Researchers from Microsoft have developed a new language called Droidspeak. This language allows AI agents to communicate with each other 2.78 times faster than before, with minimal loss of accuracy. AI agents are advanced models capable of executing complex, multi-step tasks autonomously. Unlike AI bots, which act as tools responding to requests, AI agents can prioritize and complete tasks independently without further interaction.
AI agents typically use natural language, like English, for communication. However, human languages are not ideal for machine communication. The new method developed by Microsoft allows agents to communicate in a mathematical language underlying large language models (LLMs). This approach offers a significant advantage. When AI agents communicate in natural language, they not only share the result of their current task but also the entire conversation history. This back-and-forth exchange can lead to communication delays, limiting the scalability and responsiveness of multi-agent systems.
The researchers’ new approach enables models to share data directly from the computation steps before language generation, bypassing natural language. This new language, Droidspeak, is inspired by the beeping and whistling language of robots from the Star Wars series. However, Droidspeak currently has limitations. It only works for communication between versions of the same underlying LLM. For broader practical use, a method allowing models of different sizes and configurations to communicate would be necessary.
According to Philip Feldman from the University of Maryland, the speed gains from bypassing natural language could help multi-agent systems tackle larger and more complex problems than possible with natural language. Whether this will succeed remains to be seen.
The development of Droidspeak highlights the ongoing advancements in AI communication methods, aiming to improve efficiency and problem-solving capabilities in AI systems. As AI technology continues to evolve, finding effective ways for machines to communicate will be crucial in unlocking their full potential.
The potential of Droidspeak could lead to new applications and improvements in AI systems, making them more efficient and capable of handling complex tasks. The future of AI communication looks promising, with researchers exploring new ways to enhance interaction between AI agents.