later this year.”
************
Make sure you check out the latest edition of AI Magazine and also sign up to our global conference series - Tech & AI LIVE 2024
************
AI Magazine is a BizClik brand