#Meta invests in AI models for mobile devices – Computerworld #technology

Meta bets on AI models for mobile devices – Computerworld

Meta, the parent company of Facebook, has been working on developing a new small language model (SLM) that is compatible with mobile devices. This SLM aims to run on-device applications while reducing energy consumption during model inferencing tasks. Large language models (LLMs) have a significant number of parameters, with models like Mistral-22B having 22 billion parameters and GPT-4 having 1.76 trillion parameters. In contrast, smaller language models like Microsoft’s Phi-3 family have fewer parameters, starting from 3.8 billion.

Meta researchers believe that effective SLMs with less than a billion parameters can be developed, enabling the adoption of generative AI on mobile devices with limited computing infrastructure. They conducted experiments with models having 125 million and 350 million parameters and found that smaller models prioritizing depth over width enhance model performance. The researchers emphasized the importance of model architecture for sub-billion scale LLMs.

The MobileLLM models with 125 and 350 million parameters were as effective as large language models like Llama 2 in handling tasks like chat and API calling. While MobileLLM is not yet available for public use in Meta’s products, the researchers have made the code and data for the experiment available along with the paper. This research highlights the potential of small language models for common on-device use cases and the importance of model architecture in determining model quality.

Source link

Source link:

What do you think?

Leave a Reply

GIPHY App Key not set. Please check settings

3D GENERATION from a single image

The evolution of technology: the rise of 3D generation #3DGeneration

Insights 2024 | Attitudes toward AI

Attitudes toward AI in 2024: Insights and Predictions #AIInsights