Yeah, that's what Gary Kasparov said. What he forgot is that computers double their capabilities every 18 months.
There's no reason why the robot-piloted plane itself needs to be any more expensive, less mechanically sound, or worse in communication than any human-piloted aircraft; the worst-case scenario is that you simply replace the pilots of existing aircraft. But that's already a winner: computers are capable of executing much higher-g turns than humans. Also, they're much better at taking calculated risks, because they know to several decimal places how much leeway they have...and they are only as risk-averse as they're told to be.
What I expect, though, is that the "fighter planes" of the future will essentially be intelligent missiles. Extremely intelligent missiles.
I don't see that happening anytime soon. My posit for this opinion is that it takes an equal competitor to move technology forward. Thus, aerospace science and aeronautical engineering (along with automation) only moved forward during WWI, II, Korea, the Cold War and Vietnam. Since we have no competitors in this realm at this time (China is still 10 years away) I suspect we will rely on stagnant technology until something earth shattering hits us. (My bet is China landing on the moon).
Capacity, is only number crunching. The rules for Chess do not change. ACM tactics change constantly, and it is the pilot that recongnizes those changes, in able to best use his aircraft against any one of dozens of potential fighter threats, at different speeds, altitudes, weather conditions, day-night, radar and hundreds of other factors, that will win the day. From the ground, you stand as much chance against another pilot, as some child with a video game.
You will never create a computer that can tell me whether my tally is losing energy in a turn, or how hard he is pulling based solely on the vortex coming from his LEX. Oh wait, does he have a LEX, or canards, what might his fuel state be; depending on what base he probably came from, how much burner time has he got available? Is he a 2ship or a 4ship from my radar return alone? Can I beam him if he gets a snap off against me? I could go on for days.
Now, someday, someone could likely program all those sensors into a fighter aircraft and return that data to the ground in about 3-4 seconds allowing someone to take positive action. The only problem is that if I'm in the plane, and I cant figure out and act on that information in less than 2 seconds, I will probably get my ass shot off. I wont be holding my breath waiting for some badass computer fighter jock to take the place of a man in the seat.
"Open the pod bay doors, HAL!"