AI Agent, a tool based on large language models, has one of its main applications in content generation, especially in the content-intensive gaming industry. At present, several projects have already showcased the potential of this technology, such as the international chess game ‘Leela vs the World‘ developed by Mudulas Labs, ‘AI Town’ launched by a16z and its portfolio companies, and the virtual companion game ‘AI Waifu‘ that incorporates the Virtual Protocol.

Under this trend, traditional game development teams have also begun active exploration of ‘AI Agent +Web3’. Among them, the game studio behind Void Labs plans to introduce AI Agent into its developed AAA action role-playing game ‘Blade of God X‘, to enhance game interactivity and diversity, thereby enriching the player experience.

From Web2 to Web3: Exploring New Development Paths

Since 2006, the game studio behind Void Labs has embarked on an exploration in the gaming industry, experiencing various stages of development from standalone games to MMORPGs, and then to mobile games. With globalization deepening, the team gradually recognized the limitations of the Web2 gaming market in terms of purchasing power and international Go-to-Market (GMT) and began to seek new development paths. The maturation of blockchain technology has provided new business models and user interaction methods for the gaming field, and its decentralized characteristics have broken through the geographical and platform restrictions of traditional games, enabling global players to game in a fair and transparent environment.

In addition, they also pointed out that with the development of AI technology, the “production relationship” has been redefined, and the traditional currency system is no longer applicable to the new production relationship. The transparency, tamper-resistance, and decentralization of blockchain technology provide more ideal support for this new production relationship. Combined with AI technology, it can not only enhance the richness of interaction within the game but also build complex economic and social structures.

Therefore, based on insights into the future development trends of the gaming industry, they began exploring Web3 game development years ago and launched the AI-integrated action role-playing game Blade of God X in 2023. This game recently obtained millions of dollars in financing from well-known institutions such as Delphi Ventures, OKX Ventures, Yield Guild Games, and Longhash Capital.。

“Blade of God X” is an action role-playing game that blends elements of Norse and Cthulhu mythology, adopting a Souls-like gameplay. In the game, players will portray an “Inheritor” reborn through cycles, setting out from Muspelheim and crossing through mysterious regions such as Voidom, Primglory, and Trurem. Players can choose to sacrifice or redeem to acquire artifacts or gain wisdom from deities like Allfather Odin and the evil god Loki, thereby changing their fate. In addition, “Blade of God X” is actively developing an AI Agent system to enhance the game’s interactivity. The game will also periodically adjust its development direction based on player performance, enhancing player immersion and engagement.

Reinforcement Training and Core Technology: Enhancing AI Strength

Previously, a research team from Stanford University and Google launched a game called “Stanford Town”, in which 25 virtual characters driven by large language model technology interact and live in a sandbox environment. This attempt not only increased game developers’ interest in AI Agents but also attracted a lot of player attention. However, these types of technology applications still face challenges such as high generation costs, slow response speeds, and repetitive content. The Void Labs team also recently announced they will deploy AI Agents in the Blade of God X game in the next round of testing and attempt to overcome these challenges.

In the Blade of God X game, the behavior and decisions of AI Agents can be controlled by models of different sizes. Small models handle simple tasks such as combat actions, while large models are responsible for handling more complex strategies and decisions. This model adjustment allows complex game logic to be implemented without sacrificing performance. At the same time, unlike the limited control of character behavior through simple state machine behavior trees, Blade of God X introduces the concept of “reinforcement training”, supporting AI behavior decisions through a neural network, improving AI autonomy and efficiency, and making game content less predictable and exploitable by players. The Void Labs team stated that during the training of AI Agents, they used over six million user behavior data to build reinforcement training environments for different gameplays and conducted hundreds of thousands of stack simulations for each profession.

After reinforcement training is completed, Blade of God X will implement 3D Laber using vision and language as inputs, allowing AI to more accurately recognize and understand the game environment and enhance the game’s immersive experience. To alleviate server load and significantly reduce response time, the optimized AI can run directly on the user’s mobile device GPU. In addition, to reduce the difficulty of player operations, the Void Labs team has designed a customizable AI Agent dashboard for users, allowing players to adjust the AI’s combat style and skill configuration according to personal preferences, making the AI’s thinking and processing ability closer to the player’s own, further enhancing the personalized experience. The team calls this “Play ARPG to Train AI Agents”.

‘Agents as a Service’: Creating Personalized AI Agents

Unlike previous attempts to enhance AI strength through machine learning alone, the Void Labs team focuses more on enhancing the player’s gaming experience, introducing a new concept ‘Agents as a Service’. They indicate that in the ‘Agents as a Service’ mode, the AI Agent can not only act as an auxiliary role when the player is online but also continue the game as a digital twin, AI Twins, when the player is offline. That is to say, the player’s AI Agent will highly resemble the player in terms of thinking and processing ability. For instance, you can invite a friend’s AI Agent to assist in combat when they are not online, and their AI Agent will mimic their gaming style and decisions, such as equipped skills, combat habits, and output methods, thus enhancing the realism and engagement of the game. Moreover, through blockchain technology, Blade of God X can ensure that AI Agent activities are transparent and verifiable and can generate actual economic value for players. For example, even if your AI Agent takes your place in a team task and achieves victory, you can still receive corresponding rewards. The team also revealed plans to develop a more complex, interactive Group AI Agents feature in the next generation AI Agent system, allowing AIs to cooperate and compete with each other.

The core of Agents as a Service is to use Visual Language Action (VLA) reinforcement training based on large language models. This method integrates visual, language, and action dimensions, which can not only help AI better recognize objects and scenes in the game, interpret and execute player text instructions, but also decide its behavior in the game based on visual and language inputs.

Furthermore, the Agents as a Service mode is a continuous automated service, which usually occurs when the player is offline or inactive, so it will not interfere with the player’s normal gaming experience. On the contrary, through this mode, Blade of God X has not only reduced the MMORPG game’s dependence on players being online at the same time but also significantly improved the overall contribution of active users to the game ecosystem. In the future, the team also plans to introduce on-chain random numbers and other components into Balde of God X, allowing AI Agents to autonomously generate unique emotions and strong behavior, to break the problem of content death cycle, and truly realize a scalable, value-supported AIGC mode. Therefore, in the third generation of Blade of God X AI Agent functions, AI Agents will have the ability to independently generate game content based on the game engine and material library, achieving ‘Generative AI Agents’.

Does Web3 Really Need AI?

Despite the introduction of AI Agents being regarded as a noteworthy innovation in the crypto space, it has sparked a discussion: Do these projects truly need AI? Or are they simply following a concept trend? Addressing this question, the Void Labs team has provided their insights from the gaming field.

Firstly, from a production relationship perspective, most AI tools on the market are focused on enhancing productivity, such as AI problem-solving tools and AI drawing tools, etc. However, there is currently no technology or team dedicated to addressing the production relationship issues between AI, like the asset interaction between AI and humans, and transaction and empowerment issues between models. The Void Labs team pointed out that Blade of God X is attempting to fill this market gap by integrating the game, AI, and Web3, for example, through token systems and AI Agent markets.

Secondly, in terms of gaming experience, the application of AI shows significant differences between Web2 and Web3. In a Web2 environment, the main purpose of AI is to increase player retention and purchasing potential, but its personalization is limited, typically adopting a centralized data processing and decision-making system, lacking sufficient flexibility and personalized customization. In contrast, in a Web3 environment, an AI Agent can be customized and optimized according to the specific needs of the player. This means that each user actually has a unique AI Agent solution to enhance each user’s experience in PVP games as much as possible. This is what we call “Agents as a Service”.

Furthermore, the Void Labs team mentioned that there is currently a centralization issue in the development of AI model networks, with each model striving to be a leader in its vertical market by improving computational efficiency and the accuracy of results. However, they hope that in the future, AI models can evolve towards a direction of free invocation, where each model can operate independently according to its own characteristics and market needs, no longer constrained by a centralized network. In such a decentralized network, each AI Agent can be an independent economic entity that receives rewards based on the value it produces. The Void Labs team pointed out that in this way, the decentralized characteristics of Web3 provide a new mode of operation and value distribution for AI Agents, which may fundamentally change our current understanding of AI interaction and economic contribution.

Recent Progress

The first round of Alpha testing for “Blade of God X” concluded successfully at the end of March, with a total of 2445 accounts created and 3660 codes distributed. Additionally, it received a score of 5.0 in the Epic Store’s Alpha testing. The Void Labs team has announced that they will fully deploy AI Agents in the upcoming second round of testing, enabling players to interact in over ten different scenarios. They also emphasized that their exploration of AI Agents is not limited to the current stage; the ultimate goal is to transform AI into an intelligent entity capable of parsing game engines and logic. These AI will be able to use user-generated content tools and game engines to create new game modes and rules.

In summary, the Void Labs team expects AI Agents to be more than just auxiliary tools in the game. They want them to become a core technical force capable of participating in all aspects of game operations, content innovation, and community development.

statement:

1. This article is reprinted from [substack], the original title is “Web3 + AI Agent: How does BOGX break the traditional game imagination?”, the copyright belongs to the original author [LINDABELL], if you have any objection to the reprint, please contact Gate Learn Team, the team will handle it as soon as possible according to relevant procedures.

2. Disclaimer: The views and opinions expressed in this article represent only the author’s personal views and do not constitute any investment advice.

3. Other language versions of the article are translated by the Gate Learn team, not mentioned in Gate.io, the translated article may not be reproduced, distributed or plagiarized.