LG Uplus and OptAI Partner on AI Token Optimization Technology
South Korean telecommunications firm LG Uplus has partnered with AI model optimization specialist OptAI to develop technology aimed at enhancing AI processing efficiency. The collaboration seeks to reduce graphics processing unit (GPU) and power consumption.
LG Uplus, a South Korean telecommunications company, announced on October 2 its partnership with OptAI, an AI model optimization specialist. The collaboration aims to develop token optimization technology designed to decrease graphics processing unit (GPU) and power usage while improving overall AI processing efficiency.
Tokens are the fundamental units of data that artificial intelligence systems process when handling user queries and generating responses. OptAI will focus on developing lighter and faster AI model execution tools, while LG Uplus will utilize its AI service operations for testing and applying the new technology.
In prior work involving a small language model based on LG AI Research's Exaone model family, the companies reportedly enabled the model to run on smartphone neural processing units (NPUs). This achieved performance comparable to CPU-based methods, with an 82% reduction in model size and a 78% decrease in power consumption.
The companies are now extending this development to server GPU environments. LG Uplus stated that the technology has the potential to increase the number of tokens processed on a single GPU by up to four times compared to existing methods. This advancement could significantly impact the efficiency and scalability of AI operations in data centers.