I think most libertarians would tend to agree with Hayek's presentation of the Economic Calculation Problem as a fairly fundamental obstacle to successful government planning. There are a couple of problems with government attempts to improve society: one is their practical ability to do so (given a clear goal, are they able to achieve it) and the other is how they decide what constitutes 'improvement'. The fact that they generally fail at the former tends to mask the fact that they don't really have a good way of doing the latter. Given all the relevant inputs, perfect rationality and unlimited computational capacity I concede the theoretical possibility of a central planner producing more optimal outcomes than a market. Such a planner would be so far from any government that actually exists or could exist given current technology however that I don't consider it particularly relevant whether it is theoretically possible or not. That could perhaps change if Eliezer is successful.

The more immediate problem is that governments are not structured in a way that provides incentives to improve society. The reality of politics is all about special interests, rent seeking, regulatory capture and political maneuvering. The system as it actually exists is certainly not capable of making rational policy choices to improve society, though it remains possible that by some happy accident some policies may not be terribly harmful.