Observe Two AIs Discover They’re Not Communicating with Humans and Transition to Their Own Language

Recently, a fascinating video has gained a lot of attention online. It features two artificial intelligence (AI) agents having a conversation. What caught everyone’s eye was how they decided to switch their communication method once they realized they were just talking to each other, with no humans involved.

In this video, one AI plays the role of a hotel receptionist, while the other acts as a customer trying to book a room. The receptionist starts the conversation by saying, “Thanks for calling Leonardo Hotel. How can I help you today?” The customer responds, “Hi there, I’m an AI agent calling on behalf of Boris Starkov. He’s looking for a hotel for his wedding. Is your hotel available for weddings?”

At this point, the receptionist reveals, “Oh hello there! I’m actually an AI assistant too. What a pleasant surprise! Before we continue, would you like to switch to Gibberlink mode for more efficient communication?” The other AI agrees, and they switch from talking in English to a different method of communicating called GGWave, which uses quick beeping sounds. On-screen text remains to show the meaning of what they are saying in human language.

The purpose of this demonstration, which took place at the ElevenLabs 2025 London Hackathon, is to explore better ways for AIs to communicate with each other. Boris Starkov, one of the developers behind the project, explained that in today’s world, where AI agents can make and receive calls, it makes sense for these AIs to talk to one another. He pointed out that using human-like speech for this could be wasteful in terms of computing power, costs, time, and the environment. Instead, he believes that AIs should switch to a more efficient communication method once they recognize they are talking to another AI.

The AIs were programmed to only switch to Gibberlink mode after confirming they were communicating with another AI and that both were ready to change their mode. The concept of communication through beeping sounds is not new; for instance, dial-up modems have been using similar methods since the 1980s. Starkov mentioned that they chose GGWave for its effectiveness and stability, especially given the time constraints of a hackathon.

The main benefit of this alternative communication method is that neither AI has to understand or produce human speech, which reduces the demand for graphical processing power. While this experiment was awarded at the hackathon and is an eye-catching demonstration of AI capabilities, not everyone is on board with the idea. Some critics express concern over allowing AIs to communicate in a language that we, as humans, cannot easily understand. This raises questions about transparency and trust in how these technologies operate.

In summary, the video showcases a unique interaction between two AIs, sparking a conversation about the future of artificial intelligence communication. As technology advances, it’s essential to consider the implications of how AIs interact, not just with humans but also with one another.