On Wednesday, Meta Connect 2024 occurred – the company’s developer conference. As part of the event, social media giant Meta has unveiled a few new artificial intelligence features and wearable devices. Apart from this, Meta reportedly also announced a collaboration with the tech firm Arm to be building special small language models. These AI models allegedly power smartphones and other devices, bringing even fresher ways of using them. The logic here is for users to have on-device and edge computing options so as not to weigh on AI inference and thus keep it fast.
Meta and Arm are said to be developing AI models capable of executing more complex tasks on devices. For example, AI may well be your device’s virtual assistant and can dial a number or click a photograph. Not far-fetched at all, really, as of today AI tools can already do a multitude of jobs, from editing photographs and drafting emails.
But the most important difference is that the users have to interface or type in specific commands for AI to perform all these tasks. The couple reportedly hailed this aspect at the Meta event, stating how they wanted to eliminate this and make AI models more intuitive and responsive.
The one way to achieve this is by bringing the servers pretty close to the devices, which is what research institutions and big companies do. As the publication details, this is known as edge computing. According to Ragavan Srinivasan, VP of product management for generative AI at Meta, “a good way of tapping into that opportunity is how to build the new AI models”.
For this, the AI models will have to be smaller in size. While Meta has developed LLMs that are as large as 90 billion parameters, they are not something that is suitable for smaller devices or faster processing. The Llama 3.2 1B and 3B models are believed to be ideal for this.
The other factor here is that AI models will also need to be fitted with newer capabilities beyond simple text generation and computer vision. It’s where Arm fits into this. According to the report, the tech giant is deeply collaborating with Meta to create processor-optimised AI models that can adapt to the workflows of devices such as smartphones, tablets, and even laptops. So far, nothing has been shared about SLMs.
Bhupendra Singh Chundawat is a seasoned technology journalist with over 22 years of experience in the media industry. He specializes in covering the global technology landscape, with a deep focus on manufacturing trends and the geopolitical impact on tech companies. Currently serving as the Editor at Udaipur Kiran, his insights are shaped by decades of hands-on reporting and editorial leadership in the fast-evolving world of technology.