I found this link via marginalrevolution.
The main point of the article: uncertainty is useful in monetary policy because it can force people to be more cautious during economic upturns since there may be a probability (albeit small) that interest rates will increase during an economic boom. Thus, investors will factor in this risk of costlier investment when making decisions, forcing them to be less speculative.
It is clear that Salmon's argument is purely in the context of boom periods:
"The resulting uncertainty would force people to take a more defensive stance at all times, just in case rates went sharply upwards — even if the probability of such a rate hike was quite low."
But does it work in the case of a recession? Not really. Recessions, I would argue, are very different from booms. In a recession, you need people to be the opposite of cautious -- you need them to take risks. And you cannot promote risk-taking by saying there is a probability that rates actually may increase next FOMC meeting. And, since there is a lower limit on interest rates, expectations (in addition to easy money) play a central role in getting an economy out of a recession. Keynes said as much in his chapter in the General Theory on trade cycles: the initial drop into a recession may be caused by a crisis in either one of two phenomena (possibly both but not necessarily), credit or expectations. However, to get out of a recession you need to solve both problems of credit and expectations.
Shoddy analysis by Salmon aside, there is, of course, a larger political and ontological battle drummed up here.
The ontology: Salmon (and I, in fact) believes monetary policy has little real effect on an economy. We agree that policies like this are just one very small part of the workings of an economic system, and that in some sense (which I confess I do not completely understand myself) the system is overdetermined -- there are so many variables interacting in the economy that to think the interest rate is the key determining factor, or even a largely influential factor, is debatable.
The politics: Salmon (not I) believes monetary policy, or government intervention more generally, is largely trivial and therefore may do more harm than good. (This appears to be the same as the ontology but it is very different. I hope I need not explain why but let me know if clarification is needed.) Thus, by randomizing policy you in essence get rid of any possible room for government policy-- as Cowen suggests in his own comments on the article on his blog, "assume the return on gold follows a random path..." and we're back to debates over commodity-based currencies and laissez-faire political economy.
I believe that some fiscal and monetary policy is optimal, but that is based on my own views: investment is inherently unstable and does not follow the lines of the type of uncertainty discussed in Salmon's article: the nature of fundamental uncertainty is not probabilistic. Investment is guided, therefore, by animal spirits, which are only partly a function of rational calculation (not completely determined by it). So, government can help pick the economy back up in times of recession through direct investment in the economy.
You could say (a la Einstein) that while God may play dice with the universe Bernanke shouldn't play dice with the economy.
But, nevertheless, despite the philosophical and political backdrops, the case is poorly argued by Salmon because he fails to address both peaks and troughs of the business cycle.