Researchers from the VUB’s Artificial Intelligence Lab and the University of Namur (UNamur) display in a new large-scale study how autonomous AI systems can spontaneously develop a functional, shared language without human intervention.
Unlike large language models together with ChatGPT, which memorize existing human language, those AI agents construct their very own communication system, directly rooted of their particular sensors and environment.
This discovery signifies a step forward for communication among autonomous systems, together with self-driving cars, robots and industrial sensors.
The research is published in the journal Transactions of the Association for Computational Linguistics.
A layer of abstraction for nonhuman senses
Recent AI language models focus on understanding and generating human languages including Dutch or English. These human languages are traditionally rooted in our personal sensory perceptions. When physical or virtual AI agents are equipped with sensors for which humans have no direct sensory equivalent—together with lidar, sonar, infrared or chemical analysis—existing human language proves insufficient.
Jérôme Botoko Ekila, a doctoral student at the VUB’s AI Lab, describes, “Rather than of forcing AI systems to communicate in a human language that does not correspond to their sensors, we permit them to develop their own ‘artificial natural language.’ They construct effective concepts that correspond exactly to what they understand through their sensors, even if the data is incomprehensible to human.”
As of sensors on different machines or vehicles can never be calibrated same in practice, the researchers consider that this language gives the perfect solution.
Paul Van Eecke, a professor of artificial intelligence at the VUB, mention, “In industrial or physical applications, it’s simply unrealistic to anticipate all of the sensors on different robots or vehicles to be calibrated in exactly the same way.
“Humans solve this issue through language: if I say I am standing near to the ‘red car,’ we understand each other without our eyes registering precisely the same colour values. Language gives a layer of abstraction above the sensor readings, and that is exactly what we’re now permitting those AI agents to construct autonomously.”
Self-organization and evolution
Rather than of preprogrammed words or fixed codes, the researchers offered the agents with the basic mechanisms underpinning the evolution of human language: the local invention of words, the adaptation of conceptual prototypes and the incorporation of feedback following a successful or failed communication.
Katrien Beuls, a professor at UNamur, describes, “What’s fascinating is that there’s be no central computer imposing the rules. Each time, we let two sellers communicate with each other at a local level. It is solely via these individual interactions and updates that a worldwide convention emerges spontaneously across the whole population.
“This principle of self-organization permits the language to optimize itself automatically: the agents forms precisely the number of categories required to carry out their tasks efficiently, without superfluous information.”
Tested on 37 data sets
The universal applicability of the method has been showed throughout 37 varied and complex data sets. The agents successfully developed their own vocabulary to communicate, drastically regarding physicochemical analyses of wine, NASA data on exoplanets, visible data sets and images of birds or faces. In all these situations, the agents successfully synthesized the data into new, shared concepts.
Transparent and adaptive
Transparency is a remarkable benefit of the systems developed. Unlike many “black-box” AI models, these systems permits researchers to precisely decide the which means of a mainly developed word, as the prototype can be directly connected to the values of the underlying sensors.
Furthermore, these evolving linguistic systems are robust. When the environment changes, when sensors produce noise or when individual agents have a different structure, the language automatically adapts to the new circumstances through continuous interaction.
