[AI Revolution](81) Krafton Creating AI for Everyone to Enjoy

The AI development philosophy of game company Krafton is ‘fun.’ Everyone should be able to enjoy AI, and AI should bring innovative fun to games. Krafton is rapidly moving toward this goal. Soon, the moment will come when players can enjoy games together by having conversations with AI at the level of actual human interaction.

Krafton has been operating an AI research organization called the ‘Deep Learning Headquarters’ since 2021. CEO Kim Chang-han has been directly involved from the beginning, leading the business with significant investment in AI research and application. This year, the organization was expanded and reorganized to strengthen AI research. The Deep Learning Headquarters develops various AI technologies and applies them to games. These include a TTS model that reads sentences with delicate emotions and voice, a chatbot that maintains long conversations with characters (personas) and interacts like a human, and technology that converts human faces into various styles to create 3D avatars.

Krafton's flagship title 'PUBG: Battlegrounds'

Krafton also presented a blueprint for playing games as a team with AI. The technology that makes this possible is called ‘Virtual Friend.’ It incorporates natural language processing and language models at the level of ChatGPT, enabling conversations at the level of real humans. Additionally, STT (speech-to-text technology) and TTS help facilitate smooth communication through various means such as voice or text. The plan also includes integrating ‘gameplay AI’ technology so that the AI can understand and solve the game on its own.

‘Virtual Friend’ is expected to be applied to games as early as next year. For example, in the flagship title ‘PUBG: Battlegrounds,’ it will become possible to fight alongside AI teammates instead of human players. In such a scenario, players will be able to communicate with AI in real time using commands like, “Anna, check if there are enemies at the pinged location. Winnie, hand me the Eotteop (name of a sniper rifle).” This will make it important who can give more efficient commands to the AI, adding a new element of fun to the game.

Regarding this, Krafton CEO Kim Chang-han explained, “Applying AI deep learning can break the traditional gameplay style. In games where four players play together, even if a gamer plays alone, AI can create an environment where it communicates in real time through text and conversation as if playing with three other friends.”

Krafton Virtual Human Production Process Scene

Krafton is striving to make AI a source of fun in games. To this end, it established a new independent studio, Relu Games, in June. The goal is to apply deep learning game production experience and ideas to new projects. The first project, launched in July, was ‘FOONDA: AI Puzzle.’ Deep learning generated puzzle stages that provided users with a highly personalized puzzle experience, receiving positive reviews. Plans are underway to pursue projects that explore fun in games using voice recognition.

Krafton regards AI as a next-generation technology that will determine the future industry and has proactively introduced it across all company operations. The aim is to improve work productivity and satisfaction by enhancing all employees’ understanding and utilization of AI. Since March this year, the company has supported all fees for deep learning solutions such as ChatGPT, Diffusion, and Copilot for all employees. Additionally, the Deep Learning Headquarters distributes AI tools and manuals developed in-house to help employees quickly apply ‘deep learning technology’ to various work situations. Currently, over 90% of all employees use AI in their work.

Krafton does not focus solely on AI development. It also makes efforts to ensure AI is healthy. To this end, it established the ‘Krafton AI Ethics Committee’ (hereafter AI Ethics Committee) in April. This committee aims to prevent ethical issues that may arise during AI research or development and to proactively prepare for them. The AI Ethics Committee is composed of volunteers from various departments (legal team, data team, privacy team, etc.) to continuously deliberate and discuss AI ethical issues. They hold regular monthly roundtable meetings to establish AI ethical guidelines and codes and discuss various AI ethical issues and cases.