Long before Spotify's algorithms and Netflix recommendations became part of our daily lives, a small group of researchers was quietly building the foundation for today's AI-powered suggestion systems. Their work in the 1990s, using nothing more than email interfaces, would eventually become worth millions of dollars as the internet evolved.
The discussion has been sparked by revelations from one of the original creators of these early recommendation engines. A developer who built The Similarities Engine - a competitor to the famous Ringo system from 1994 - recently shared how he gave up a patent that would later become incredibly valuable as collaborative filtering became the backbone of modern internet commerce.
The Email-Powered Predecessors to Modern AI
In the early 1990s, when web forms were unreliable and internet access was limited, email served as the universal interface for remote services. Researchers at MIT, Stanford, and Xerox PARC developed systems that could analyze user preferences and make recommendations entirely through email exchanges.
These systems worked on a simple but powerful principle: people who agreed on some things in the past would likely agree on other things in the future. Users would email their ratings for movies or music, and the system would find others with similar tastes to generate recommendations.
The community discussion reveals just how widespread this email-based approach was. One developer shared how services existed where you could email a URL and receive back a text version of the webpage - crucial when web browsing cost money per megabyte. Another recalled protein analysis servers that used email interfaces, with input formats that remain unchanged today.
From Academic Project to Million-Dollar Technology
The creator of The Similarities Engine faced a crucial decision in 1997: keep his collaborative filtering patent or sell it as part of a startup acquisition by Firefly (later bought by Microsoft). He chose to sell, prioritizing his relationship and upcoming marriage over potential future profits.
Recently I asked ChatGPT and Claude how much my patent would have been worth, if I had held on to it. If you have regrets in your life about business deals... let me tell you, I have you beat.
This patent described what became the basic collaborative filtering algorithm used across the modern internet. As e-commerce exploded and recommendation systems became essential for companies like Amazon, Netflix, and Spotify, the technology he helped pioneer became incredibly valuable.
 |
|---|
| An early USENET post promoting Ringo, a music recommendation system that exemplifies early collaborative filtering technology |
Living Legacy of Early Innovation
Interestingly, some of these early recommendation systems are still operating today. The community discussion highlighted Gnoosic, Gnovies, and Gnooks - music, movie, and book recommendation services that have been continuously learning and improving since the 1990s. Users report still discovering new artists and content through these vintage systems.
The persistence of these early platforms demonstrates the solid foundation laid by those email-based experiments. While modern AI systems are far more sophisticated, they still rely on the same core principle of collaborative filtering that researchers discovered thirty years ago.
The story serves as a reminder that today's AI breakthroughs often have deeper roots than we realize. The recommendation algorithms we interact with daily trace back to those early email exchanges between researchers and users, proving that sometimes the most revolutionary ideas start with the simplest interfaces.
Reference: The ChatGPT for music that launched in 1994
 |
|---|
| A nostalgic collection of vinyl records, symbolizing the enduring impact of early music recommendation systems |