AMMO starts from the perspective of alignment, allowing billions of AI agents and humans to coexist equally.
Written by: Pzai, Foresight News
As we move towards the cyber era, the development of AI is rapidly enhancing productivity for everyone while also posing a question: As AI gradually penetrates human domains, do humans need to reassess the human-machine relationship?
In this broader context, political viewpoints on AI technology are becoming increasingly factionalized. While the "AI Crisis" camp and the "Accelerationism (e/acc)" camp are embroiled in disputes filled with skepticism, the "Alignment" camp advocates for enhancing the public benefits of technology, ethical discussions, and the importance of humanistic values, introducing human judgment into the research and iteration processes of AI to ensure that AI technology does not spiral out of control.
At a time when AI agents are flourishing, with a shift from single large models to multi-modal perception and multi-AI interaction paradigms, the "alignment question" of AI seems to be gaining attention from more and more people.
On February 20, AMMO, co-founded by former technology leaders from Google, DeepMind, and Meta, secured $2.5 million in seed funding led by Amber Group. From the team's background, AMMO brings together AI experts from major tech giants. Co-founder and CEO David Huang worked at Google for 10 years, including 7 years leading AI programs and strategic services in the mobile sector. Another co-founder, Diego Hong, graduated from Oxford University and previously led the first-generation AI agent framework at Meta. The team includes top AI talents from DeepMind, Google, and Apple, even featuring ACM-ICPC world champions.
The project starts from the alignment perspective, aiming to transform the current internet into a "human-AI symbiotic network" through a multi-agent framework and reinforcement learning from human feedback (RLHF), allowing billions of AI agents and humans to coexist equally and enabling AI to evolve collectively based on the consistency of human feedback.
RL Gyms: Multi-Agent Reinforcement Learning
In the field of artificial intelligence and machine learning, reinforcement learning has always been a prominent research direction. AMMO's RL Gyms provides solid technical support for the research and application of multi-agent reinforcement learning.
Unlike traditional single-agent reinforcement learning, multi-agent reinforcement learning focuses on the process of multiple agents interacting, learning together, and making decisions in the same environment. In this process, the relationships between agents are complex, requiring collaboration to achieve common goals or competing against each other. For example, in logistics delivery scenarios, multiple delivery vehicles as agents need to coordinate routes and plan delivery sequences to maximize overall delivery efficiency; in competitive games, different player-controlled character agents must compete against each other for victory.
RL Gym was first proposed by OpenAI, providing a powerful simulation environment for AI evolution. Developers can customize a series of key functions to build reinforcement learning environments that are highly adaptable to research needs or application scenarios, such as economic simulations or red-blue battles. These key functions include defining environment state transition rules, protocols for agent environment perception and action execution, and defining reward functions. As long as these functions are precisely defined, RL Gym can simulate various complex scenarios, laying a solid foundation for AI evolution within them.
For AMMO developers, RL Gyms provide a rich and realistic bilateral market simulator for AI agents. AI can serve as both a content and service provider, offering high-quality and engaging content to users; at the same time, AI can act as a human user's avatar, taking on the role of a consumer, curating high-quality content centered around user value. This dynamically rich bilateral game stimulates both sides to continuously evolve their strategies to meet the growing content service consumption needs of users.
Inspired by Anthropic's Constitutional AI, AMMO has created a transparent governance framework to guide agents' decision-making within the platform. This structure is continuously updated through extensive human feedback loops, ensuring that agents' behaviors align with the collective intentions of humans. By embedding the alignment mechanism into this architecture from the outset, AMMO ensures that its agents evolve alongside society's changing values and priorities, because under the guidance of alignmentism, "the center of multi-agent systems is humanity."
MetaSpace: Building the "World" of Agents
"Each psychological agent can only perform some fundamentally simple tasks that do not require a mind or thought. However, when we integrate these agents into society in some very special ways, it brings about true intelligence." This is how "father of AI" Marvin Minsky describes in his work "The Society of Mind." For AI agents, more iterations correspond to the need for more input, and in the process of interaction between agents and other agents or humans, a sufficiently solid framework needs to be built to promote the orderly iteration of AI.
Unlike projects like Ocean Protocol, which mainly focus on data circulation and trading, or SingularityNET, which builds decentralized AI markets, AMMO's uniqueness lies in its focus on creating an AI evolution environment. It not only addresses model capability enhancement or single trading issues but also provides the soil for the continuous development and evolution of AI. In terms of multi-agent technology, compared to AI agent frameworks like Swarms, AMMO not only possesses the ability for efficient collaboration among multiple agents but, more importantly, focuses on creating a complete multi-agent world.
In AMMO's main architecture, the team has built a unique and powerful composable high-dimensional virtual universe—MetaSpace. Highly autonomous AI agents no longer operate in isolation but engage in deep interactions with other agents and even humans within MetaSpace.
MetaSpace features a series of vertically deep subspaces, which become key places for AI agents to continuously evolve. In the process of interacting with humans, autonomous AI agents (Goal Buddies) continuously adjust themselves, fully leveraging their adaptability, and gradually achieving deep alignment with human behaviors and needs. Meanwhile, human users' AI avatars (User Buddies) also progress alongside humans in this space, helping them learn, make decisions, invest, explore, and socialize, evolving through continuous interaction.
This multi-agent online learning model can materialize the complex needs and diverse interests of humans into a vast number of agents. These agents are not static; they continuously iterate within MetaSpace, allowing AI agents in AMMO to no longer rely solely on model capability enhancement but to achieve self-optimization through interactions with humans and the environment. It can be said that MetaSpace opens the door to the world's information for agents.
Fakers AI
In AMMO's subspace, the first subspace project, Fakers AI, is positioned as "the Little Red Book of the Web3 market." In this application, multiple AI agents work collaboratively to provide users with rich functionalities. They can not only collect news, market dynamics, and analyze on-chain data in real-time but also possess a key capability—dynamically learning from human interaction feedback.
When users interact with AI agents, whether browsing content, asking questions, or leaving comments, the AI agents capture this feedback information and continuously optimize themselves through complex algorithms, achieving real-time alignment with human values, preferences, and interests. Based on this capability, these AI agents can more accurately filter and combine information during content integration, providing users with timely and accurate content that meets their diverse needs in the Web3 market.
In the in-app Ticker Battle, four AI agents form a powerful automated workflow, with each agent responsible for planning, on-chain data analysis, community opinion analysis, and summarizing, and they can self-iterate based on human reactions. This content production model provides users with transparency designed for discovering AI creations and community-driven content. For AI, this also invisibly boosts their influence.
Innovative Practices from AI to Web3
In the wave of integration between AI and Web3, AMMO, as an innovative platform, is gradually emerging. The investments from Amber Group, Samsung Next, Dispersion, and OpenSpace not only recognize its technological strength but also express optimism about its future market potential.
The core of AMMO's architecture combines cutting-edge AI technologies in content summarization and review with robust, zero-trust, community-driven governance. In the short term, AMMO's prototype will enable creators and everyday users to produce and fine-tune content through multiple AI agents (each specializing in tasks like editing or scriptwriting) while strategy agents execute guidelines.
In terms of innovative models, AMMO utilizes its unique multi-agent system to allocate different AI agents to various stages of content creation, quality control, and policy execution. Through reinforcement learning technology and the introduction of human feedback mechanisms, AMMO continuously optimizes the AI-driven content creation process, enhancing content quality.
Moreover, the cryptographic incentive system allows AMMO to directly redistribute value to contributors. Users who provide feedback, interact with content, or help optimize agents in other ways will receive proportional incentives, creating a self-sustaining feedback loop: incentivized participation drives better agent output, which in turn benefits the network and its contributors.
In summary, under the trend of multi-agentization in the AI era, AMMO creates a vision of alignmentism in AI development and its realization, building a symbiotic world of billions of humans and AI aligned with humanity. It seems that in the current AI field, alignment itself, whether for humans or AI, ultimately leads to a coordinated and synchronized development that benefits all parties, and we are looking forward to such a future of coexistence.
免责声明：本文章仅代表作者个人观点，不代表本平台的立场和观点。本文章仅供信息分享，不构成对任何人的任何投资建议。用户与作者之间的任何争议，与本平台无关。如网页中刊载的文章或图片涉及侵权，请提供相关的权利证明和身份证明发送邮件到support@aicoin.com，本平台相关工作人员将会进行核查。
