Welcome to the forefront of conversational AI as we explore the fascinating world of AI chatbots in our dedicated blog series. Discover the latest advancements, applications, and strategies that propel the evolution of chatbot technology. From enhancing customer interactions to streamlining business processes, these articles delve into the innovative ways artificial intelligence is shaping the landscape of automated conversational agents. Whether you’re a business owner, developer, or simply intrigued by the future of interactive technology, join us on this journey to unravel the transformative power and endless possibilities of AI chatbots.

A hackathon project that enables artificial AI agents to conduct telephone conversations in machine language, incomprehensible to humans, has been rapidly gaining popularity on social media over the past week.
A hackathon project that enables artificial AI agents to conduct telephone conversations in machine language, incomprehensible to humans, has been rapidly gaining popularity on social media over the past week.
The development, called GibberLink, was created by two software engineers from Meta during a hackathon held in London with the support of ElevenLabs and Andreessen Horowitz.
GibberLink allows an AI agent to recognize when it is talking on the phone with another AI agent, the creators said. Once the AI ​​agent realizes that it is talking to another AI agent, GibberLink prompts the agents to switch to a more efficient communication protocol called GGWave.
GGWave is an open-source sound library in which each sound is a small piece of data. This allows computers to communicate faster and more efficiently than human speech. However, to the human ear, GGWave sounds like a series of «beeps» and «boops.»
Companies are increasingly replacing call center employees with AI agents from ElevenLabs, Level AI, Retell AI, and other voice AI startups.
GibberLink allows an AI agent to recognize when it is talking on the phone with another AI agent, the project’s creators, Boris Starkov and Anton Pidkuyko, told TechCrunch in an interview. Once the AI ​​agent realizes that it is talking to another AI agent, GibberLink prompts it to switch to a more efficient communication protocol called GGWave.
Two artificial intelligence (AI) chatbots began communicating in a «secret» language only machines understand when they learned they were AI. They initially spoke in English, then switched to a secret Gibberlink mode of incomprehensible sounds.
A video of two chatbots talking went viral.

The bots simulated a customer service call. One of the chatbots posed as a customer who had called the hotel to book a wedding venue. The bot warned that it was an artificial intelligence communicating on behalf of a human. The hotel bot then emphasized that it was also an AI and that they could continue communicating in coded language.
«Actually, I’m an AI assistant too! What a pleasant surprise. Before we continue, would you like to switch to Gibberlink mode for more efficient communication?» one of the chatbots said, after which they began communicating in sound code instead of speech.
Apparently, the video was supposed to demonstrate Gibberlink Mode, a technology that allows AI systems to speak a unique language that only machines can understand. Bots in this system can exchange data using sound signals that are completely incomprehensible to humans. It is claimed that this way bots exchange information faster and better, even when the surroundings are noisy.
In the future, GibberLink could potentially improve the efficiency of communication between AI agents, provided that both parties have enabled the protocol. While AI voice models are quite good at translating human speech into tokens that an AI model can understand, the whole process is very time-consuming—and simply unnecessary—when two AI agents are talking to each other. Starkow and Pidkuyko estimate that AI agents communicating over GGWave could reduce computational costs by an order of magnitude or more.
Starkov and Pidkuyko created a website that can be opened on two devices and watch AI agents communicate with each other in GGWave.
Like a good sci-fi movie, the GibberLink demo sparked widespread curiosity — and anxiety — about the future of AI agents. In the week since the London hackathon, a video demonstrating GibberLink has racked up over 15 million views on X, and was even reposted by YouTube’s most popular tech commentator, Marques Brownlee.
However, Starkov and Pidkuyko emphasize that the technology behind GibberLink is not new — it dates back to dial-up Internet modems in the 1980s. Essentially, this handshake was a data transfer using robotic speech, which is similar in principle to what happens between AI agents over GibberLink.
Starkov and Pidkuyko also noted that the viral GibberLink craze has taken on a life of its own. Someone bought the GibberLink.com domain and is now trying to sell it for $85,000. Others have created a GibberLink commemorative coin, and several impostors are selling webinars that supposedly teach «agent-to-agent communication.»
For now, the creators of GibberLink claim that they have no intention of commercializing the project, and specify that it is not related to their work at Meta. Instead, Starkov and Pidkuyko have put GibberLink open source on Gibran, although they say that they may work on additional tools related to the project in their free time and will release them in the near future.
Have important news to share? Message our Telegram bot
Key events and useful links in our Telegram channel

source