Warren Pearce has written an interesting layman's guide to sceptical arguments on climate change at the Making Science Public blog. He covers issues such as Climategate and science by press release in what I think is a reasonable summary.
The section on computer models was also quite good:
Computer models are critical to climate science and the projected effects of carbon dioxide emissions on global temperatures. Criticisms are levelled at these models sometimes focus on the assumptions upon which they are based. More broadly, there are worries about the weight afforded to these models over empirical observation. In other words, can we not learn more from existing temperature data than projections?
However, this elicited a comment from Gavin Schmidt that misrepresents things completely:
...the a priori demonisation of ‘models’ as a tool for making forecasts makes no logical sense – since of course we don’t have ‘empirical data’ from the future. Instead, results from coherent and physics-based models are dismissed in favor of untested and unevaluated heuristics – ‘no change!’, ‘new ice age!’, etc.
Who said anything about a priori demonisation of models? Sceptics dismiss the output of climate models as tools for policymaking because GCMs have no proven ability to make valid forecasts - to the extent that their predictions have been tested, they are running much too hot. If climate scientists want to play with GCMs, they are welcome to try. They will learn much along the way. Just don't expect us to believe the output is a valid forecast (in Gavin's words) without some evidence.