How does AI empower GameFi?

Author: Dylan Wang, Dr. Nick

I. AI Empowering the Traditional Gaming Industry

Large games contain a rich array of sub-content, including 3D animation, special effects, audio, art, and text. The development process generally takes a long time, is complex, and requires a high degree of industrialization, real-time performance, and interactivity. Development units often need to invest a significant amount of human, material, and financial resources over the course of years. For a long time, the traditional game development field has been characterized by a "triple constraint" of cost, quality, and efficiency, where only two can be chosen at the expense of the third. For example, the production of "Red Dead Redemption 2," which features a realistic map of over 28 square miles and 1,000 NPCs, took 8 years, employed over 1,200 dedicated developers, and cost nearly $300 million. Today, thanks to the powerful capabilities of AIGC, game developers can break free from the constraints of traditional game R&D, lower development thresholds and costs, and significantly enhance game development efficiency.

AIGC can provide precise and powerful empowerment for game development, storyline planning, character interaction, and online operations through technologies such as machine learning, deep learning, and natural language processing. Specifically, AI can empower traditional games in the following three areas: pre-production, in-game experience operations, and disruption of the game industry structure.

1. AI Empowering Game Production

In terms of game production, both art and planning AI provide tremendous support for workers.

AI Used for Generating Various Game Assets, Achieving Cost Reduction and Efficiency Improvement

Different functionalities of AIGC can be applied in various aspects of games, including storyline design, character design, voice acting and music, art concept design, game animation, 3D modeling, and map editing, allowing for the generation of various game assets.

Taking the AI image generation company Scenario as an example, Scenario is the first commercial AI-generated game asset platform that can train AI models based on user-input text and reference images. Users upload training data on the platform, such as characters, props, vehicles, weapons, skins, buildings, concept art, pixel art, and sketches, and then, with just a few clicks, users can create their own generative AI engine to produce game assets that meet production standards based on the input text and reference images.

When conducting concept art design and 3D modeling, AIGC tools like Midjourney can provide good creative assistance, while Stable Diffusion, as an open-source model, offers more precise control. Epic Games showcased their latest Metahuman Animator technology at GDC 2023, which allows developers to create high-quality animations in minutes, generating facial bones from just three frames and supporting animation generation from video and audio. During a live demonstration, the presenter recreated the expression details of a live performance in the digital space in just 2 minutes. Opus.ai developed a new text-driven 3D world generation method that allows users to create dynamic lighting, camera controls, terrain, trees and animals, buildings, roads, and humanoid animated characters through text input. In this application, users can create 3D game scenes and various components and dynamic effects simply by typing.

In game planning, Tencent AI Lab showcased AI level auto-generation technology at GDC 2022, where AI can generate over a thousand high-quality levels in just a few hours. In terms of writing, AI can also provide outlines and dialogues for games, accelerating content production.

The powerful generative capabilities of AICG not only improve the efficiency of game producers but also significantly reduce costs. The gaming industry is riding the wave of generative AI towards accelerated growth.

AI Used for Assisting Game Testing

Game testing is a crucial phase in the game development cycle, aimed at ensuring game quality, reducing post-release risks, and providing players with a better gaming experience. In game testing, testers will use different testing techniques for various testing objectives, testing gameplay, game flow content, game systems, device compatibility, etc., recording issues found in the game and reporting bugs to developers through management tools.

In specific game testing applications, for example, the Japanese card game "Shadowverse" utilizes AI to test card combinations, training an AI that can play the game to automate game matches, helping to discover potential bugs and balance issues, greatly improving efficiency and reducing testing costs. The mobile game "Endless Runner" uses AI to automate testing of various actions and the rationality of obstacle arrangements in the game.

2. AI Agents Enhancing Player Game Experience

Increasing Playability in Level Design and Game Difficulty

In terms of game experience, AIGC's native games and innovative gameplay have already taken shape around AI's natural language generation, image generation, and intelligent NPC capabilities. The AI "Wang Zhe Rong Yao" has been widely used in level evaluation and testing; in "Left 4 Dead," a system called "Director" adjusts zombie spawns, item placements, and overall difficulty based on player performance to ensure playability.

Real-time Diverse Dialogue of NPCs Enhancing Game Experience

In traditional games, storylines are pre-designed, and players can only experience limited game content in a fixed sequence, making the revenue of role-playing games highly correlated with content reserves. With the advent of AI Agents, NPCs in games are given "life." AI NPCs design daily life trajectories like humans, remember events of the day, and every interaction between players and NPCs will alter the NPC's life trajectory, creating infinitely possible game content.

Ubisoft's La Forge has developed an internal AI tool called Ghostwriter to assist game writers in generating draft dialogues for NPCs during triggered events, such as dialogues between NPCs, enemy dialogues during battles, or dialogues triggered when players enter certain areas. These tasks were previously time-consuming for game writers (planners). With the Ghostwriter tool, AI can automatically generate samples for selection and modification based on the NPC's basic settings, saving a significant amount of time and allowing writers to focus on other core storyline elements.

Currently, a representative application case in China is NetEase's use of an intelligent NPC system in "Nirvana in Fire," where over 400 NPCs are powered by NetEase's Fuxi AI engine, each with independent personality traits and behavior patterns. Compared to previous game NPCs, interactions between players and intelligent NPCs are no longer pre-set and procedural, but rather highly free and completely open. An example from abroad is "Mount & Blade II," which has implemented dynamic responses from NPCs.

AI Used for Game Marketing and Optimization

AI can be utilized in user acquisition, from creating marketing materials to analyzing advertising effectiveness and executing user acquisition campaigns, improving efficiency and quality while reducing costs.

3. AI Disrupting the Game Industry Structure

l Lowering the Entry Barriers for the Gaming Industry

AI has significantly reduced the development costs for game producers. Previously, generating game assets and animating required manually drawing each frame, but now, using AI can greatly lower costs. This breaks the traditional model of high investment in games, providing conditions for small game developers to emerge.

l Increasing Player Power

With the enhancement of AICG, every player can become a planner and complete tasks such as coding, art, and testing through AI-assisted UGC, allowing players to create their own games and achieve "decentralization of games."

II. How AI Drives GameFi Growth?

In summary, the reasons for the explosive growth of GameFi projects followed by a death spiral are as follows:

The short-term attraction for players was the "P2E" model, which was extremely popular in 2021, but the appeal of "P2E" has declined in the later stages. In 2021, as many as 67.9% of respondents believed "P2E" was the biggest driving force in the industry. This model was unique when the gaming industry was just emerging, prompting many players to eagerly participate. However, over time, the P2E mechanism has lost its significance as a driving force in the industry, dropping to eighth place in surveys conducted in 2023.

The speculative nature of games outweighs their gaming attributes, making them unfriendly to players. Most users participating in blockchain games are cryptocurrency players, and the quality of games varies widely. Game content emphasizes strategy and competition, and the economic models have short lifespans. Some players outside the crypto space even view these blockchain games as online gambling, leading to many high-risk speculators in GameFi. New players must purchase in-game items or assets to play, and although they can earn back from the game, as the circulation of in-game tokens increases, if the game is not engaging, has poor playability, or has an unstable economic model, the tokens will depreciate, making it difficult for players to recoup their investments. The focus on asset value rather than gameplay characteristics makes it hard for GameFi projects to sustain. This issue stems from several inherent factors of the play-to-earn model. A key problem is the unsustainable token economy that typically drives these projects. When token rewards exceed the actual value generated in the game, an unsustainable inflation cycle occurs, leading to price drops and loss of player interest.

Poor game playability fails to retain users and attract new players. GameFi games are generally single-scene, NFT-based, and cannot attract new players. Compared to traditional games, they lack playability, and the development difficulty and cycle are shorter. Many low-quality mini-games quickly seize market share. Increasing application scenarios to extend the lifecycle of blockchain games and sustainable revenue models is a primary goal for GameFi.

During this quiet period for GameFi, major companies are focusing on addressing the issues left over from the previous period.

In user acquisition, expanding and innovating the P2E model to continuously attract users. For example, Stepn's explosive popularity in May 2022 expanded "play to earn" into "move to earn." The basic gameplay of STEPN is simple: players equip NFT sneakers in the game app and earn token rewards by running in the real world; the rewards can be used to upgrade and repair sneakers to improve "earning" efficiency or sold directly. This was one of Stepn's most popular weeks, with 10.7% of super users, and the number of super users entering SOL reached 1.79 million, almost equal to the total number of users who deposited funds, making this project truly powerful, contributing 39% of high-level users entering SOL, nearly matching the 920,000 ordinary users. This data is particularly impressive, achieving a user base of millions in just a few months. Based on this, X2E games have emerged. The "X" in X2E can be defined as any human behavior, such as gaming or moving, to earn rewards. Unlike earlier purely profit-oriented games, X2E places a high emphasis on community culture and values, often using meaningful real-life or creative actions as core game mechanics.

In user retention, first improve game playability. Start developing higher-quality game products, also independent game teams with blockchain as the main background, but compared to the previous stage, more AAA studios and members are joining, significantly improving overall game quality. A small portion of curious traditional gamers are attracted to relatively high-quality games and join the blockchain gaming community, gradually bringing GameFi into the public eye as a normal attribute of gaming.

Secondly, address financial sustainability issues. If the game itself encounters financial mechanism problems, DAOs play a crucial role in maintaining the stability of tokens in the secondary market or repurchasing tokens from the market through bond mechanisms to maintain healthy liquidity. GameFi will solve the issue of single game economic models based on the previous stage, avoiding circulation within a single game.

Currently, the focus in the GameFi field has shifted to the intrinsic value of games, making the continuous attraction of new users and retention of existing users a new development priority. AI empowerment also plays a role at these levels.

1. AI Further Expanding the P2E Model

With the support of AI, GameFi can further expand "play to earn." The previous round of GameFi explosion was driven by AI technology achieving "move to earn." This time, AI is expected to lead GameFi into the "love to earn" era and drive a new round of growth.

(1) Sleepless AI

Sleepless AI extends the pay-to-earn model further. By shaping "love to earn" through AI, it attracts users. Sleepless AI is an innovative Web3 + AI virtual companion game that uses AIGC and LLM to create rich story-based gameplay and continuously evolving interactions with characters, providing players with emotional support and immersive gaming experiences.

The flagship game of this project is HIM, which employs advanced large language models, enabling virtual characters in the game to generate natural and distinctive dialogue content based on different game situations and personality traits. This technological application greatly enhances the realism of characters and the immersion of players, creating a comprehensive and three-dimensional virtual image that becomes part of players' daily lives. HIM's long-term goal is to establish a virtual universe centered around AI boyfriends, AI girlfriends, and AI pets, providing users with support and companionship. HIM not only marks the natural fit of AI with gaming but also demonstrates how Web3 games can leverage AI technology to create entirely new interactive experiences.

Project Information:

X Account: https://x.com/SleeplessAI_Lab

MCP (as of 2024.6.15): 109,802,401

FDV (as of 2024.6.15): 884,633,856

(2) Palio

The Palio project also utilizes large model technology to join the "love to earn" wave. Unlike Sleepless AI, the companion for players is not a virtual partner but a virtual pet. Technically, Palio is based on using large language models (LLM) as the core of the agent system, developed by a team of researchers and engineers from OpenAI, Stanford University, and Google Brain. Specifically, Palio's large language model performs excellently, scoring 155 on the WAIS benchmark, surpassing the average person and reflecting its outstanding language comprehension abilities. To better enhance the companionship attributes of the game, Palio has made efforts in enhancing AI empathy. To convincingly convey emotions, it must be fine-tuned using real emotional datasets. PalioAI has collected a large amount of such data, allowing it to improve its algorithms. This gives PalioAI an advantage over other existing knowledge-based LLMs, enabling it to resonate more effectively with users and provide detailed feedback. Finally, PalioAI's large model excels in long-term memory compared to other companion games, using retrieval-augmented generation (RAG) to give PalioAI a true sense of memory.

Here are specific applications of Palio in the game:

• Helping players complete tasks: Palio can provide explanations and hints for game tasks, assisting players in completing them smoothly.

• Providing game information: Palio can offer background stories, character introductions, gameplay explanations, etc., helping players better understand the game.

• Providing game suggestions: Palio can offer strategic advice based on players' game progress and performance, helping them improve their gaming skills.

Project Information:

X Account: https://x.com/PalioAI

MCP (as of 2024.6.15): 813.25 ETH

(3) SpaceCatch

Twitter: https://x.com/spacecatch_io

Official Website: https://spacecatch.io/

White Paper: https://whitepaper.spacecatch.io/

AI can also empower the previously popular "move to earn." SpaceCatch is an AI-driven "move to earn" game, with a team experienced in implementing AI, especially in game tasks. It utilizes AI to create game tasks tailored to individual players based on their behavior, level/avatar specifications, and in-game activity, ensuring players feel the game is personalized to their preferences.

Project Information:

X Account: https://x.com/spacecatch_io

MCP (as of 2024.6.15): 916.57 ETH

2. Players Creating AI to Join Game Competitions

In this new model, players can create their own unique AI in the game and let the AI participate in competitions against other players. Each created AI is generated based on the player's unique data. This model combines creativity and competitiveness, making it noteworthy.

(1) AI Arena

AI Arena is the first game to integrate Human x AI collaboration. In this game, players transfer their personal data to the AI through a process of imitation learning. In fact, the process of players creating AI is also the process of training the AI large model.

Creating NFTs: When players purchase new NFTs for game competitions, the core neural network parameters are randomly generated. This means the neural network will initially perform random actions, as the network has not yet developed any skills.

Data Collection: Players demonstrate operations to the AI, and while playing the game, they are essentially creating an operation list for the AI to replicate and learn. In machine learning terms, players are creating the dataset that the AI will train on. Players need to consciously create useful datasets for the AI to learn from during gameplay.

Choosing Appropriate Datasets and Applications: How much does the player want the AI to remember from previous training? Determine the functions the AI focuses on during training. For example, players can instruct their AI to focus on a specific training course.

Training: After collecting and configuring data, the AI updates its parameters. The player's AI will continuously evolve and adapt to new strategies.

In summary, this process can be understood as the AI learning from you, capturing your skills, and ultimately competing for you. The better you perform as a trainer, the better your AI will be. Once a unique AI is created, it can participate in competitions. Players purchase, train, and battle AI-driven NFTs. Players compete to develop better training techniques to earn rewards.

Project Information:

X Account: https://x.com/aiarena_

(2) ASTO

In this project, players autonomously generate AI agents.

The AI agents generated by players consist of three parts: the brain NFT, which is the core of the agent; the appearance and operational methods of the agent; and memories stored in the brain memory tree, which encode the behavioral strategies learned by the agent. The ASM platform formed by this core technology provides the ability to create AI agents owned as NFTs for any application (e.g., games, metaverse, finance).

Players can train the AI agents they create in a "gym." The gym is a networked GPU cloud computing provider (similar to miners) that chooses to run training algorithms for specific ASM applications (e.g., popular games or trading robots in the DeFi market). Owners pay training fees using ASTO tokens.

Currently, three games driven by the ASM project are:

Artificial Intelligence Football Association (AIFA)

AIFA is a 4x4 football game autonomously conducted by ASM Brain-powered All-Stars. Players can collect and trade all-stars to build winning teams. Players will be able to train their ASM brains in the AI gym to enhance the skills of their all-stars and watch them evolve into metaverse football champions.

Muhammad Ali - The Next Legend

ASM collaborates with Muhammad Ali Enterprises and Non-Fungible Labs to create the world's first AI-driven metaverse boxing game. Players will combine trained boxers with any ASM Brain to train and compete for legendary titles, following in the footsteps of the greatest of all time, Muhammad Ali.

AI League: FIFA Women's World Cup AU·NZ·2023™ Edition

ASM collaborates with FIFA to bring a new era of street football into the metaverse. In the AI League mobile game, players start as managers of four completely unique AI all-star football players.

The aforementioned model of players creating AI to participate in competitions represents a significant advancement in competitive gaming. In traditional competitive games, players' ability to create characters is determined by game developers, limiting players' autonomy. However, in this new model, each AI character created by players is unique, making the competitive process closer to reality and greatly enhancing players' operational space.

Project Information:

X Account: https://x.com/altstatemachine

MCP (as of 2024.6.15): 16,703,261

FDV (as of 2024.6.15): 49,591,301

3. AI Promoting Cost Reduction and Efficiency Improvement in Games, Enhancing Game Quality

The cost reduction and efficiency improvement of AI in the traditional gaming industry mentioned in the first part also applies here and shows even more effective results compared to traditional industries. From the perspective of the games themselves, although Web3 games have begun to focus on enhancing gameplay and playability, high-quality game content in the market from 2021 to 2023 is relatively scarce.

According to Game7 Research data, currently, more than half of Web3 games are developed by independent producers or small teams, while AAA or AA-level productions account for only 6%. Some projects focus too much on technical and economic aspects, neglecting the refinement, innovation, and richness of game design. This may lead to player attrition and market saturation risks, limiting the long-term development of the industry.

At this stage, Web3 games are moving towards high-quality graphics, rich content, and excellent player experiences, even advancing towards AAA production, aiming to attract more Web2 players by enhancing game playability. However, before the emergence of generative AI, the enormous investment costs of high-quality AAA games deterred small teams in the Web3 space. The advent of generative AI has lowered the costs of game production, allowing more small teams to participate in the pursuit of high-quality games. From an industry-wide perspective, this round of GameFi trends will see the emergence of more high-quality games, with AI accelerating the emergence of truly valuable and playable games.

(1) EXVERSE

EXVERSE is a AAA-level first-person shooter game that combines intense shooting gameplay with innovative Web3 mechanisms built on Unreal Engine 5. EXVERSE employs various technologies to enhance the quality and realism of the game. It uses the virtualized geometry system Nanite for pixel-level detail rendering. It employs the Lumen fully dynamic global illumination and reflection system to present infinite reflections and indirect specular reflections in large detail environments ranging from millimeters to kilometers. It utilizes Chaos functionality to generate cinematic-quality destruction, shattering, and demolition of large-scale scenes. Chaos also supports dynamic static meshes, cloth, hair, vehicles, and rigid body animations, integrated with Niagara for auxiliary effects like dust and smoke.

Niagara is a visual effects system that enhances traditional particle systems. With Niagara, we will create impressive particle simulations in Expverse.

Realism and quality comparable to traditional games become the core selling points of this game. With the continuous empowerment of AI technology, more high-quality AAA-level Web3 games will be presented to the public in the future.

Project Information:

X Account: https://x.com/exverse_io

MCP (as of 2024.6.15): 1,484,312

FDV (as of 2024.6.15): 18,461,655

(2) KOMPETE

This game is also a AAA-level shooting-dominated game. It integrates multiple gameplay styles, placing players in a battle royale-style landscape filled with points of interest, racetracks, basketball courts, and golf courses. This free-to-play multiplayer game has no entry barriers and can be accessed on any device.

The highlight of this project is that it is an all-in-one multiplayer game featuring gunfights, racing, battle royale, basketball courts, golf courses, social deduction, and tournaments. Players gain a more diverse and richer gaming experience within the game.

Project Information:

X Account: https://x.com/KOMPETEgame

MCP (as of 2024.6.15): 14,536,828

FDV (as of 2024.6.15): 18,286,126

4. AI Agents Increasing Game Playability

Most games are incorporating AI agent technology to provide players with richer and more diverse gaming experiences. For example, dynamic NPCs can make non-player characters in games behave more realistically and flexibly, increasing the challenge and fun of the game. Additionally, players can receive personalized interactive experiences based on their behaviors and preferences. Every interaction between players and NPCs will alter the NPC's life trajectory, significantly enhancing the richness of the game.

(1) AI Hero

For example, AI Hero is a Web3 game that combines "RPG + Battle Royale + DND," generated with AI participation. Players first choose their race and profession based on their preferences and explore different storyline options through AI-generated plots combined with their attributes to earn equipment rewards and enhance their character attributes. During the competition, players will encounter other opposing players, and the last surviving player will win rewards. In the game, AI tracks the player's current character status and attributes, as well as the terrain, creating corresponding storyline narratives and development paths. For instance, if a player is a warrior, they may encounter a dragon in a volcanic terrain, facing combat challenges; however, if the player is a rogue, they might bypass the fight in the same scenario. During the initial training of the AI model, the player's character status and attributes are introduced, allowing AI to tailor the script for them. To avoid imbalances and uncontrollable factors, the AI directly provides players with storyline options instead of allowing players to advance the game through dialogue. With the help of AI agents and generative AI, AI Hero creates more possibilities for players.

Project Information:

X Account: https://x.com/binary_x

MCP (as of 2024.6.15): 389,304,902

FDV (as of 2024.6.15): 2,220,965,504

(2) Parallel Colony

Another project utilizing AI agents is Colony. In April 2023, researchers from Stanford University and Google published a paper known as the "Smallville Paper," detailing how they simulated human behavior by filling a sandbox digital realm with 25 AI agents. Inspired by this paper, Parallel Colony was born. In this game, NFT characters based on AI agents are the main participants, and players only need to guide these NFT characters in the game. The AI agents in the game are powered by Avatars, which are custom large language models (LLMs) fine-tuned by the Parallel team to help the agents better remember experiences and make decisions based on their game memories. As a result, other characters in the game world are not just non-playable characters (NPCs); they actually possess unique memories and can make autonomous decisions in the game world.

Specifically, the unique architecture of Colony's semi-autonomous AI avatars means that although players own personalized avatars, they must work collaboratively with their avatars through general instructions to guide them through various tasks. Users might suggest to the game, "Maybe you should plant an apple tree for future food?" NPCs will respond to the suggestions, and in some cases, they may interpret them based on other goals they might have—if they believe the player's suggestion does not align with their characteristics or conflicts with another goal they are pursuing, they may even refuse it. In such cases, owners may find themselves debating or persuading their avatars to complete tasks. Ultimately, the avatars will determine their intended goals—sometimes even ignoring their owners' suggestions—and begin pursuing what they believe to be the most beneficial tasks. Through AI agent empowerment, the game world becomes closer to the real world, enhancing the authenticity and playability of the game.

Project Information:

X Account: https://x.com/ParallelColony

5. AI Discovering User Needs to Increase User Stickiness

The Ultiverse project launched the first AI-driven open metaverse protocol, Bodhi, in early December. The Bodhi protocol serves as a supporting infrastructure and collaborative ecosystem integrator for Ultiverse's AI-driven collaborative world vision. Targeting players driven by different gaming motivations, Ultiverse's Bodhi protocol can mine user preferences and segment player groups using AI, feeding player data back to another AI engine based on multiple real-time algorithms. This allows for real-time adjustments based on player preference data, providing the best experience for players.

AI is the underlying driver, and games are the practical scenarios. AI + Web3 breaks the limitations of traditional games, offering players more personalized experiences and achieving refined operations.

Figure 33: Ultiverse Project Roadmap

Source: "AI Assisting Games, Can Ultiverse Lead a New Trend in GameFi?" Dot Labs

Project Information:

X Account: https://x.com/UltiverseDAO

MCP (as of 2024.6.15): 38,988,211

FDV (as of 2024.6.15): 278,487,225

6. AI Improving the Sustainability of Token Economies

Zara Finance

Zara Finance is becoming a leader in the industry with its unique AI GameFi Pool operating mechanism and focus on the GameFi and NFT niche market. Zara Finance combines the concept of DeFi's vaults with GameFi's guild strategies to develop a complete AI strategy system. This system not only enhances the effective allocation of resources but also optimizes returns for players and investors, further promoting the integration of blockchain gaming and finance.

Zara Finance utilizes AI technology to filter high-quality WEB3 products, maximizing benefits for users. Its products, such as Z-Swap, GameList, and Slaunch, not only increase user engagement but also enhance the interactivity and sustainability of the entire ecosystem. Particularly in the design of GUSD algorithmic stablecoins and Zara governance tokens, Zara Finance demonstrates advanced thinking in financial engineering, providing users with more stable and trustworthy investment options.

Project Information:

Twitter: https://x.com/ZaraFinance

7. Providing Infrastructure for AI Empowering GameFi

(1) KARRAT

KARRAT is a decentralized gaming infrastructure layer supported by $KARRAT and backed by a truly decentralized community with a shared vision, including catering to the new era of gaming, entertainment, and AI products. The KARRAT protocol aims to support the gaming and entertainment industries as well as transformative AI innovations in these sectors.

Specifics include:

Funding community projects and rewards

Developing infrastructure solutions for gaming and entertainment ecosystems

Developing story films for Web3 native IPs

Establishing alliances with esports organizations

Partnering with gaming and entertainment studios for their title projects

Building content delivery networks

Facilitating innovation challenges

The highlight of this project is its backing studio, AMGI Studios, which produces entertainment IPs, games, and Web3 native assets through proprietary technology, AI, and real-time animation innovations. AMGI Studios was founded by talents from Pixar, DreamWorks, EA, and Amazon, and has a diverse team of over 60 creative personnel and developers. AMGI is supported by Epic, Netflix, Coldplay, Polygon, as well as Tony Robbins, Marc Cuban, and Eric Yuan.

Project Information:

X Account: https://x.com/karratcoin

MCP (as of 2024.6.15): 81,821,582

FDV (as of 2024.6.15): 819,161,582

(2) GameGPT

The GameGPT project technically supports game production. GameGPT is an AI-driven game engine that powers the next generation of blockchain games. GameGPT can integrate multiple AI agents to automate parts of the game development process. Different agents perform specific roles, working in an organized manner. Some agents review game design plans and make corresponding modifications and adjustments; others convert tasks into specific code; some check the code generated in the previous step and review the results; and there are agents responsible for verifying whether all work meets initial expectations.

By refining and breaking down workflows, GameGPT simplifies the work of AI agents. This division of labor is more efficient and easier to implement than having a single omnipotent agent do everything.

First, in the game design phase, upon receiving user requests, GameGPT's tasks include generating the entire development plan for the game. This planning phase is one of the critical steps that greatly influences the seamless progress of the entire development process. This phase is planned by an LLM-based game development manager, who proposes an initial plan, which is then broken down into a task list.

It is worth noting that due to the inherent limitations of LLMs, this initial plan often suffers from hallucinations, leading to unexpected tasks, including irrelevant or unnecessary redundant tasks. To address these issues, researchers have proposed four strategies that can mitigate these challenges, which are orthogonal to each other and can be executed in layers for better results.

Figure 38: GameGPT Project Architecture

Source: "GameGPT White Paper," Dot Labs

Project Information:

X Account: https://x.com/Gamegptofficial

MCP (as of 2024.6.15): 15,751,978

FDV (as of 2024.6.15): 95,654,353

III. Risk Warning

Risk 1: Price Volatility

• Cryptocurrency prices are highly volatile, and future prices cannot be guaranteed or predicted.

Risk 2: Financial

• Projects may go bankrupt due to poor management.

Risk 3: Underperformance

• The overall development of the GameFi sector may fall short of expectations.

Risk 4: Legal

• Some countries and regions prohibit such activities, hindering development.