In the 1970s, MIT's Artificial Intelligence Lab built its own computer network called Chaosnet. It was designed to connect the lab's Lisp machines—specialized computers dedicated to AI research. The network worked over short distances, roughly 1,000 meters (a bit over half a mile), linking machines within the lab and nearby buildings.

Chaosnet was built for a specific job: getting AI workstations to communicate with each other reliably and quickly. It was not designed as a general-purpose network that could connect computers across cities or countries. It was built for one place, one purpose, one research community.

The Problem: Keeping Information Accurate

The lab's engineers faced a particular challenge. When Lisp machines calculated things, they worked with symbols—short pieces of code or data. If those symbols got corrupted during transmission between machines, the computer would not catch the error. A calculation could silently produce wrong results, making it nearly impossible for researchers to track down where the mistake came from. To prevent this, Chaosnet's designers made a simple choice: always check that data arrived intact, and if it did not, send it again. Speed mattered, but correctness mattered more. (MIT DSpace)

Sharing Files Across the Network

One of Chaosnet's most important features was the ability for one computer to access another computer's files. Today, this is so common that most people do not think about it—cloud storage and remote access are ordinary. In the 1970s, it was not obvious how to do this safely and reliably. Chaosnet's FILE protocol made it possible for researchers to share large AI programs and datasets across the network without physically moving them on tape or disk. (Symbolics/CSAIL document)

Each Lisp machine was its own independent computer—not like older systems where many users sat at terminals connected to a single mainframe. Chaosnet let these independent machines work together as a team.

Why Chaosnet Disappeared

Chaosnet never left MIT. When Ethernet and, shortly after, TCP/IP (the protocol stack that powers the modern internet) became available, they spread across the industry. These were designed to work with any computer brand, not just Lisp machines. Symbolics, a company born from MIT's Lisp machine project, used Chaosnet at first but eventually switched to TCP/IP like everyone else.

Charosnet's real value was not that it "won" the competition with other networks. Its value was solving genuine problems before anyone else had figured out how. The questions Chaosnet's designers asked in the 1970s—how do machines talk reliably, how do they share files, how do they work as a connected system—are the same questions that led to modern internet protocols.

History often tells the story as one straight line from ARPANET (an early military-funded network) to TCP/IP dominating everything. But AI research created its own way of thinking about networks. The Lisp machine community cared about speed and accuracy over short distances more than about connecting distant computers. That difference—local reliability versus long-distance reach—is still a key tension in how networks are designed today.

For anyone curious about where modern networking came from, the original Chaosnet documentation is still available online, along with research papers in MIT's archives.