Bruce Sterling, SF writer, has pointed out that the Gobi Desert is far more hospitable than Mars, so before we ever settle Mars we’ll have settled the Gobi Desert (i.e. not any time soon). He also points out that by the time we have the ability to terraform Mars, we’ll be doing much more interesting things.
Sterling is missing the point here. People don’t want to settle Mars just because it happens to exist, they want to settle it because it happens to be very interesting. The Gobi Desert is not that interesting and it does not require settlement to find out whatever little that is interesting about it. However, Mars is a giant world-sized sandbox of possibilities that could tell us about the origin of the solar system, the formation of planets and perhaps life in the universe. There is no more closer or easily-reached place than Mars for finding those things out. Robots and other kinds of unmanned exploration can find out a lot from Mars, but scientists are agreed that humans beat the pants off robots when it comes to science and discovery, and that they will probably continue to do so for the next few decades.
Given that we may want to explore Mars and that using humans may be a very good way to do it, a permanent human settlement makes sense where the settlement of the Gobi Desert does not. The terraformation and material exploitation of Mars, however, is a completely different issue.
mssv 
The question of robot versus man has been going on for a long time. The rise of expert system, later reverted to knowledge-based system to avoid obvious embarassment in the 80s, offered a fair amount of excitement but they faded away soon due to problems in knowledge extraction and representation. More recently, the flavour of the month has been Knowledge Management. It is interesting to note that the age-old debate can be looked at again from ‘Knowledge’ point of view. Machine or robot is very good in dealing with ‘explicit’ knowledge as represented by explicit facts, axiomatic propositions, mathematical expressions, etc. However, we also have ‘tacit knowledge’ which is individual-based, context-related, analogous and practice-related knowledge. The best way to handle, deploy, transmit and interpret tacit knowledge is by a human being, not a robot. Hence, there remains very limited penetration of expert systems in the real world. From this angle, manned exploration of Mars will be far more valuable than sending a robot.
Terraforming the Gobi’s gotta be easier than terraforming Mars… you gotta admit it would be good practice…
Terraforming the Gobi might save Beijing, and prevent a war for usable land. It might prevent fine particles from destroying lungs or climate. People would also be more willing to pay for a major expidition to Mars if Mongolia became grassland again. What is not interesting about that?