Swarm Intelligence Cup

AI through Emergence

SIC Resuming later this Year!

Stay informed:

SIC Waitlist

Intelligence from Swarms

With the Swarm Intelligence Cup, we want to explore the role of collective behavior in designing intelligent systems. Achieving human-level AI requires crucial elements that are currently absent from state-of-the-art AI systems. A prime aspect of human intelligence, and life on Earth in general, is the manifestation of complex behavior through countless simple cells – just like the human brain, where neurons, in their individual simplicity, work together to produce the rich tapestry of human intelligence.

Your Mission

The Swarm Intelligence Cup is designed to explore the role of collective behavior in designing intelligent systems. Spark the beginning of a new form of intelligence by designing the agents of your swarm. Only swarms capable of exploring their environment and learning how to exploit its features will flourish. It is your challenge to form a surviving swarm!

First SIC Disclipline

A Game of “Swarm Deathmatch”

Future Discipline

A Game of “Supplies Hunt”

Join the Public Beta!

November 10 - December 17, 2023

Community-Driven

We’re excited to unveil the Swarm Intelligence Cup (SIC) — a groundbreaking challenge in the domain of AI, diverging from typical machine learning approaches. With a public beta round, our intent is to merge the creative talents of the AI community to establish a dynamic, community-driven project!

Open-Source

At the heart of the competition is a swarm simulation framework coded in Python that is open-source and available on Github. Participants can script their own agent behaviors and guiding their swarm within the simulation. In order to establish a sound Cup framework, the goals of the beta are to:

  1. Develop Unique Agent Behaviors: Flex your coding muscles and invent fascinating, efficient swarm behaviors based on the many skills available.
  2. Creating benchmark swarms: Develop and refine powerful swarm capabilities, setting benchmarks for future agent-based modeling innovations.
  3. Enhance the Simulation: Collaboration is key. We invite you to actively refine, optimize, and expand the simulation framework.

Win a Christmas Present!

1st Place

2nd Place

3rd, 4th & 5th Place

Cup Format

The goal of SIC is to learn how simple agents – like processing nodes in a neural network – become intelligent when operating as a collective. Since the inner workings of today’s neural networks remain largely opaque, particularly in researching the emergence of intelligence in depth, we delve into the realms of swarm intelligence. The SIC framework is specifically built to retrace what properties make swarms truly intelligent.

The SIC Beta Phase with the first game mode called “Swarm Deathmatch” was completed in December 2023. The SIC main tournament will entail numerous disciplines which all have to be mastered by a single swarm from each participating team. In the meantime, the Quick Start Guide and our Github repo below stay open for you to try out some swarm compilations. Stay tuned for what’s to come in 2024!

Swarm Deathmatch!

In this contest, swarms go head-to-head to see whose agents are best at adapting to their environment and utilizing a diverse skill set: Attack & defend, merge items for upgrades, manage energy resources, reproduce, communicate, and more! Find out about all swarm capabilities in our Cup documentation.

Quick Start Guide

Collaborative Ethos

As a participant, you will be granted access to the swarm simulation code on our Lab42 Github repo where you can sript the behavior of your agents in Python. While the competition-spirit is upheld, we encourage sharing insights and working together to bolster the simulation. Your feedback is essential to improve both the simulation framework and the world specifications over time! 

SIC Github Repo

Join the Lab42 Community

The Lab42 Discord server is a place for all AI enthusiasts, challenge participants and experts from different fields around the globe. Collaborate, discuss and exchange insights across disciplines!

Submission License & Data Use

Open Source Code under Apache 2.0 license.

 Any purpose, licensed under Apache 2.0.

Your Contact

Hansueli Jud

AI Engineer at Lab42

For Humankind

Lab42
Villa Fontana
Obere Strasse 22B
7270 Davos, Switzerland

Lab42

Powered by

For Humankind

Lab42
Villa Fontana
Obere Strasse 22B
7270 Davos, Switzerland

Lab42

Powered by

For Humankind

Villa Fontana
Obere Strasse 22B
7270 Davos, Switzerland

Lab42

Powered by