Science as a whole can only be a group activity, of course - without replication of findings, it isn't really science (just
mad science).
But the peer review process as it stands necessarily entails a certain barrier to entry. While we all would like to believe that every deserving idea gets peer review, in practice, most people don't have the time to bother with a submitted paper unless it's coming from an institution they've more or less heard of. How could it be otherwise? There are huge numbers of crackpots out there, and institutional affiliation is an efficient way of filtering them out.
My wife wants to do theoretical physics research (she's qualified: she has a doctorate in theoretical physics) but unless she's affiliated with an institution, it's unlikely she'll get anything published, or so she says. The problem is that unless you're a rock star, affiliation with an institution entails teaching, which in turn translates into all-out effort for roughly 22 hours a day by my own observations of the past semester. The resulting Catch-22 is that my wife can't reasonably do research at all, and that's illogical; she needs no resources but time and access to books and preprints, which she has at home.
So I'd like to see some research done using these new-fangled informational access tools we have now - research that is both sound and eventually makes its way into traditional peer review simply on the basis of its soundness, and not because the institutional affiliation of the researcher makes it plausible. This is one more thing that I think the Internet can do for us, and I'd like to see it happen.
Of course, the romantic in me just likes the idea of the lone mad scientist in his lair discovering room-temperature superconductivity, because that is the twenty-first century I signed up for. Today room-temperature superconductivity, tomorrow super-intelligent lab rats and cheap immortality. It's all good.