In theory current chess programs *could* play perfectly, and easily destroy a human. The limitation is the processing speed which limits how fast computers can evaluate moves or positions. In order to complete a game in a given amount of time only so many moves can be evaluated before one must be chosen. Therefore computer performance is dependant on processing speed (as well as memory, etc. as you have to store all those moves somehow...this is probably less of an issue if any, I'm not sure how intensive it gets) and will steadily improve as technology improves.
In the more abstract sense, computers will never be able to play in a human-like fashion. Humans make moves based on a variety of factors, only one of which is cold calculation. Computers rely on being able to calculate dozens of moves which is far beyond human capability. However humans also employ intuition, long-term positional sense, and so forth. For example, a human might see that a BvN is winning, but the win requires perhaps 100 moves to bring to fruition which probably eclipses the vision of his computer opponent (at present). Similarly, a computer may make a move which appears to go against any positional sense, because 30 moves later an exchange is won.
The point is that computers can easily eclipse human players (this is almost always the case today, and in the future the human side will certainly be lost 100% of the time), but they cannot emulate their styles. There is a subjective element to chess which cannot be replicated by purely objective machines. In theory, one could come close by studying the games of some grandmaster and trying to weight moves based on that grandmaster's flavors, but in practice this would be difficult as the effort required to emulate even one GM effectively would be enormous. Perhaps teaching a computer to emulate Tal might be easy (if can-attack == TRUE then ATTACK!!!!!!!!!! end-if 
), but how would one teach a computer to evaluate the positional nuances like Petrosian? A daunting task indeed.
I have been reading articles about advancement of computer chess programs and how they may or may not ever be able to play like humans do. I assume that means playing with logic rather than brute force ? To me that is nonsense. With future advancements in today's computers and development of quantum computers there won't be a grandmaster on earth that wouldn't look like as poor a chess player as I am. Machines will always....eventually...outperform humans in physical and mental endeavors as easily as a top fuel dragster would beat Usain Bolt in a sprint !
What I would like to see is a program that could be modified to play, as closely as possible and with human limitations, like the great grandmasters of today and yesterday, then play one against the other....impossible or not ?