Computers may improve all the time in processing power, but this is a crude measurement. It is quite conceivable that some groups of tasks will remain beyond the ability of computers forever - for instance ones requiring creativity or originality - and that more will not be possible for the foreseeable future at least. More importantly, the question is not just whether or not a computer could be made to do it, but whether it would be practical and economic to do it. You might be able to build an aircraft that could fly without human guidance, but be exorbitantly too expensive to produce or operate. Further, such things could well go through a long period as laboratory curiosities, too hedged in by limitations for anything but extremely specialised applications. Finally, there exists a certain class of mathematical problems which computers cannot, for reasons of the scientific philosophy underpinning computing, resolve.
I once saw a documentary concerning a Mercedes Benz project to build an automatic road vehicle. They did, indeed, get it to work within extremely restrictive parameters, and one saw it driving around a test track as the operator left the steering wheel and walked into the back (it was a van) to show off the equipment that made it work...which filled the load space completely. Merc later abandoned the project as a waste of effort.