There is a real possibility that all you would end up with is an on-line version of the current crop of dead-tree publishers. In addition to simply democratising publishing there needs to be some thought applied to the subject of the publisher as an institution. Learned societies and universities fit this approach so tools that made publishing easier for these group to manage would be a step forward. However any concentration of power or influence results in some of the current set of problems quickly resurfacing. Perhaps it is simply a facet of any enterprise where humans are involved and a revolution every so often is needed to shake things up a little.
Now the act of publishing is trivially cheap, so you can simply let everyone publish everything first, and apply aggregation and filtering second.
If you took all the attendees of those two conferences and gave them a collaborative website for rating and reviewing papers, you'd probably get something just as good as the old journals, and probably faster.
You seem to be criticizing a straw man. Nobody is suggesting that peer review should stop. We're suggesting that there's no longer any reason to link peer review with publication and access.
The real value is added by the reviewers and the authors, and they can keep on providing that value the same as always. It's the publisher who has become nearly superfluous, yet it's the publisher who charges boatloads of money for access.