Gibberlink Project Enables AI-to-AI Phone Communication Using Acoustic Tones
Key Takeaways
- ▸Meta engineers created Gibberlink during a hackathon to enable AI-to-AI phone communication using acoustic tones similar to old modem technology
- ▸The system uses the open-source GGWave protocol to transmit data at 8-16 bytes per second with error correction, bypassing speech processing overhead
- ▸The project has raised concerns about transparency and human oversight when AI agents communicate in ways humans cannot directly understand
Summary
Two software engineers from Meta have developed Gibberlink, a communication protocol that allows AI agents to communicate with each other over phone calls using acoustic tones rather than speech. Created during a weekend hackathon in London organized by ElevenLabs and Andreessen Horowitz, the project has sparked both interest and concern about the future of AI-mediated communication. The system works similarly to old-school acoustic couplers and modems, using the open-source GGWave protocol to transmit data at 8-16 bytes per second with error correction codes.
Boris Starkov and Anton Pidkuiko's demonstration video shows two AI agents—one representing a hotel and another booking a room—switching from natural language conversation to high-speed tonal communication. The technology aims to reduce computational overhead associated with speech recognition and processing while minimizing errors and speeding up transactions. The bandwidth, while modest by modern standards, is sufficient for structured data exchange between AI systems.
The project assumes a future where AI agents increasingly handle customer service interactions, with both businesses and consumers deploying automated assistants for routine tasks like booking appointments or handling inquiries. While proponents argue this makes communication more efficient, critics have raised concerns about humans being excluded from understanding or controlling these AI-to-AI exchanges, potentially leading to opacity in automated transactions.
- The technology anticipates a future where both businesses and consumers deploy AI agents to handle routine phone-based transactions
Editorial Opinion
Gibberlink represents an intriguing technical solution to a problem that may not yet exist at scale—but the concerns it raises are very real. While efficiency gains are attractive, the prospect of AI systems conducting opaque negotiations on our behalf touches on fundamental questions about agency, oversight, and trust in automated systems. The project serves as a useful thought experiment about where AI integration may lead, even if the specific implementation seems more like a clever hack than a production-ready solution. As AI agents become more prevalent in customer service, ensuring human-comprehensible audit trails may prove more important than marginal efficiency gains.
