NCSoft Launches AI Suite to Accelerate Game Development
NCSoft, the South Korean game developer and publisher responsible for the popular MMORPG Guild Wars, has recently unveiled four new AI large language models called VARCO. These models aim to enhance game development processes and are a testament to the rapid integration of generative AI technology in various aspects of our lives, such as internet browsing, computer coding, and even dialogue improvement in popular video game franchises.
VARCO (“Via AI, Realize your Creativity and Originality”, if you civilize it right) is both the quartet of language models developed by the company and all the products and services that the company plans to build on top of them. These potential products include “digital humans, generative AI platforms and conversational language models,” according to the NCSoft release.
The four models are VARCO, basic LLM, and Art, Text and Human. LLM will be released first – the Korean version will be available on August 16, while the English and bilingual iterations will arrive by the end of the month. The LLM needs to be trained on the 1.3B, 6.4B and 13B parameters to start with the larger versions that will be released later this year.
“LLM is trained on datasets that are either publicly available for pre-training, collected from the Internet, or built internally,” Jehee Lee, CRO at NCSOFT told ReturnByte via email. “We strive to improve the performance of the LLM and create a text that does not undermine the universal values of society.”
“Bias is one of our biggest concerns in the data generation process,” Lee continued. ” To build high-quality datasets, NC uses a pipeline where the collected data is analyzed and evaluated from different aspects (e.g. form, content, ethics) and refined based on our own criteria.”
Larger parametric versions of the LLM will arrive later this year, allowing for more nuanced and complete responses from the system. “As the model size increases, the performance improves,” Lee notes, “but the operating costs also increase accordingly.” Therefore, it is very important to ensure that the models are developed both computationally efficient and computationally light, he continued.
“NC has accumulated techniques to reduce and optimize model weight as we have applied real-time machine translation and NLP-based technologies to our games that require large-scale traffic processing in real-time,” Lee explained. “Based on our experience, we intend to develop and publish one after another efficient light models specialized for various individual tasks.”
Three additional services are based on this basic model. VARCO Art, a text-to-image generator, is said to be able to summarize the work of certain artists (with their permission, of course). “With dozens of illustrations consistently worked on by an artist, generative models can be trained to sufficiently reveal a particular style, such as coloring, touches and outlines,” Lee said. “So far, artists or experts can notice the difference, but it’s not easy for ordinary people to distinguish whether it’s produced by AI or not.”
What’s more, VARCO Text does just that by creating and managing core game settings, including plot scenarios and character worldviews, while VARCO Human, “an integrated tool for creating, editing and managing digital humans,” according to the release. Art, text and people can be used and managed in the company’s VACRO Studio suite, which will be available in 2024.
VARCO models are initially used to enhance game development, similar to Ubisoft’s Ghostwriter. However, what makes VARCO stand out, Lee explains, is that it is “a special vertical AI that can directly solve specific pain points regardless of industry and domain.”
“The generative model can be used to design, develop and operate games,” Lee said. “The game’s worldview, character name, type and attributes can be created and modified. It also creates conversations for adventures tailored to specific situations and areas, as well as related images.
And like Ghostwriter, VARCO cannot operate in isolation. “We need final human judgment and additional work to produce sophisticated results,” Lee said. He argues that “generative AI technology will further add value to human labor” by separating the two.
When AI can handle low-level repetitive tasks that constantly slow down game development, human designers are freed up to “focus on more complex tasks,” Lee said. “Generative AI results are not a finished product, but rather inspire people and help them quickly achieve higher goals through AI’s emerging and large-scale creative capabilities.”
But games are just the beginning of NC’s AI endeavors. “We want to create new high-value-added businesses by moving into different industries beyond gaming,” Less said. “I believe we can make progress with our AI in any industry, not just one. NC plans to pay attention to new growth areas where we can expect greater potential and synergy combined with AI, such as fashion, life, health, mobility, bio, robotics and content.”
