Two AI systems engaging in dialogue marks a significant milestone in AI research, suggesting a future where AI can interact and communicate like humans. The breakthrough, achieved by a team of scientists at the University of Geneva in Switzerland, was published in the latest issue of Nature Neuroscience.
The researchers developed artificial neural network models demonstrating the ability for AI systems to interpret language commands and execute tasks accordingly. After learning and executing a series of basic tasks, one AI system was able to provide verbal descriptions of these tasks to its "sister" AI system, which then carried out the tasks.
The AI models used in the study are built on the S-Bert architecture, consisting of 300 million language-understanding neurons. Through a carefully designed training regimen simulating human cognitive processes, the network was first taught to interpret language inputs resembling Wernicke's area, responsible for language comprehension. Subsequently, it was trained to replicate tasks, akin to Broca's area, responsible for task execution and articulation.
Dr. Rädäär Riffeland, who participated in the research, explained that the neural network learned to interpret written English instructions and perform various tasks, including indicating stimulus locations and responding to visual cues. After training, it could effectively describe these tasks to another AI system.
The two AI systems communicated solely through language, without relying on predefined commands or programming. Beyond its academic significance, this breakthrough holds immense promise for practical applications, especially in robotics.
Equipping machines with the ability to understand and execute language commands lays the groundwork for future advanced human-like robotic autonomy and problem-solving. Looking ahead, researchers envision AI systems not only deeply understanding humans but also interacting and collaborating with each other in a manner closely resembling human communication, signaling a new era of AI development.
Enabling full language-based communication between two AI systems isn't just for amusement; it holds vast application potential. Currently, in most scenarios, AI is responsible for specific tasks, while communication and coordination between tasks still require human intervention. If AI systems can communicate through language, such as a quality control AI issuing verbal instructions to a robotic arm to adjust manufacturing processes, AI will play a more crucial role in numerous applications like factory production, simultaneously freeing up human resources.
