The argument about Open Access publishing is one that is going to run for some time, I am sure. (See yesterday's piece in The Economist, for example. One spectre that raise its ugly head from time to time, though, is the notion that Open Access is bad because it encourages bad science. In essence, the argument goes: Open Access is cheap for the publishers, therefore they are willing to publish anything. In support of this, I have seen people cite the high acceptance rates at PLoS and BioMedCentral of (apparently) 50-70% as an indicator that they will publish pretty much anything. I've even seen this used in favour of removing pre-publication peer-review altogether in favour of post-publication peer-review, as in the case of WebmedCentral - the implication being that, in the Open Access world, pre-publication peer-review is useless.
I think this is all rather unfair on Open Acess journals, especially the likes of PLoS and BMC. The latter have just done away with the need for the science to be deemed "interesting" as long as the science is sound. (And this for only some of their titles.) This is a far cry from publishing anything and peer-review being worthless.
The peer-review process involves highlighting things that need to be changed to meet the "scientific soundness" criterion. Most authors then go ahead and make the required changes, usually improving the paper as a result. I would argue that this is the single biggest utility of peer-review; as an author, I want to go through this process myself, even if it is sometimes irritating. The fact that PLoS and BMC publish 50-70% of papers just indicates that scientists are capable of producing scientifically sound work (following revisions) 50-70% of the time. The higher rejection rate at Science and Nature is more to do with what science is perceived as being of general interest, rather than the soundness or quality of the work. (I also suspect the acceptance rate for the flagship PLoS and BMC journals that do stipulate a need for impact is considerably lower than across the whole series.)
The peer-review process and quality of reviewers is not always perfect but it is still better than none at all. I am very reluctant to ever publish in a journal that does not have pre-publication peer review and it is not just about the journal's "Impact Factor"; the quality of the papers is going to be lower without the opportunity for revisions in the face of peer-review and at least the threat of being in that 30-50% that aren't deemed scientifically sound enough to publish.
The biggest problem in the modern scientific literature is the pressure on scientists to publish too soon and too often, not the presence of journals that are willing to publish boring science. Until we get rewarded for the quality of our science and not the quantity or impact of our papers (first is not always best), I cannot see this changing, sadly.