LG AI Research Institute Demonstrates Leadership with 19 Papers Presented at Major Machine Learning Conference

151 Papers Published and 48 Patents Filed Since Launch in December 2020

LG AI Research Institute demonstrated its AI technology leadership by presenting 19 papers at NeurIPS 2023, the world's largest conference in the field of machine learning. NeurIPS started on the 10th and will be held until the 16th in New Orleans, USA. Since its launch in December 2020, LG AI Research Institute has been increasing its research achievements, presenting 3 papers at NeurIPS in 2021, 12 papers in 2022, and 19 papers this year.

At the Expo Day held on the opening day of Newlips, Lee Muntae, Head of Advanced ML Lab at LG AI Research, is demonstrating the ExaOne Universe.

On the 13th, LG AI Research Institute announced that it presented numerous papers on advanced technologies in the generative AI field, including 'AI Agents,' at this year's NeurIPS. AI Agents refer to artificial intelligence that autonomously performs decision-making by accurately recognizing human commands, similar to a personal assistant. This technology is more proactive and advanced than existing generative AI.

The research also included results showing that the 'LLM (Large Language Model) Actor,' which plays the role of executing human commands by asking questions on its own, and the 'LLM Critic,' which predicts outcomes of executions and prioritizes them, work complementarily to enable AI to make decisions that can assist humans. Applying this technology allows home appliances that assist with household tasks, such as online shopping or robot vacuum cleaners, to better understand and execute human commands in simulated environments.

Additionally, LG AI Research Institute unveiled ▲ Task-Oriented Dialogue (TOD) technology that enables AI to perform specific human commands ▲ and technology that automatically optimizes prompts, which are commands or questions input into generative AI.

LG AI Research Institute also introduced ‘Exaone Universe,’ a generative AI service for AI researchers. Lee Moon-tae, head of the Advanced Machine Learning lab, said, “Exaone Universe is a platform designed to logically reason by synthesizing evidence found in professional literature to overcome the hallucination problem of generative AI.” He added, “We are preparing to expand the service areas based on the Exaone Universe core model for AI experts into fields such as bio, chemistry, medical, pharmaceutical, patent, finance, and law.”

This content was produced with the assistance of AI translation services.