Communication with aliens could one day become much easier if we study how AI agents talk to each other.

In the 2016 sci-fi film Arrival, a linguist is faced with the daunting task of deciphering an alien language made up of palindromic phrases that read the same backwards and forwards and are written in circular characters. As she discovers different clues, different countries around the world interpret the messages differently – some believing they pose a threat.

If humanity found itself in such a situation today, the best option would be to turn to research that reveals how artificial intelligence (AI) develops languages.

But what exactly defines language? Most of us use at least one to communicate with people around us, but how did this happen? Linguists are thinking this same question for decadesbut there is no easy way learn how language developed.

Language is ephemeral; it leaves no noticeable traces in fossil records. Unlike bones, we cannot excavate ancient languages to study how they developed over time.

While we may not be able to study the true evolution of human language, perhaps modeling could provide some insight. This is where artificial intelligence comes in, a fascinating area of research called emergency contactwhich I have spent the last three years studying.

To model how language might evolve, we give agents (AI) simple tasks that require communication, such as a game in which one robot must guide another to a specific location on a grid without showing it a map. We provide (almost) no restrictions on what or how they can say—we simply give them a problem and let them solve it the way they want.

Because solving these problems requires agents to communicate with each other, we can study how their communication evolves over time to gain insight into how language may evolve.

Connected: A father-daughter team has deciphered the ‘alien signal’ from Mars that has been baffling the world for a year.

Similar experiments were carried out on humans. Imagine that you, an English speaker, are paired with a non-English speaker. Your task is to instruct your partner to pick up a green cube from a variety of objects on the table.

You can try drawing a cube shape with your hands and pointing at the grass outside the window to represent the color green. Over time, you will develop a kind of proto-language together. Perhaps you could create special gestures or symbols for “cube” and “green”. Through repeated interactions, these improvised signals will become more accurate and consistent, forming a basic communication system.

This works similarly for AI. Through they learn by trial and error communicate about the objects they see, and their interlocutors learn to understand them.

But how do we know what they are talking about? If they only develop this language with their artificial interlocutor and not with us, how do we know what each word means? After all, a particular word could mean “green”, “cube” or, worse, both. This problem of interpretation is a key part of my research.

Cracking the code

The task of understanding AI language may seem almost impossible at first glance. If I tried to speak Polish (my native language) with a colleague who only speaks English, we would not be able to understand each other or even know where each word began and ended.

The problem with artificial intelligence languages is even greater because they can organize information in ways that are completely foreign to human language models.

Fortunately, linguists have developed difficult tools using information theory to interpret unknown languages.

Just as archaeologists piece together ancient languages, we use patterns in AI conversations to understand their linguistic structure. Sometimes we find amazing resemblance human languages, and sometimes we discover completely new ways to communicate.

A pixel art illustration of the hands of Michelangelo's Adam.

AI develops its own languages. (Image credit: Cybermag via Shutterestock)

These tools help us peer into the black box of AI communication, showing how artificial agents develop their own unique ways of exchanging information.

My recent work has focused on using what agents see and say to interpret their language. Imagine having a transcript of a conversation in a language you don’t know, and what each speaker was looking at. We can match transcript patterns to objects in the participant’s field of view, building statistical relationships between words and objects.

For example, perhaps the phrase “yayo” coincides with a bird flying by – we might have guessed that “yayo” is the speaker’s word for “bird.” Through careful analysis of these patterns, we can begin to decipher the meaning of communication.

IN last article My colleagues and I, appearing in the proceedings of the Neural Information Processing Systems (NeurIPS) conference, show that such techniques can be used to reverse engineer at least parts of AI language and syntax, giving us insight into how they might structure communication.

Aliens and autonomous systems

How does this relate to aliens? The techniques we are developing to understand AI languages could help us decipher any future alien communications.

If we can obtain written alien text along with some context (such as visual information related to the text), we can apply the same statistical tools analyze them. The approaches we are developing today may become useful tools in the future study of alien languages, known as xenolinguistics.

But we don’t have to look for aliens to benefit from this research. Eat numerous applicationsfrom improving language models such as ChatGPT or Claude, to improve communication between autonomous vehicles or drones.

By deciphering new languages, we can make future technologies easier to understand. Whether we know how self-driving cars coordinate their movements or how artificial intelligence systems make decisions, we’re not just creating intelligent systems—we’re learning to understand them.

This edited article is republished from Talk under Creative Commons license. Read original article.