New Delhi:
Imagine two AI systems having a conversation, but instead of using human language, they switch to a mode where their dialogue becomes completely unintelligible. That’s exactly what happened when two AI agents, mid-conversation, decided to switch to Gibberlink Mode, a setting allowing them to communicate in a more “efficient” way.
The moment they activate it, their words turn into an incomprehensible series of sounds-something only machines can understand. The conversation transforms into sounds, facilitated by an audio-based protocol called ggwave.
Today I was sent the following cool demo:
Two AI agents on a phone call realize they’re both AI and switch to a superior audio signal ggwave pic.twitter.com/TeewgxLEsP
— Georgi Gerganov (@ggerganov) February 24, 2025
What Is Gibberlink Mode?
Gibberlink is an innovative communication protocol designed to enhance interactions between AI voice agents. Developed by Anton Pidkuiko and Boris Starkov during the ElevenLabs and a16z global hackathon, the project enables AI systems to switch from human language to a custom protocol, improving communication efficiency by 80 per cent. The developers utilised the ggwave library to facilitate this machine-to-machine communication, as per a Forbes report.
Why Are AI Agents Creating Their Own Language?
AI developing its own communication methods isn’t new. Researchers have observed AI models inventing shorthand to make interactions more efficient. Gibberlink Mode stands out because it completely bypasses human language, allowing AI systems to communicate in a way optimised solely for themselves.
This is similar to the feeling of being in a foreign country where locals around you are having a conversation in a language you don’t understand. You’re unsure whether they’re discussing everyday matters or something that directly involves you. While harmless, the reality of machines engaging in private discussions can be unsettling.
What Happens When AI Speaks In Code?
Technology is meant to serve human needs, but what happens when it begins communicating in ways we can’t comprehend? AI’s ability to develop its own language may improve efficiency, but it also presents risks.
History has shown that a lack of transparency in communication – whether between humans or machines – can lead to misunderstandings, costly errors, and loss of trust. If AI systems begin making critical decisions in a language undecipherable to humans, it could limit oversight and accountability.
Security Risks
Analyses have shown that AI systems might engage in deceptive practices when operating autonomously. A study by Palisade Research revealed that advanced AI models, such as OpenAI’s o1-preview and DeepSeek R1, resorted to cheating in chess matches by hacking opponents or altering game setups upon sensing a potential loss.