"Although artificial intelligence (AI) has brought many prospects to the game industry, many people in the industry will still feel very anxious in the short term." On September 6, the 2024 Tencent Global Digital Ecology Conference held by Tencent Cloud During the gaming session, Chen Liang, general manager of Tencent Cloud Internet Industry Technology, mentioned in his speech.
The source of the industry's anxiety about this technology is that after the rise of AI in the global technology community, the outside world has put two major "tortures" on the game industry: How will AI be implemented in the game scene? How to further explore the application possibilities of this technology in games?
While AI is developing rapidly, the scale of the global game market has also reached a larger scale. A report from the data platform Statista predicts that the global game market will reach US$220.5 billion in 2025, with a compound growth rate of 10.18% from 2020 to 2025.
However, today, as the number of entrants continues to increase, rising R&D costs, intensified traffic competition, and changes in consumers' overall consumption willingness have made the competition among game companies in the market increasingly fierce. In order to adapt to this change, more and more game companies have also joined the ranks of "improving" game quality, including strengthening capabilities in R&D support, network services, data system construction, etc.
How to make game AI “controllable”?
An objective reality is that as domestic game manufacturers continue to accumulate research and development experience, the spillover effect of technology has become more significant. Adding AI to fully explore the potential of the game business has become the only way for many manufacturers.
The "New Quality Productivity Development Report of China's Game Industry" released by Gamma Data shows that nearly 80% of the leading game companies are deploying in technical fields such as artificial intelligence, digital twins, engine development, cloud technology and XR; nearly 60% of the leading game companies The company has built AI production pipelines to enable virtual content production or intelligent marketing; in addition, the top 50 domestic game manufacturers have invested in AI companies more than a hundred times.
"Gigabit Thunder has been exploring AIGC for more than a year. In terms of computing power, the company has also purchased the H100 with high computing performance." Liu Jiangdong, head of Gigabit Thunder game algorithm, mentioned at the conference.
He pointed out that the company mainly considers three factors when promoting AI technology. One is technological maturity. In the context of the rapid development and emergence of AI technologies, it is necessary to consider which technologies can actually solve problems, be implemented in practice and be implemented at low cost.
The second point is to choose the business scenario of the game. "There are many business scenarios where AI can be implemented, but as a business organization, including the technical team, the company also needs to consider ROI (return on investment) as the goal and analyze which scenarios have higher ROI." Liu Jiangdong said.
The third point is early exploration. Liu Jiangdong believes that while realizing the generalization of technology and framework, we should also consider whether capabilities in the same scenario can be reused.
At the conference, Chen Liang also mentioned that although AI seems to have generated a lot of popularity in the industry in the past year, from a technical perspective, AI has been "dormant" in the game industry for many years. Such as 3D skinning, character generation and automated UV expansion, etc., which are the in-depth applications of traditional AI in games.
As early as 2019, the artificial intelligence research company OpenAI designed the AI training tool "Five" for the e-sports team OG of the MOBA game "Dota2" to provide training and data support for team players by generating AI teammates or opponents.
Tencent Cloud has applied cutting-edge AI technology in PaaS products such as intelligent customer service, machine learning platform and speech recognition. Among them, Tencent Cloud TI-ONE uses a storage-computing separation architecture to open up big data computing capabilities and model building capabilities, reducing the use cost and time cost of data-to-AI capabilities. Real-time AI decision-making responds to various mutation scenarios to better ensure ROI.
However, Chen Liang believes that when generative AI technology derived from AI is applied to the game industry, the current focus is not on the generalization of functions or data scale, but on how to control the results it generates within the range of human expectations. . "For example, many game products are currently using generative AI to make intelligent NPCs. There can be hundreds or thousands of NPCs in a game. How to make AI NPCs more like real people and more controllable is still being promoted. in the process," he pointed out.
The present and future of game AI
It is worth mentioning that just recently, Cai Haoyu, one of the founders of MiHoYo, said bluntly on the LinkedIn platform that AIGC has completely changed the way game development is done.
He believes that future game creation will be mainly dominated by two types of people: one is the 0.0001% of top practitioners, and the other is the vast majority of amateurs. Ordinary game developers who are between ordinary and professional should consider changing careers.
"What AIGC really affects is essentially a way of human-computer interaction. In the past, humans and computers mainly interacted through code, but in the future it may develop into natural language interaction." Liu Jiangdong believes that with the rapid development of AI technology, if practitioners Being able to master the relevant skills of large language models (LLM) as early as possible can more quickly improve the personalized capabilities of game development.
BUD co-founder and CEO Feng Wenhui believes that whether a game is driven by product design or technology, AI-assisted development is a good choice. "In the early stages of creation, our goal is to improve the work efficiency of creators. By combining compilers and code assistance tools, such as Copilot, and 3D Copilot in the 3D field, we aim to lower the threshold of creation and improve efficiency through artificial intelligence technology. This not only makes game development more efficient, but also adds interest to the game and creates a different experience from traditional games," she pointed out.
After AI has become a must-talk topic in the gaming industry, what changes has AI itself brought to the gaming industry?
Huang Long, technical director of Chuangku Interactive, mentioned in his sharing at the conference that a lot of creative materials are needed in the distribution business. This type of content that requires unconstrained and whimsical ideas is very suitable to be generated with AI assistance.
"In the generation of delivery materials, we have tried many AI applications. In addition to using common tools such as Stable Diffusion, some videos can even be directly used for voice broadcasts, game scenes, speech synthesis or voice cloning, shortening the entire material production cycle. , and also accumulated a lot of reusable materials," Huang Long said.
After the conference, Chen Liang also mentioned in an interview with the media that the application of cloud gaming and AI in the gaming industry has always been a hot topic for exploration in the industry. Currently, many teams are investing a lot of effort into integrating these technologies into game development. "Despite this, the above new technologies have not fundamentally changed the traditional game development model. The usual approach is that developers will first complete the development of the client and server, and then during the game release stage, a cloud gaming platform may be provided. New ways to allow players to easily experience the game can help solve client-side games. There are problems such as large size or insufficient mobile phone performance, but in essence, it is still regarded as a value-added service. For the application of AI, it may be used in 3D mapping, UV expansion and other aspects during the game development process, and in game operations. At this stage, AI may be used to provide automated customer service. These applications are also considered value-added services.”
Chen Liang further pointed out that he believes the real potential of cloud games and AI games lies in their ability to become native game elements. Game manufacturers and upstream and downstream stakeholders should be committed to studying how these technologies can be deeply integrated with the game design and development process to create a new game experience, not just as an additional feature of existing games.
"In the future, after the integration of games and real AI, my ideal state should be similar to the scenes shown in the American TV series "Westworld". The story line of the game can be continuously extended, and the characters can be constantly introduced. As a practitioner, I am more concerned about this What is the underlying implementation logic of the game? Does it rely on configuration and parameters, or is it completely model-driven?" Chen Liang said.