Introduction of NC AI
We'll give a short introduction of NC AI before describing the Applied AI Lab's research areas and organizational culture.
The NC AI organization was established in 2011 with a small team of specialized researchers and engineers in the gaming industry. Today, we have grown to include over 200 exceptional researchers and engineers across multiple business sectors, not just the gaming industry.
We aim to create artificial intelligence working with people that allows us to realize the true value of human beings by maximizing work efficiency.
Through researching various AI areas like computer vision, natural language processing, speech recognition, recommender system, and so on, NC AI is pursuing to develop of AI-driven systems such as a 'digital human,' 'machine translation system,' 'AI-powered financial advisor,' 'game AI', and etc.
Who are we? What are we heading for?
In new business areas such as finance and media, we are creating AI-based technologies that can provide, predict, interpret, and optimize useful functions.
At companies like Amazon and Meta, applied data scientists and research scientists are actively leveraging AI technologies to overcome business challenges.
In general, colleges and industries that have big R&D organizations such as Amazon, Meta, Google and etc. publish their papers in top journals or conferences in order to get an evaluation of their research results.
Our lab is working to develop AI technologies that can be applied to real-world services to enhance user convenience and satisfaction by optimizing existing operations, as well as publishing research papers.
In general, researchers and engineers frequently encounter the cases in which the excellent performances in papers is hard to show practically. We also had a lot of similar experiences.
We think that it is difficult to achieve state-of-the-art performance in live services as in the paper because the most of technologies in the research papers were verified only based on the refined and static benchmarking data with a fixed period as experimental data.
In live service environments, there are many unexpected variables and various factors are confounded, and we have to face challenges such as data sparsity, imbalance, and missing training data.
Applied AI Lab conducts plenty of data analysis to discover the challenges and is seeking an appropriate way to solve the problems we found so that we are trying to develop reliable high-performing technologies in the live services.
What are we focusing on?
The Applied AI Lab consists of 1) researchers who study and develop AI technologies and 2) engineers who develop and provide AI technologies to live services.
Researchers belong to one of the three teams: 1) the curation team, 2) the anomaly detection team, and 3) the sequence modeling team.
The curation team develops recommendation algorithms for delivering selected contents of interest based on user's preference and history, and also devise automatic user profiling techniques for analyzing users' interests, characteristics and behavioral patterns.
The anomaly detection team defines "anomalies" that indicate an unexpected patterns from data and designs a model that detects new defined anomalies. Currently, this team is focusing on the graph neural networks to consider the relationships between anomalies and normals.
The sequence modeling team analyzes causal relationships between the various sequential data including time series data and rarely occurred discrete data (e.g. event). Based on the analyses, researchers in this team develops predictive models using event sequences.
The engineers belong to the data platform team. They build infrastructures, develop data pipelines, and provide AI models devised by researchers to live services stably.
Among the many research results, here are some recommendation-related examples we developed.
In NC's AI-generated baseball app for Korean league, called 'PAIGE', we have recommended posts, news, video contents to users using various algorithms for each content features. In Bufftoon, NC's webtoon service, we suggest webtoons of which thumbnails with similar drawing style inspired by image style transfer technology.
In the media domain, NC made a technical agreement with Yonhap News and has been doing R&D projects. When you click any news on the PC web and mobile app of Yonhap News, you may find "AI recommendation news" in the right wing of PC web and right under the news in mobile app unlike the past when we can see "Popular news" or "Hot news".
Only "Popular news" or "Hot news" used to be shown in the recommendation areas in the past, but now online A/B Test has been running for comparing popular news lists to our recommendation models.
To do this, we have developed a number of news recommendation models, and we are choosing the best recommendation model that provides a satisfactory user experience by running online A/B tests.
We will explain the more details and examples of recommendation models, anomaly detection models, and sequence modeling techniques in the next time.
How are we working together?
NC has a horizontal corporate culture - calling each other's name with "Nim (님)" regardless of his/her job title. Based on such culture, with a more open and free work atmosphere, we discuss research challenges and questions.
We hold two regular technology meetings on a regular schedule. During these meetings, we discuss a pre-determined topic, review papers from top-tier journals and conferences, and stay current with new technology trends.
In our weekly work-sharing meeting, we take turns presenting our current projects to the group. Through listening to our colleagues' updates, we often gain inspiration for applying the technology discussed in one project to another.
While individual efforts are essential for achieving significant research results, we believe that collaboration among researchers is equally important. Our research culture fosters cooperation, making us stronger data scientists and machine learning researchers.
Confucius, he said, 知之者不如好之者 好之者不如樂之者 (지지자불여호지자 호지자불여락지자, 'They who know the truth are not equal to those who love it, and they who love it are not equal to those who delight in it.')
Like the above old saying, how about joining us and enjoying research works with us?
