🧠AI and Machine Learning

Core Feature

The need for Smart AI

Artificial Intelligence is growing at an exponential rate, and will become an integral part to technology. Descend Online aims to be amongst the first-movers to adopt machine learning in our gameplay, especially within our mobs. This provides a unique experience in each mob encounter, and allows for an unpredictable and fun gameplay experience.

This is why we are training our own set of ML Agents to interact with the world and ensure NPC behaviour changes over time. This will add more pressure to the game economy and make the game much more entertaining. We will be keeping a close eye on player competence and how well they are at engaging with our AI-assisted mobs in order to keep the game balanced.

Some AI will be pretrained, some won't require any training at all. Below is an example of model-free learning currently in development.

Rude Boars (an example of ML agents)

This is a Q-learning algorithm, a type of model-free machine learning method. It's specifically designed with tailored weights to guide a boar to evade the player. The core concept of this algorithm is to make the boar highly adept at dodging attacks when faced by players.

To efficiently farm boars, players may need to collaborate with others in order to corner the beast, making room for social dynamics and mechanical mastery. P.S. boars drop resources vital to crawling dungeons!

Other potential use-cases for ML Agents

  • Incrementally increasing difficulty within the dungeons whilst not making it too difficult for the player so that they don't quit the game (completely)

  • Chatting with other players

  • Dynamic questlines which respond towards what a player wants to see

  • Boss fights that get harder over time

  • Placing game content in correct areas which increases user retention over time

  • Pets that respond to how you play and develop behaviours of true companionship

  • And many many more!

Last updated