Saturday's Telegraph contained an interesting interview with Nova Spivack, the founder of Radar networks. The story focuses on radar's development of Twine, a new product described as feeling like "a cross between Facebook, del.icio.us and Wikipedia." According to Spivack, Twine is powered by a "semantic understanding" of the web. Meanwhile, the Telegraph asserts that "Twine is one of several companies trying to take the lead in the fast-growing movement known as the semantic web -- or 'Web 3.0'."

In a video presentation on his blog, Spivack explains that today's social media has driven greater connections between people -- "people to people," Spivack calls it. The next evolution, he believes will include "connecting anything to anything."

The implications of such a shift are broad and potentially significant. And by Spivack's own admission, will require yet another major change in the way that web sites are coded.

At the end of the day, Spivack's vision rests on the principle that we can teach computers to truly think. That is, give them the artificial intelligence to anticipate our needs and serve up content that we want before we realize that we want it.

On this, I confess that I am quite skeptical. Even in 2008, I still believe that the very 1980s axiom "garbage in, garbage out" still applies. Or, put another way, that a computer's output is only going to be as good as the human being who instructed it one what to do. (I freely admit, here, that my thinking may be hopelessly backward and outdated, and I am happy to be proven wrong.) According to Richard MacManus, who had a chance to review Twine and discuss it with Spivack, the product is designed to bring people together "for a purpose." But Twine is still reliant on its (human) participants for information.

Perhaps Spivack and his team -- and the others working along similar lines -- have, indeed, built a better mousetrap. I'm now on the waiting list for the Twine beta, and I'm anxious to give it a try. But the real test for me will be seeing whether or not Twine can actually learn and serve up better information based on my unique needs and interests consistently over time.

The test for Twine is simple: I know that I can find what I want in Google. Can Twine give it to me before I ask for it?