We had the opportunity to interview Jun Shimoda, a Senior Gaming Solution Architect from Microsoft’s Worldwide Gaming Vertical Team, during LEVEL UP KL. With his extensive background in the gaming industry, including roles at Epic Games and AWS, Shimoda shared valuable insights into the current state and future of AI in gaming.

AI in Game Development: Balancing Innovation and IP Protection

One of the key topics Shimoda discussed was the integration of AI in game development and operations. He emphasized that while many game developers are eager to harness AI’s potential, there are concerns about generative AI, particularly regarding copyright issues.

“Many game developers are interested in using AI for both game development and operations. However, some gaming companies are concerned about the risks of generative AI, particularly regarding copyright issues,” Shimoda explained. “That’s why we’re working closely with them to ensure their IP and security are protected, using Azure solutions to safeguard their assets.”

To address these concerns, Microsoft is collaborating with gaming companies to implement tailored AI solutions that protect intellectual property. Shimoda highlighted an interesting approach, particularly for fantasy titles:

“Sometimes, we introduce a small language model, which has limited intelligence compared to a large language model. A large language model has a vast knowledge base from the internet, but not all game developers want that. For example, in fantasy games with a medieval world view, developers want to maintain the integrity of their world. They don’t want an NPC suddenly referencing real-world topics. So, in those cases, we use a smaller language model that fits the specific fantasy setting, ensuring it aligns with the game’s narrative.”

AI-Powered Quality Assurance and Localization

The interview also touched on the use of AI in game debugging and quality assurance. Shimoda revealed that some QA companies are now using advanced AI systems to create AI agents capable of playing games and identifying issues like collision detection problems.

“Some QA companies use large language models, such as OpenAI, or other advanced AI systems to create AI agents that can play games and identify issues like collision detection problems,” Shimoda said. “Previously, QA companies would hire human testers to manually identify collision errors, which is costly. Now, they are developing AI agents that can detect these issues automatically, significantly reducing costs.”

In the realm of localization, AI is making significant strides. Shimoda explained:

“Another area is LQA (Localization Quality Assurance), where companies use AI to translate game scenarios into multiple languages. While the quality of AI translations isn’t perfect yet, it speeds up the process, allowing developers to quickly produce drafts that can later be fine-tuned by human translators.”

He also highlighted the use of AI in image processing: “AI is also being used to handle images, like OCR (Optical Character Recognition), to detect text within images. For example, some games have lots of on-screen text, and sometimes there are errors. AI with OCR capabilities can quickly find and correct these mistakes.”

Addressing Copyright Concerns in AI Implementation

When asked about the widespread concern regarding copyright issues in AI use, Jun Shimoda acknowledged the cautious approach many companies are taking:

“Many companies are concerned about it. Executives and their lawyers recommend them to use AI for internal use only and don’t use generative AI for production use. But we are discussing with them to use a smaller language model with their IP. For example, Microsoft has many pictures of Minecraft, and small language models with their own Minecraft pictures to generate Minecraft-like content with their data only—that’s safe. That is the solution for that IP.”

The Future of AI in Gaming

Looking to the future, Shimoda believes that AI will play an increasingly significant role in gaming. He shared his vision for the coming years:

“Some RPG game companies want to use AI for conversation with NPCs. Some companies have already started to use that. And in the near future, many companies will use AI for in-game features with the small language model to reduce risk.”

He also touched on the potential for AI in content creation: “Microsoft and other companies are trying to generate 3D models with AI. But artists always tell us the quality is not up to their expectations yet. But in the future, we can create generative AI for production level quality. I think it will be in five years. In five years, we can create that.”

Throughout the interview, Jun Shimoda provided a view of AI’s potential in gaming, acknowledging both the exciting possibilities and the challenges that need to be addressed.