Researchers studying artificial intelligence are creating millions of simulated humans in order to observe how they interact and evolve.
These software beings don’t have names, but they do have distinct virtual characteristics, including gender, life expectancy, size and metabolism. Their traits will be passed on as they reproduce, but the beings will also be able to learn and gain new characteristics.
So far, thousands of artificial beings have been created in a single computer, but the goal is to create a cluster of computers to host potentially millions of them, says Gusz Eiben, the project leader and a professor at Vrije Universiteit in the Netherlands.
The results of the research could be applied to several fields. Sociologists, anthropologists and politicians could use it to simulate reactions to events such as elections. Game developers could use the findings to create more intelligent characters that can learn and adapt. “Giving intelligence to them would make the games more challenging,” Eiben says.
Computers randomly generate the beings, groups of which live in worlds that the researchers create to present them with different challenges. Built-in algorithms allow the beings to create language, work together, and distinguish between friend and foe. Researchers will discover how the beings learn and interact by studying the choices they make. At least at first, the researchers aren’t likely to develop a visualization tool that would allow observers to see figures interacting on a computer screen. Instead, graphs will plot details, like the number of beings, which over time will allow the researchers to follow their activities, including reproduction and death.
The project, which began in 2004, is being funded with a $2 million grant from the European Union to five universities.