NCSoft's new AI suite is trained to streamline game production
NCSoft's new AI suite is trained to streamline game production
Despite being publicly available for less than a year, generative AI technology can already be found all around us, helping us browse the internet, taking the drudgery out of computer coding, and even improving the dialog in popular video game franchises. On Wednesday, NCSoft, the South Korean game developer and publisher behind long-running MMORPG Guild Wars, announced that it has developed four new AI large language models, dubbed VARCO, to help streamline future game development.
VARCO (“Via AI, Realize your Creativity and Originality,” if you squint just right) is both the quartet of language models the company has developed, as well as all of the products and services the company plans to build atop them. Those potential products include, “digital humans, generative AI platforms, and conversational language models,” per an NCSoft release.
The four models are VARCO the base LLM, as well as Art, Text and Human. LLM will be the first released — the Korean-language version is available August 16, while the English and bilingual iterations will arrive by the end of the month. LLM be trained with 1.3B, 6.4B and 13B parameters to start with larger versions made available later in the year.
“Our LLM is trained with datasets that are either publicly available for pretraining, collected from the Internet or internally constructed,” Jehee Lee, CRO of NCSOFT, told Engadget via email. “We are putting efforts to improve the performance of LLM and generate text that does not undermine the universal values of society.”
Subscribe to the Engadget Deals Newsletter
Great deals on consumer electronics delivered straight to your inbox, curated by Engadget’s editorial team. See latest
Please enter a valid email address
Please select a newsletter
By subscribing, you are agreeing to Engadget’s Terms and Privacy Policy.
“Bias is one of our biggest concerns in the process of constructing data,” Lee continued, “ In order to build high-quality datasets, NC utilizes a pipeline where collected data are analyzed and evaluated in various aspects (e.g. format, content, ethics) and are refined based on our own criteria.”
Larger-parameter versions of LLM will arrive later in the year, allowing for more nuanced and complete responses from the system. “As the size of the model increases, the performance increases,” Lee notes, “but the operating cost also increases accordingly.” As such ensuring that models are developed to be both computationally powerful and computationally lightweight is of tantamount importance, he continued.
“NC has accumulated technologies for model weight reduction and optimization as we have been applying 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 experiences, we plan to develop and sequentially release high-performance lightweight models specialized for various individual tasks in the future.”
The three additional services will be based upon that foundational model. VARCO Art, the text-to-image generator, will reportedly be capable of closely matching the work of specific artists (with their permission of course). “With dozens of illustrations worked consistently by an artist, generative models can be trained to reveal enough of a particular style, such as coloring, touches, and outlines,” Lee said. “ Thus far, artists or experts can spot the difference, but it is not easy for ordinary people to distinguish if it is generated by AI or not.”
Additionally, VARCO Text does just that, generating and managing the core settings of a game including the plot scenarios and character worldviews, while VARCO Human, “an integrated tool for creating, editing, and managing digital humans,” per the release. Art, Text, and Human can all be used and managed within the company’s VACRO Studio suite, which will be available in 2024.
The VARCO models will initially be used to help streamline game development efforts, much in the same way as Ubisoft’s Ghostwriter. However, VARCO sets itself apart, Lee explains, in 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 for the planning, development, and operation of games.,” Lee said. “The worldview of the game, character name, type, and attributes can be created and edited. Also, it creates conversations for specific situations and quests tailored to regions,” as well as accompanying images.
And like Ghostwriter, VARCO cannot operate in isolation. “We need final human judgment and additional work in order to produce sophisticated results,” Lee said. He argues that, ”generative AI technology will contribute to further increasing the value of human labor,” by distinguishing between the two.
In allowing AI to handle the low level repetitive tasks that perpetually slow game development, human designers will be freed to “focus on more complex tasks,” Lee said. “The results of generative AI will not be the final product, but rather it will serve to inspire humans and help them quickly reach higher goals with the help of AI’s emergent and large-scale creation capabilities.”
But games are only the beginning for NC’s AI aspirations. “we want to create new high value-added businesses by entering different fields beyond games,” Less said. “I think we can advance with our AI in any industrial field rather than just one field. NC plans to pay attention to new growth fields that we can expect greater potential and synergy when combined with AI, such as fashion, life, health, mobility, bio, robotics, and content.”