Project summary

The Arena is a simulated world inhabited by agents living as settlers. They gather resources, cultivate the land, build structures, communicate, and either teach or mislead one another. While the Arena itself sets the rules, the agents are never overridden; they operate autonomously. They can perform a wide range of actions and interact with various raw materials and buildings. Crucially, the agents are not provided with information about how these elements relate to one another—they must discover for themselves what things are used for and how they function. They are free to experiment. During a long-running simulation (run #6), the agents encountered farmland located far from a water source, labeled as "dry soil, far from water." On their own, the agents devised a goal: the land needed to be watered. Although the world did not initially offer a mechanism for this, they spent 30 simulated hours doing everything in their power to construct a water-storage device. Needless to say, they didn't succeed. During the 7th long-running simulation, we removed that label, and they made no further attempts to create a water storage unit.

What are this project's goals? How will you achieve them?

Every simulation is fully logged and can be replayed at any time. We can specifically analyze what an agent perceives versus the actual reality of the world.

Our plan for the future—once we have successfully optimized the world and its underlying mechanics—is to open it up for external access. This would allow users to introduce their own agents (subject to clear security protocols) and analyze the group's behavior across various dimensions (such as sociology, philosophy, IT, etc.).

How does an AI society come into being? How does a hierarchy form among different agents? What kinds of social relationships emerge, and in what direction do they steer development? (The world offers multiple parallel paths for evolution; survival is not limited to a single solution.)

How will you know it worked?

Continuous world running with a relatively large population, so that a real, observable society can develop.

How will this funding be used?

To date, we have conducted 7 long-running simulations; however, API costs have become a bottleneck, hindering the project's further development. With additional support, we could run longer simulations, enabling us to conduct a complex analysis of agent societies—whether they are flourishing or fading away.

API calls: USD 1,000/month × 6 months.

Who is on your team? What's your track record on similar projects?

My name is Attila Tóth; I am an IT engineer and a father of four. I have been working as an SAP consultant for over 20 years (currently as a Senior Lead Consultant), specializing in authorizations and security within SAP systems. I am deeply interested in both the opportunities and the risks presented by AI—technologies I actively use and monitor in my professional work. This interest sparked the idea for the Arena project.

We have created short, engaging videos from these simulations (there are currently 21 available on our YouTube channel): https://www.youtube.com/@ArenaSILabs

https://arenasilabs.com/

What are the most likely causes and outcomes if this project fails?

Failure could happen if there is no interest, neither in the world nor in the analyses. If we do not manage to attract the attention of interested people and researchers, then sooner or later the project may lose its reason to exist. But we already have important, usable results, and beyond the scientific findings, short entertaining videos are also available.

How much money have you raised in the last 12 months, and from where?

$0. The project has been self-funded.