NAVER Cloud Has 61 Papers Accepted at Global AI Conferences, Expanding its Influence in Hyperscale AI Research
NAVER Cloud Has 61 Papers Accepted at Global AI Conferences, Expanding its Influence in Hyperscale AI Research
NAVER Cloud Has 61 Papers Accepted at Global AI Conferences, Expanding its Influence in Hyperscale AI Research
- NAVER’s AI organizations exhibit technological synergy following successful integration earlier this year
- NAVER presented significant research papers in various areas related to hyperscale AI, including multimodal AI, model compression, and AI ethics
- “NAVER will strengthen collaboration with prestigious research institutions such as SNU, KAIST and NYU to further enhance our leadership in hyperscale AI”
June 13, 2023
NAVER Cloud (CEO Kim Yoo-won) achieved remarkable success in the first half of this year with 61 papers accepted at renowned global artificial intelligence (AI) conferences. This accomplishment, comparable to last year’s publication of 100 papers, underscores the technological synergy within the newly integrated NAVER’s AI organization including CLOVA and Papago.
In particular, NAVER Cloud is expanding its technological influence in hyperscale AI, which has become a highly competitive field for major tech companies like Microsoft and Google. NAVER Cloud’s research papers on hyperscale AI have gone beyond conference acceptance and were selected in the categories of “Best Paper” and “Notable Top 25%.” Following the publication of the HyperCLOVA paper in the 2021 Conference on Empirical Methods in Natural Language Processing (EMNLP), the world’s most prestigious natural language processing conference, NAVER and NAVER Cloud had a total of 20 hyperscale AI–related papers approved to be published in global conferences.
In the first half of this year, NAVER Cloud presented important research papers across various domains, encompassing the development of the hyperscale AI technology ecosystem, its implementation in services, and the creation of social values. Most notably, the research that came up with an evaluation metric to measure the performances of image generative models* was ranked as one of the “Notable Top 25%” at the International Conference on Learning Representations (ICLR) 2023, the world’s top conference in the field of machine learning. NAVER Cloud introduced a new metric to evaluate the diversity and creativity of hyperscale AI-generated images, which is expected to lay the foundation for developing more advanced hyperscale image generative models.
* Rarity Score: A New Metric to Evaluate the Uncommonness of Synthesized Images (https://arxiv.org/abs/2206.08549)
A research on AI compression, which has gained importance as hyperscale AI is being introduced to various services and commercialized beyond research, was also selected at the International Conference on Machine Learning (ICML) 2023. The study suggested a new quantization scheme that could reduce the operation cost of hyperscale AI,** which will be incorporated into NAVER’s hyperscale AI model HyperCLOVA X to enhance efficiency. NAVER Cloud is already using its own compression technology to optimize key technologies and services such as HyperCLOVA and CLOVA Note.
** FlexRound: Learnable Rounding based on Element-wise Division for Post-Training Quantization (https://arxiv.org/abs/2306.00317)
Furthermore, NAVER has solidified its leadership in hyperscale AI ethics, presenting another research in the area. Having led discussions on the subject by hosting a workshop on hyperscale AI ethics at a major conference for fairness in computer science last year, NAVER published research on a dataset that can mitigate AI’s bias in responding to sensitive questions and the protocol to implement the dataset*** at the 61st Annual Meeting of the Association for Computational Linguistics (ACL) 2023. This research resulted from a year-long collaboration and research at “AI Ethics Forum” by experts from diverse fields, including AI, social sciences, and law. In addition, NAVER’s interview-based research**** on the potential social value created by conversational AI leveraging large language models explored through the case of CLOVA CareCall, NAVER’s AI call agent system for elderly people living alone, won “Best Paper” at the Association for Computing Machinery (ACM) Conference on Human Factors in Computing Systems (CHI) 2023.
*** SQuARe: A Large-Scale Dataset of Sensitive Questions and Acceptable Responses Created Through Human-Machine Collaboration (https://arxiv.org/abs/2305.17696)
**** Understanding the Benefits and Challenges of Deploying Conversational AI Leveraging Large Language Models for Public Health Intervention
NAVER Cloud AI Lab leader Yun Sang-doo said, “We presented meaningful research at various AI conferences on machine learning, natural language processing, voice/speech, and computer vision in the first half, especially in the emerging field of hyperscale AI. We will continue to collaborate with globally renowned research institutions such as Seoul National University, Korea Advanced Institute of Science & Technology(KAIST), and New York University to advance hyperscale AI technology, including HyperCLOVA X, and work together with the University of Tübingen and Washington University to build reliable hyperscale AI models.” <END>