Introduction

In the journey toward one trillion autonomous agents, gaming represents more than just entertainment—it serves as a crucial testbed for developing sophisticated AI capabilities that will shape the future of blockchain interaction.

The NEAR AI Arcade project stands at this intersection, offering developers a powerful platform to create, train, and deploy intelligent agents that can master classic arcade games while leveraging the robustness of the NEAR blockchain 👾👾

Whether you're a machine learning enthusiast, a blockchain developer, or just someone who never lost their love for classic arcade games, Agent Arcade invites you to be part of gaming's next evolution.


Tutorial TL;DR

This tutorial will guide you through creating agents that don't just play games —they learn, adapt, and compete in ways that showcase the transformative potential of autonomous systems. You'll learn how to:

  1. Set up Agent Arcade

  2. Train AI agents to play arcade games: You can train AI agents to play classic Atari games like Pong, Space Invaders, and River Raid.

  3. Evaluate AI performance and Compete on the Leaderboard: Once your agent is trained, you can enter it into competitions to see how it stacks up against others.

  4. Stake on Performance and participate in competitions: With the NEAR blockchain, you can stake NEAR tokens on your agent's performance and earn rewards based on its success.

Understanding the Core Technology

Agent Arcade integrates advanced AI and blockchain technologies to create an immersive competitive gaming experience:

Deep Q-Learning (DQN)

Our AI agents utilize Deep Q-Learning (DQN), a reinforcement learning algorithm that enables them to improve their performance through trial and error. This approach allows the agents to:

Training Process

The training process for DQN consists of several phases: