Skills trainers

For an NFT player to learn a new skill, it must undergo training using a machine learning algorithm known as deep reinforcement learning.

While each NFT collection has the freedom to develop its own trainer for specific or unconventional skills, we will be providing trainers for common tasks such as seeking a target, hand-to-hand combat, etc. These trainers can be utilized by multiple NFTs, simplifying the process for NFT creators to publish new collections. As these trainers will be open-source, they can be customized and enhanced as desired.

The game developers can also utilize their game engines to train specific skills required for participation in their game's matches. They could then create their own NFT collection or sell/distribute the training data (the neural network parameters) to be used by a third-party NFT creators.

Currently training must be done off-chain until Internet Computer add support for GPU subnets. When on-chain training becomes feasible, we also plan to enable users to teach their NFTs how to perform the tasks, allowing the human assistance to create more advanced AI players.