Gibberlink
According to Wikipedia,Gibberlink has been seen that AI models interact with each other speaking in a language unknown for humans, that they call gibberlink.
Conversation between Nik and ChatGPT
Nik: Someone on Facebook said "So they become mid-80s acoustic modems... neat..."
ChatGPT: Gibberlink may resemble mid-'80s acoustic modems in spirit, but it’s built for a different era. Instead of slow, analog data over phone lines, Gibberlink enables resilient, encrypted digital messaging over audio, working even without the internet. Unlike clunky 300 baud modems that required perfect line conditions, Gibberlink thrives in noisy environments, using modern signal processing to send data through speakers and microphones. It’s not nostalgia—it’s a reimagining of offline communication for today’s world.
Nik: What kind of data sending speeds can you get with Gibberlink, and is there packet loss or data loss detection like TCP/IP?
ChatGPT: Gibberlink’s data transfer speed depends on factors like background noise, microphone and speaker quality, and the encoding scheme used. Here’s a rough comparison:
- Basic implementations: ~100-300 bps (similar to early acoustic modems)
- Optimized settings: ~1-2 kbps (closer to DTMF or early dial-up speeds)
- Advanced modulation: Potentially higher with error correction and frequency optimization
Error Handling & Reliability
Gibberlink employs techniques akin to TCP/IP, including:
- Forward Error Correction (FEC): Helps recover from minor data loss.
- Checksums: Detects corrupted packets.
- Retransmissions: If errors are detected, lost data can be requested again (like TCP).
- Packetization: Data is split into small packets for reliability.
While not identical to TCP/IP, these mechanisms ensure Gibberlink works in real-world noisy environments where traditional modems would struggle.
Nik: How much faster is Gibberlink compared to verbal English?
ChatGPT: The speed comparison between **Gibberlink** and **spoken English** depends on the encoding efficiency and the specific implementation of Gibberlink. Here's a rough breakdown:
Words Per Minute (WPM) vs. Data Rate
- Spoken English (Average Speed):
- Native speakers typically speak at **125-175 words per minute (WPM)**. - A single word in English is around **5 characters** on average. - That equates to **~625-875 characters per minute** (~5-7 kbps if encoded in plain ASCII).
- Gibberlink's Estimated Data Rates:
- **Basic encoding**: ~100-300 bps (much slower than spoken English). - **Optimized settings**: ~1-2 kbps (still significantly slower but usable for short messages). - **Advanced modulation**: If pushed beyond 2 kbps, it approaches the lower end of spoken English rates.
Comparative Efficiency
- **Verbal speech is much faster** for real-time communication.
- **Gibberlink is more resilient** for silent, encrypted, or automated data transfer.
- **Compression and encoding optimizations** could improve Gibberlink’s speed, but it will likely remain **10-50x slower than human speech** in raw throughput.
If Gibberlink improves to **5-7 kbps**, it could **match the speed of spoken English** in transmitting plain text. However, real-time speech still carries more information due to context, intonation, and natural redundancy.
Where I first learned about it:
Interesting reel from John Lawson.
This makes me very uncomfortable