EMIE

Description

This project focuses on the research and development of a new generation of compact, energy-efficient language models and AI systems using quantum-inspired technologies, particularly Tensor Networks. Its purpose is to transform the way companies access, query, and leverage their internal knowledge, enabling the use of advanced models in resource-constrained environments and significantly reducing the traditional computational and energy requirements of current LLMs.

Objective

The overall objective is to research, design, and validate quantum-inspired technologies that enable the creation of lighter, more efficient, and scalable LLMs and AI models, with reductions of up to 70% in both size and energy consumption, while maintaining their accuracy and operational capacity. The project aims to enable the deployment of specialized models in strategic industries, ensuring more sustainable and democratized access to advanced artificial intelligence.

Actions

• Investigate quantum and quantum-inspired technologies applicable to LLMs.
• Study Tensor Networks for model compression and efficiency.
• Design algorithms to reduce the size and energy consumption of LLMs.
• Analyze deployments of compressed models compared to traditional LLMs.
• Investigate the servitization of compact LLMs in resource-constrained environments.
• Validate results through Proofs of Concept in real-world use cases.