Alongside the U.S.-centric Pandora, Last.fm was a pioneer in the web music-streaming space. Its core function, ‘scrobbling’, even threatened to become a widely-used word at one point. These days the company is trying to find its feet again, but co-founders Felix Miller and Martin Stiksel are onto a new project now, called Lumi.
Where Last.fm was all about recommending new music based on the user’s tastes (which were recorded through the aforementioned scrobbling), Lumi is trying to do the same thing for general web content. Only this time it’s even more of a sit-back experience: all the user needs to do is install a Chrome, Firefox or Safari browser plugin and let it watch pretty much everything he or she does online, barring the secure and private stuff.
“Ever since we left Last.fm, we were thinking about applying this concept and mix of technology on a much larger basis, to help people discover all sorts of content,” Stiksel told me. “How we are trying to do this is by allowing people to anonymously record their browsing data with Lumi to let Lumi know what things you are interested in. It’s the Last.fm concept applied to all sorts of web content.”
The service is very much in prototype phase right now, which is why Lumi is looking for early-birds to try using it over the next few weeks, to see what happens. As with all machine-learning systems, nothing valuable comes out until a good deal of data goes in. Lumi needs to see what people look at online.
How does it work? There, the two get tight-lipped, although Stiksel hinted to me that they were “using a bunch of technologies and seeing what works”. What they do stress is that it’s all about effortlessness – it doesn’t require you to rate content, like StumbleUpon does.
Privacy is naturally a concern so, as with other omniscient services such as Archify, Lumi is keen to tout its security credentials. “We assure you that your browsing activity is secure, completely anonymized, and only ever accessible by you,” the blog post reads.
It’s hard to tell how this will turn out. Indeed, I get the impression from Miller and Stiksel that even they aren’t fully sure where it’ll end up – that seems to be one point of this new testing phase. What I can say is that web content recommendation is threatening to become a hot market again, what with the advent of Lumi and Futureful, which we covered last month.
Give it a year or two. It seems a burning question down the line will be this: who’s better at recommending content, machines or your friends?