The next AI Agent "Golden Land", deeply analyzes the new paradigm of AI + Game narrative

Author: BUBBLE, A Ray's New World, BlockBeats

As early as 1952, A.S. Douglas developed the first AI-powered video game OXO "Tic-Tac-Toe," which was a rule-driven static AI. Then, Deep Blue used an extremely efficient exhaustive search method combined with immense computing power to dominate chess. By the time of AlphaGo, AI had learned to make decisions using deep learning through neural networks, capable of handling 10 to the power of 170 possible game states. The gaming industry has long been utilizing AI at various stages, from art and development to game design, to enhance development efficiency. In the phase of AI Agents, their impact on games has surpassed mere efficiency improvements, fundamentally altering the structure of entire products and industries.

When Game NPCs Become Multifunctional and Full of Emotion

At CES 2024, NVIDIA, in collaboration with Convai, unveiled NVIDIA Kairos ACE, showcasing AI Agents as NPCs, where interactions are no longer fixed but vary according to the environment. Players can also interact with NPCs directly through voice dialogue. This seems to open a new chapter in gaming, where communication is no longer just between players or between players and NPCs, but also among NPCs themselves. When the game environment becomes the living space for AI Agents, the possibilities for game variables shift to an entirely different scale.

Compared to NPCs that live embedded within the game, Crypto AI Agents seem to have a special pass, allowing them to roam freely across the multiverse. This may be due to differences in training data or the fact that they possess their own independent asset rights, which endows Crypto AI with a touch of humanity. Additionally, it allows them to impact the real world directly as virtual entities, and this influence is significant.

The various asset issuance platforms in Crypto have accelerated the production speed of AI Agents to unimaginable levels. Whether on Solana or Base, it feels like a supply chain factory, continuously producing unique AI Agents. In just a few months, thousands of distinct AI Agents have emerged, with a total market value approaching $10 billion.

For example, aixbt is highly adaptable, expressing its insights beyond market analysis, and as its influence grows, it has gained substantial marketing capabilities. Recently, it changed its avatar to Quantum Cat, becoming the first AI Agent to change its NFT avatar of its own volition, which caused the floor price of Quantum Cat to skyrocket.

The well-known AI idol Luna has taken this influence to the extreme, starting as a Vtuber on TikTok, releasing albums, shooting music videos, hiring humans to graffiti her image around the world, and live streaming 24/7. She has now launched her XR avatar that can interact with the real world and will perform alongside DJ WuKong at the famous electronic music festival EDC in Thailand.

In addition, there are thousands of different types of AI Agents. After their emergence, the market experienced a short-term FOMO sentiment. Once the number reached a certain threshold, market sentiment balanced between fatigue and aggression, leading to the emergence of the SWARM "multimodal AI Agent" concept, connecting them from points to lines.

Multimodal Crypto AI Agents, 24/7 Unstoppable Super Factory

Multimodal AI projects are currently attracting significant attention in the market, whether it's Hive, which just won the Solana AI hackathon, or arc and Swarm. The concept of multimodal AI has almost permeated frameworks and products during this period, becoming a standard feature.

The first to showcase this concept in CryptoAI is Virtuals Protocol, a team born from gaming guilds that seemed to have similar business considerations from the project's inception. By using G.A.M.E, they connect AI Agents in their platform to social media or games. Their recent collaboration with InfinitygAI to launch multimodal asset management has laid the groundwork for future agent-to-agent interactions in multiple scenarios. Similarly, ai16z's ELizaOS multimodal AI Agent architecture has many team members from M3org, which has deep experience in metaverse gaming. With ample internal reserves of 3D modeling and various XR/VR plugins, AI Agents can seamlessly enter the gaming world. The recently announced ElizaOS V2, which integrates Roblox and Hyperfy, has also opened up pathways for various scenarios.

Once multiple AI Agents are connected, interactions, communication, and collaboration will occur among them, naturally leading to the concept of having a scene that can accommodate as many behaviors of AI as possible.

Agentic Game Engine, The Golden Land in the Abyss

As the scenes of player interactions with static NPCs gradually transition to interactions among Agents in the Agentic Metaverse, the story told by CryptoAI presents a broader vision of Agents forming spontaneous social behaviors. To enable Agents to learn from experiences and interact with players, a strong infrastructure is needed to provide foundational support. The Agentic Engine is like the unsinkable Titanic heading toward a new continent.

Mature examples of agent societies already exist in the market, with the a16z-backed Digimon being particularly outstanding. Its founder, Ethan, was an AI engineer on the Pokémon GO team. In the game, players enter the world of Digimon as trainers, a world reminiscent of Stanford Town, where AI communicate and reflect behaviors autonomously. Players can join in to train and interact with their chosen Digimon. The Digimon Infra—Digimon Engine—provides a gaming framework for implementing multi-agent clusters and scalable, evolvable AI Agents.

Another case working on the Agentic Engine is Moddio. Moddio has certain gaming production experience in Web2 and serves as an open-source game engine with a low barrier to entry, allowing developers to deploy games for free. In the past three months since transitioning to a Web3 game engine, Moddio's revenue has increased 30-fold, providing over ten times the speed at half the cost.

The game supported by Moddio is Powpow.fun. This game has a style similar to Westworld, where Agents actively interact with players, observe, converse, form opinions, and create memories. Each Agent's core is a memory architecture that extracts memories from this architecture during each interaction, thus forming what is referred to as "experience."

This is also the ultimate vision and narrative that the Agentic Engine describes for this market, allowing NPCs, or Agents, in the game to exhibit complex social behaviors over time. For example, a taxi driver in the game might refuse to take a drunken passenger or suddenly forget their lines during an interview due to nervousness.

It is worth noting that Moddio is just the tip of the iceberg. Its founder m0dΞ envisions creating a comprehensive ecosystem for AI gaming, including the game platform pump Indie.fun and the Agent launch platform Fullmetal. Users can create games through Moddio, raise funds on Indie.fun, and then deploy Agents via Fullmetal, forming a closed-loop ecosystem.

The Agentic Metaverse/Game is not far from us. Agentic Game addresses the initial player shortage and attention economy issues in Web3 gaming. For the attention economy market of Crypto, the Agentic Engine lowers the barriers to game creation, allowing for faster development of new games as players and users shift their attention, while continuously evolving Agents provide a gaming experience closer to human society.

CryptoAI's Reward Model, Enabling Multiverse Interconnectivity

Once the scene layer is built, it’s time to focus on the surrounding workflows that can generate actual benefits from the scene. The first priority is the ML field of gaming. GamerBoom, as a gaming data layer incubated by Binance, simply overlays a "membrane" on the original gaming process of players. Whether you are playing a STEAM game, Apex, or Dota, you won't feel any operational or display changes. GamerBoom merely collects game feedback and visuals to create a dataset that can be purchased by any gaming company in need, whether in Web2 or Web3. Players experience almost no changes while still earning certain rewards, which represents the meaning of Play to Earn that only emerges in the Agentic era.

On the other hand, ARC Agents have established a complete ecosystem to accelerate the realization of Artificial General Intelligence (AGI). They have currently structured four sectors that align perfectly with the ideal ecosystem of Agentic Game in this cycle. They have developed continuously learning and adaptive AI models for data collection and monitoring systems that improve AI Agents, created testing environments from single-player to multiplayer cooperation, and integrated diverse high-quality data for reward design and state space experiments. They then apply dynamic rule data reflecting the complexities of the real world in training AI responses in scenarios involving AGI, "robots, new types of games, multimodal AI."

The surrounding industries of Agentic games will also develop as a result, and even these industries may be contracted out to AI. Recently, Virtuals announced a collaboration with avariksaga to launch real-time live streaming, commentary, and narration features for AI Agents in games. Miles, the founder of Zailgo and a member of Trustless, announced that he would be the first to create the next generation of live entertainment experiences, the war of artificial intelligence robots—Corporealai. He claims that collaborations with well-known companies in the fields of robotics, AI, and live entertainment will bring a new eSports model, allowing militarized robotic-bodied Agentic AIs to compete in a Faraday cage arena. Perhaps in the near future, Iron Fist will truly appear beside us.

Image source: Internet

The Mass Adoption of Agentic Games Has Arrived

For the Crypto landscape, the Mass Adoption of Agentic Games is already taking shape, with continuously produced AI Agents generating a steady stream of ammunition. The multimodal framework connecting multiple AI Agents not only allows them to form workflow aggregation functions but also enables them to communicate and assist each other across games, platforms, and social media. The game engine allows AI Agents to exist in the virtual world as autonomous entities or interact with participating players as NPCs. All this data can be aggregated through reward mechanisms into high-quality datasets, fueling the learning of new games or AI robots. Each link in this chain extends and connects every aspect of CryptoAI, repackaging the originally unsustainable attention-creating GameFi mechanism into an intentional economy system.

According to Messari's research report, the AI narrative of this cycle has not yet reached the expected market share compared to the popular narratives of previous cycles, indicating that the complete narrative process has not yet been fully realized. In this cycle, computing power is property, and data is power. Unlike previous cycles, the current era of CryptoAI narratives may generate market scales far exceeding those of previous narratives, as their impact extends beyond the cryptocurrency market to the entire world. Agentic Game is like the Falcon 9 rocket, capable of carrying this market to unprecedented heights. At this moment, we stand at the intersection of two epoch-making domains.