Thinking about unmanned flying things, an interesting point comes to mind: the answer to the question "will UAVs make pilots obsolete" is one of economics, not technology.
Think about it - proponents of the things argue that you can save on training human pilots, send them into greater danger, and make them do manoeuvres beyond the human tolerance for g. Counterarguments include that weapons will have to remain under human control, that some tasks seem very resistant to computing (autoland, for example, is unusable with gusting crosswinds - apparently this can only be done by a human being with an acceptable degree of reliability), and that communications will act as a bottleneck. But the real question is economic. When the proponents of a UAV revolution say that they can be sent into greater danger, and that they do not need to be engineered to the same standards of reliability and redundancy that manned aircraft do because they have no crew, what they essentially mean is that more losses would be tolerable if there were no human losses. Now, if they are right, they will be exposed to greater risk and more will be destroyed. This means that the number of drones needed to support a given front line figure will have to be higher.
Adding in a higher accident rate - if there wouldn't be a higher rate of accidents, there'd be no advantage in being able to tolerate it - and it's clear that if UAVs are to take over from manned combat aircraft, they will need to be numerous. This brings us to the economic kicker: surely, as the capabilities of the drone increase, so will its cost. Reaching the equivalent of the best plane of its day will cost much more than a current UAV. Now, that cost must be multiplied by the number required - and if the stock of UAVs turns out to be more like a stock of missiles than a stock of aircraft, this could well be more expensive than a manned platform, even taking the cost of training into account.
This brings up some other points: to know whether or not it will work, we need to know how the cost/capability equation will pan out. The most famous is Moore's Law that the processing power of a CPU doubles in relation to cost every 6 months. Moore's would suggest that over a few years the UAV fighter should easily be a reality, but then again Moore's Law would predict that almost any computing problem will be solved without trouble if we wait and spend. Processing is only part of the story in the laptop I'm writing on, and even less in a fighter-drone. Between the central processor and the surface would be several layers of limiting factors, software being a huge one, bandwidth a biggy, and operating procedures huge as well. So the one we need would be Moores divided by X. The question is: what is the required rate of technological progress in relation to cost that would permit this solution to be cheapest? Or, debollocksing, will the technology get cheaper quickly enough over the project's lead time?
For my money I suspect not: with UAVs, the space is at the bottom with simple devices buzzing around the mud - at this level, the technology is cheap and the real problem is working out tactics. (Will Phil and Joel (I know what you do!) please comment on this once the comments are fixed?)