Researchers at Meta AI have released a new approach to designing efficient language models for smartphones and other resource-constrained devices called MobileLLM.
The research team consists of members from Meta Reality Labs, PyTorch, and Meta AI Research (FAIR), who focus on optimizing models with fewer than 1 billion parameters. This is only a small fraction of the size of models like GPT-4, which is estimated to have over a trillion parameters.
Key innovations of MobileLLM include:
1. Prioritizing model depth over width
2. Implementing embedding sharing and grouped query attention
3. Adopting a novel instant block-level weight sharing technique
These design choices enable MobileLLM to outperform previous models of similar size by 2.7% to 4.3% on common benchmark tasks. While these single-digit improvements may seem small, they represent meaningful progress in the highly competitive field of language model development.
It is worth noting that the 350 million parameter version of MobileLLM achieves comparable accuracy to the 7 billion parameter LLaMA-2 model on certain API call tasks. This suggests that for certain specific applications, more compact models may offer similar functionality while using significantly fewer computational resources.
The development of MobileLLM aligns with the growing interest in more efficient AI models. As the development of super-sized language models shows signs of slowing down, researchers are increasingly exploring the potential of more compact and professionally designed models. Despite its name containing "LLM" (large language model), MobileLLM's focus on efficiency and device-side deployment places it in the same category as what some researchers refer to as small language models.
While MobileLLM is not currently available to the public, Meta has open-sourced the pre-training code, allowing other researchers to build upon it. As this technology evolves, it may enable more advanced AI capabilities on personal devices, although the specific timeline and capabilities are yet to be determined.
The development of MobileLLM marks an important step towards making advanced AI more accessible and sustainable. It challenges the notion that effective language models must be massive and may pave the way for new possibilities in AI applications on personal devices.
