Here's my interpretation of the abstract: Global warming has been previously predicted in a range of 2 - 4.5C with a median estimate of 3C. The new study comes up with a range of 1.7 - 2.6C and a median estimate of 2.3C.
That is, the new median temperature change estimate is 0.7C less than the previous estimate, and the uncertainty range has been reduced from 2.5C to 0.9C. This is the way that science progresses - new models (if others can validate by replicating) provide reduced uncertainty and (hopefully) better estimates. Note that nothing in the abstract says that global warming is not occurring or is not caused by CO2 rise. Indeed, I expect to find the article is based entirely on the mechanisms by which CO2 perturbs the climate.
The article, once again, supports the science showing that rising CO2 levels lead to global climate change and an increase in the average global temperature.
Here's what I expect to see widely quoted:
"...these results imply lower probability of imminent extreme climatic change than previously thought."What will be eliminated from the quote is the initial modifying phrase...
"Assuming paleoclimatic constraints apply to the future as predicted by our model,..."When scientists begin a sentence with "Assuming...", well, you should assume they are outlining a thought problem for future study rather than discussing something proven. This common approach to scientific writing has caused much angst and misunderstanding in the global warming debate.
I'm interested if the authors have a good argument as to why paleoclimate constraints are still reliable under conditions where an external CO2 source (fossil fuel burning) is steadily forcing the system rather than part of a climate feedback cycle. I expect that what I'll find is a paper that supports the direction and general magnitude of existing global warming models, but application to the future remains speculative until the difference climates with and without human fossil-fuel burning are further analyzed.
 You shouldn't assume that what follows "assuming" is either true or necessarily discussed in the paper. In many cases, the scientists are merely presenting a thought problem illustrating the possible importance of their paper. If they actually had a good argument, they wouldn't begin the sentence with "assuming." In this case, if they were able to show evidence that paleoclimate constraints developed without human intervention should apply to a climate affected by humans, they would have written something like "It is shown that arthropogeneic paleoclimate constraints apply to the anthropogenically-affected future, so predictions of our model imply a lower probablility..."
 As an example, the entire "hockey stick" controversy revolved around an analysis methodology that written up using an "assuming" form of discussion, which was then taken out of context by people with an agenda. The authors were quite careful in their discussion - the readers not so much. However, debunking the readers out-of-context hype is not the same as proving the science to be ill-founded.