GuShiio.com In February, the AI wave led by DeepSeek had a huge impact on the game industry.
On February 17, Musk announced on social platforms that xAI would establish an AI game studio, announcing two core concepts in a high-profile manner, breaking the monopoly of large companies and using AI to reconstruct the game experience. Subsequent news also stated that the team plans to increase the ability to apply high-resolution images, and Grok 3 is expected to produce a 3A-level game next year.
same time,After Microsoft released Muse (a generative AI model that generates game visuals and controller operations), CEO Satya Nadella said the company plans to apply it to multiple video games, and gradually launch works produced by AI. In the announcement, Microsoft showed how Muse used training data from the 2020 game “Peak Team” to generate game screens.
On the other hand, the two big brothers in China are not willing to be outdone.
last weekTencent’s tactical competitive mobile game “Peace Elite” announced that it will connect to DeepSeek to inject artificial intelligence into digital spokesperson “Gilly”.
According to the plan, grayscale testing will be launched on February 25 and will be open to all users in the short term. “Peace Elite” is also Tencent’s first game product and the industry’s first batch of game products to deeply integrate large model technology. It will provide real-time intelligent services to hundreds of millions of Peace Elite users through real-time interaction.
Netease’s “Cold Mobile Games Against Water” released an update announcement saying thatOn February 21, the intelligent NPC “Shen Qiusuo” driven by the official version of DeepSeek will be launched in the game soon.According to the official statement, it is DeepSeek’s first time to enter the game, its first time to have a virtual image, and its first time to become a virtual intelligent life form.
Interestingly, according to foreign media reports, Nadella emphasized the importance of AI creating real-world value in her blog on the one hand, rather than blindly pursuing unrealistic ideas.
On the other hand, he also admitted in the blog that generative AI has not yet shown a substantial economic impact, and current AI tools are also facing technical challenges and require continuous supervision.
Excited and vigilant, applied and overheated, where have the current AI+ games reached?
Breaking the impossible triangle
From the perspective of underlying logic, AI can develop rapidly in the game industry. It is not only the result of capital market speculation, but is in line with the industry’s overall environment of reducing costs and increasing efficiency. It also has the feasibility of practical implementation, that is, AI+ Games can create more value.
Improving production efficiency through technological improvement and effective use of tools is the real key word in this round of game AI explosion.
At the ChinaJoy Summit Forum last year, Netease Vice President Pang Dazhi raised a point in his speech,”The quality improvement and efficiency improvement brought by AI is expected to become a way for the game industry to get rid of the traditional growth constraints and development path, breaking the long-standing triangle of impossible.”」。
“Impossible Triangle: Cost, quality, and efficiency are difficult to meet at the same time.Usually, a team of hundreds of people takes several months to produce content that may only take two weeks for players to consume.
We talked before that AI can help maximize cost, quality and efficiency. In fact, it includes two aspects. One is to improve efficiency and improve quality. It is the 1.0-2.0 stage, from “can do it with ease” to “How to do it better”; the second is the change in production relations.Some players have shifted from consumers to producers of new content.
Nowadays, the trend of AI+ games overseas is booming. If we compare the exploration directions of last year and this year, we can find that the logic of actual applications in enterprises has not changed.
The two main directions are still AI agents and efficiency improvement in the production process.
For example, Netease previously used Ali Tongyi Qianwen, Baidu Wenxin Yiyan, and its own large model to create intelligent NPCs to interact with players in “Reverse Water Cold Mobile Games”.
After tasting the sweetness, last year’s “Eternal Tribulation” mobile game also added AI teammates who can communicate in real time and smart voice face-pinching methods (produced by Netease Fuxi), focusing on AI teammates who listen to commands, can cooperate, and can chat.
“Egg Party” is based on the “Universal Generator” of the park map created by AIGC and has been used by tens of millions of players.
The same is true for “Yuanmeng Star”. Tencent announced at the GDC 2024 Game Developers Conference that the GiiNEX AI game engine provides players with a series of UGC level design tools available in the game. It can increase the efficiency of some work in areas such as game scene production and content generation by more than 40 times. In the “Naruto” mobile game, the large-scale intensive learning program for fighting games directly reduced the time and resources required for training by more than 90%.
In last year’s CJ speech, Kaiying actually demonstrated the interaction with the AI large model, automatically created and rendered the game background, logic, numerical values, modeling, etc., and generated the process of an applicable aircraft game.
Through the Transformer pre-training model, self-optimization is carried out and self-iteration is carried out for pre-training. Because pre-training is more efficient and accurate when solving problems, reducing repetitive linear calculation costs and time, thereby helping the AI large model to fully automatically generate games.
In the distribution and promotion process, in addition to saving 60%-80% of man-hours in the original character painting process and producing 3W + pictures and 3000+ models per week, Sanqi Mutual Entertainment’s art AI Zhongtai Turing also upgrades and covers the distribution material. Production of video and audio. Both video and audio can be improved by more than 20%, and 80+ AI videos and more than 200 audio pieces are produced every week. Including the office field, Sanqi Office Intelligent Agent Xiaoqi takes an example, integrating external information search (enterprise, product, industry dynamics), drawing, document writing, translation and other functions.
In this round,Whether it is the new actions of xAI, Microsoft or Tencent Netease, or the practice in the past set of logic, what is the real progress?
Equal rights and cost reductions benefit more developers
Industry discussions on DeepSeek are often linked to the word AI affirmative action.
According to a report by Gamma Data last year, more than 99% of game industry respondents have introduced AI technology or tools in companies or departments. At the research and development level, about 80% of respondents perceive that AI has improved project efficiency by more than 20%, and nearly 10% of respondents believe that AI has improved efficiency by more than 50%.
Currently, AI is most widely used in the production of game auxiliary content, such as “improving the freedom of NPC, plot and other content” and “assisting the creation of art, music, code, etc.” accounting for nearly 70%.
However, the reality is that in the past, most of the companies that could make achievements in AI+ games were powerful leading companies or listed companies. Everyone has the ability to build, train, and apply their own AI models and use them to produce results.
It is not difficult to see from the above cases that Tencent, Netease, Sanqi, Kaiying, Giant and other companies are sharing corresponding results based on their own AI models.
In the game industry, technology itself can become a competitive barrier in the new era. This year, DeepSeek has become the protagonist of domestic cooperation, which has also inspired overseas companies to increase their efforts to build their own models to compete with it.
Tencent Netease has its own model and will choose to cooperate, and more small and medium-sized teams can enjoy the technological progress brought by AI affirmative action at a lower threshold to try the application of “AI+ games”. The more participants participate in it, the greater the possibility of finding a new path.
From the objective environment, the current game market environment itself continues to pressure the impossible triangle. One of the most significant signs is the pursuit of Evergreen games. Newzoo’s 2024 Global Games Report states that more than 60% of gaming time is spent on games that have been in operation for six years or more. New games account for less than 10% of game time.
The logic of the long-term game product is completely different from the thinking of stand-alone and short-term games. At present, the long-term games test the content filling and iterative optimization of long-term operations, while the stand-alone and short-term games pay more attention to creativity, gameplay and a complete narrative framework. The focus on talent, ability, and capital costs is vastly different.
In other words,The open source of high-quality AI tools is actually similar to the explosion of game engines driving the development of the industry. Practitioners in the entire industry can enjoy the production dividends brought by cost and efficiency improvements.Orientation training the production tools you need. Depending on your actual situation, choose a different path between evergreen or short-term, and whether you play a stand-alone or online game.
In the next stage, we should expect more developers to train the exploration results of the models they need, and through individual verticalization and continuous exploration of track segmentation, we should jointly form a hundred flowers in the industry.
