Friday, March 1, 2024

AI

Riiid's paper accepted by CVPR, one of the world's top AI conferences



Photo: From left, Kim Jin-young, Choi Seung-taek, Lee Yun-seong, and Ko Hyo-jun, researchers. of Riiid.


Riiid an AI edtech company, announced today that it has been recognized for its leadership in generative AI, a field of increasing global interest, with the acceptance of a paper at  the Computer Vision and Pattern Recognition (CVPR), one of the world's top AI conferences.


Over the years, Lloyd has consistently published papers on its latest AI-enabled research. Since 2016, a total of 17 AI technology papers have been accepted by international conferences, including the world's leading global AI publication NeurIPS, the American Association for Artificial Intelligence (AAAI), the Association for Computational Linguistics (ACL), the North American Association for Computational Linguistics (NAACL), and the European Society for Natural Language Processing (EMNLP). 

"Our ongoing research and publications were initiated and completed to solve real-world problems in order to advance our tutor products, such as the AI education platform and Santa for TOEIC," said Dr. Young-Jun Jang, CEO of Riiid. "We will continue to build and share our world-class R&D capabilities in the field of artificial intelligence education technology."  

CVPR is the world's leading conference on computer vision and pattern recognition, and the latest research presented at the conference drives innovative technological advances in artificial intelligence and computer vision.

In particular, it is the first time in Korea and the world that an education technology company's AI technology has been recognized for its practicality at CVPR. 

This paper is related to image visualization technology using AI and employed different techniques such as 'midjourney', 'dali', 'stable diffusion', etc. which became famous for generating AI images by using existing text inputs. Among the many techniques for generating images using AI, the diffusion model, which has gained the most attention in recent years, is a technique that takes a random inputs, such as noise, and iterates on it until it gradually changes to approximate the desired image, resulting in a more sophisticated and realistic image. Because a huge amount of additional data is required to control this process, Lloyd's researchers have developed a key technique to overcome the data shortage by using diffusion. 

The paper was co-authored by seven AI researchers at Lloyd's, with Hyojun Ko and Yoonseong Lee from Lloyd's AI Tents team and Jinyoung Kim from Lloyd's AI Learning team leading the research.

"We are pleased that our efforts to overcome the challenges of using AI to create educational tents have been recognized globally," said Dr. Seung-Taek Choi, who leads the development of content AI technology that generates text, images, and videos with artificial intelligence at Lloyd's. "This research is expected to reduce the cost of data collection required to develop one-to-one AI tutors to provide educational content tailored to each student."

관련기사