Creating a civilization from scratch is no easy task: form a government, design an economy, develop a culture, and establish a religion. This has been the case for the Sumerians of Mesopotamia and the agents of Minecraft. Civilizations are born and fall, following the same rules. There are no secrets, just the challenge of building or collapsing an entire world.

The startup Altera, led by former MIT professor and neuroscientist Dr. Robert Yang, has conducted one of the most fascinating simulations in recent years. The project, called Sid, involves deploying 1,000 AI agents on an isolated Minecraft server. The aim is to simulate their collaboration, the coherence of their actions, and their long-term evolution within the video game’s controlled rules.

Then the agents began to play without prior knowledge of the environment, and within minutes a small group of 49 agents organized themselves and started to collect around 300 objects.

Later, they created a market and agreed to use gems as a medium of exchange. Step by step, transactions began to form among the agents, and before they knew it, they had created an economy. Interestingly, according to the company’s statements, it was not the merchants who traded the most, but a corrupt priest named “PastaPriest” who started bribing the villagers to convert to his religion.

They recreated this simulation over several days, and the results were always different. In another simulation, an AI agent named Olivia worked as a farmer providing food for an entire civilization until she heard the stories of the village explorer, the agent Nora, which inspired her to embark on her own journey and leave the village. However, Olivia gave up on the idea after other villagers pleaded with her to stay and focus on the harvest, prioritizing the common good over her “dream.”

In another simulation, two parallel civilizations were tested: one led by Donald Trump and the other by Kamala Harris. Both received the same constitution, which they could vote to amend.

The AI agents voted in both scenarios, but they made different decisions. The society under Trump decided to increase the number of police officers, while the one led by Harris focused on reforming the criminal justice system and abolishing the death penalty.

Another day, the simulation was different; suddenly, two villagers disappeared, and a group of agents agreed to devise a plan to light torches throughout the village and create a beacon for the missing to return to.

According to Yang, multiple agents can be organized on an “unprecedented” scale and achieve what individual agents cannot. This project demonstrates that agents can collaborate and forge alliances in the long term.

When agents socialize, they influence each other through group dynamics but also use individual power to change the system.

Altera seeks to create a truly autonomous AI with agents capable of operating independently, progressing independently, and collaborating over long periods. These agents will also consider human interests and be able to express their thoughts and emotions verbally.

According to their website, the company’s goal is “to create digital humans who live, care, and grow with us.”

This means that machines will need to be more autonomous, but they cannot simply be trained to act like humans; their deeper motivations must be aligned with human values.

The transition of Minecraft to the real world.

In the controlled simulation of Minecraft, the agents built a prosperous civilization based on cooperation, collaboration, a sense of community, and the common good. Aside from the false positive from the village priest, the simulations turned out quite well, much better than expected. However, we must understand that the rules in Minecraft are known, and the programmers who designed those agents made sure to instruct them in what we can call positive and universal human values.

In the next two or three years, we will likely reach a highly advanced state of the art in generative technology that will allow us to have advanced AI agents in the real world, and they will not be simulations in a video game. Throughout the year on the blog, I have been telling you how they would be and what they would do, and last week, the first seed was planted with OpenAI’s agent-based model, called o1. Recently, Microsoft’s suite launched its marketplace for customizable agents, and we are witnessing a revolution of advanced tools in what will be the first evolved AI agent that emulates a software development engineer 100%.

What makes me think about what will happen when that simulation is deployed in the real world, where agents will interact with each other and with us: Humans, general-purpose agents, specialized agents, orchestrators, and replicant agents (my favorite), what will that world that is being born be like? How is it going to affect us on all levels? What security measures should we take? It would make sense to have a Universal Declaration of Agent Rights, which is based on and fully aligned with the Universal Declaration of Human Rights values.

The risk exists and is inherent; Altera’s programmers have instilled good and universal values in their agents, but this could be reversed, and the same technology could fall into the wrong hands, allowing another organization to do just the opposite, potentially creating chaos in an agent-based economy.

Ray Dalio and his replicant agent.

While preparing this article, I received an email in my inbox that left a mark on me. It was an invitation to validate a proof of concept for Ray Dalio’s AI agent, an American investor and philanthropist I have followed for years, founder of Bridgewater Associates, one of the most significant hedge funds in the world, and a highly knowledgeable person in many disciplines.

And I thought, “Can you imagine being able to hire Ray Dalio as a consultant for my company? It would be amazing. I could access your digital version at any time to ask for your advice directly.” At that moment, a world of possibilities opened up for me about how the economy will function when we have AI agents that replicate experts.

We all know our physical limitations as humans: We cannot be in more than one place at the same time or have multiple conversations simultaneously. Our interactions are one-on-one, sometimes one-to-many, when we speak in front of an audience, but there is always a limit.

I have been thinking a lot about Ray Dalio’s replica agent. Ultimately, it will just be software that works on any device, available to any user, anywhere and anytime. I also considered their pricing model: imagine paying €100 monthly to access this replicant. Ray has a large following, and he could quickly gain 10,000 subscribers who use his replica in the operations of his companies. That would mean that the product could generate 12 million euros a year (100 € x 10,000 x 12 months).

I have no idea how much Ray Dalio currently earns from his in-person advisory service; it might be one of his smaller sources of income. However, their AI replica could generate at least x10 times more. Additionally, it could offer new services and products that do not yet exist. When Ray is no longer around, his family can continue to receive income from his agent.

The final reflection: AI agents, both general-purpose and those specialized in specific professions or tasks, will likely replace many of us who work in knowledge-based occupations. There will be everything: agents that replace and others that enhance our capabilities, depending on the type of profession and tasks. We must be cautious, as this will open up different scenarios.

We face uncertainty until these technologies mature and spread throughout society and the economy. But it’s not all said and done yet. After that period of adaptation and adjustment, when these technologies reach a more advanced state, I feel that we could enter the most significant era of economic prosperity in history. If not, ask Ray|yaR.

Pedro Trillo

CEO at Vizologi | Website

Pedro Trillo is a tech entrepreneur, telecommunications engineer, founder of the startup Vizologi, specialist in Generative Artificial Intelligence and business strategy, technologist, and author of several essays on technology.

Share: