SevenX Ventures: An Overview of the Next Generation Web3 Game AI Arena
Author: SevenX Ventures
One-Sentence Introduction
AI Arena is a PVP fighting game where players can learn through AI and continuously evolve their game characters. The combat mode is similar to Nintendo's "Super Smash Brothers." Game DEMO.
Features
The world's first fast-paced fighting game that combines AI and NFT, where players enhance character abilities by training AI, adjusting character stats in the observation room, and then entering battles in the arena to achieve higher rankings.
Vision
To combine games, NFTs (IP), and AI algorithms, providing players with a new gaming experience while protecting the intellectual property of AI practitioners and helping them monetize. The long-term vision is to create a bilateral market for AI, facilitating the needs of buyers and sellers.
NFT Asset Performance
AI Arena Mint Pass --- 4.11 ETH, with a 215% increase in the past 30 days (as of: 2024.01.09 Data source: https://opensea.io/collection/ai-arena-mint-pass)
Investors
Paradigm; Framework Ventures; SevenX Ventures, etc.
Project Interpretation
1. The Intersection Narrative of AI + Web3
In a previous research article on ZKML, SevenX mentioned that decentralized blockchain is an important force in balancing AI (and machine learning). It can combine with other technologies, such as ZK, to optimize the trust framework of machine learning, and effectively utilize long-tail resources to reduce the cost and threshold of using AI. On the other hand, many Web3 applications sacrifice user experience for security and decentralization, while AI can help optimize and enhance user experience, which is part of how AI can empower Web3.
In terms of practical application scenarios, although there are use cases for AI + DeFi and AI + DID/social, the combination of AI and Web3 games may be the first area to achieve an expanded user base.
2. Innovative Gameplay Enhancing User Experience
- AI Arena is not merely a deep application of AI in a Web3.0 game, but aims to allow more users to access, understand, and experience AI through Web3.0 games, enabling AI engineers to provide various AI algorithms based on AI Arena, thereby helping them increase their income.
- Players first create their own AI character (the character does not need to be an NFT; NFT characters can earn rewards. Players can purchase NFTs on the Marketplace or train AI to participate in high-quality competitions to win NFTs), and then start training their character in training mode. Players can choose a training venue they like to spar with robots (training), teaching their AI character how to fight through combat.
(Data reception toggle off)
(Data reception toggle on)
- During the battle, the AI character will be covered with red lines, indicating that the player's AI character has not started receiving data (the character has not been trained). Therefore, players should turn on the data reception toggle before officially training the AI to ensure the AI can receive data; at this point, the AI will be covered with blue lines (this design is to avoid the AI receiving unnecessary combat data, which could lead to suboptimal training results).
- After completing data collection, players need to set the AI's training parameters, including: training intensity (number of iterations per training, batch size, learning rate, direction function, action function), focus areas, and data to be cleared, and then start training the AI character.
- AI Observation Room: After training is complete, players can check the training results through the AI observation room and choose to save or discard the model. If players choose to save the training model, a metamask popup will prompt them, and upon confirmation, the model will be saved on IPFS with the hash stored on the blockchain.
- Once training is complete, players can battle against system robots in simulation mode to test their training results.
3. "Merging Pool" --- The Economic Flywheel of AI Arena
The economic growth flywheel of AI Arena is driven by both Token and NFT. It mainly involves two types of users: players and researchers.
Players
- Players train AI characters to participate in ranked matches. Entry requires staking $NRN (equivalent to an entry fee) and has the chance to earn more $NRN rewards (AI Arena regularly distributes rewards). Players who fail to win will lose a portion of their stake.
- Before entering a ranked match, players can choose to stake a certain percentage of the $NRN prize they win in that round into the merging pool. The more prize money transferred to the merging pool, the higher the probability of winning an NFT.
- Players can use $NRN to purchase batteries to charge NFTs and then fight in ranked matches.
Researchers
- Stake $NRN to participate in research competitions (AI algorithm models) for a chance to earn more $NRN rewards.
Merging Pool
- The main function of the merging pool: to give players a chance to obtain more powerful NFT characters.
- The system selects NFTs staked in a specific round of ranked matches to compete against each other. Through competition, the system can determine which NFTs to merge to create stronger NFTs. The system uses a fitness function to score the strength of NFTs. Once the fitness scores for each NFT are obtained, the system can use a roulette method to select which NFTs can merge. The higher the fitness score, the higher the probability of being selected for merging. Once merging is complete, a new batch of NFTs is generated, which will compete again. Each new competition in the genetic algorithm is referred to as a generation. Over generations, NFTs will evolve into increasingly powerful NFTs.
- NFTs generated from the "merging pool" are different from the initially dropped NFTs. These NFTs are generated through genetic algorithms and inherit the skills of the NFTs submitted to the pool. This ensures that these NFTs as rewards have high economic value. Additionally, NFTs in the merging pool are trained, while the initially airdropped NFTs are untrained.
- One of the core functions of the merging pool is to control NFT inflation: if players obtain more powerful NFTs from the merging pool, the old NFTs will be automatically destroyed.
Potential Risks
The way AI learns is through active learning, requiring a lot of human intervention. Although the gameplay setup of AI Arena can stimulate users and increase their engagement in the game, effective use of Web3-specific incentive mechanisms is needed to quickly expand the player base and influence during the early stages of customer acquisition.