Depending on exactly how/what the operator is required to do, it might be feasible for the onboard AI to hand a lot of the tasking. Current generation games have fairly respectable AI's these days. Path finding, terrain avoidance etc, are certainly well within the current scope of AI operations right now, and another 12 years of self driving cars blazing the way (and killing pedestrians) should see large scale advances. If the pilot is only responsible for basically saying "attack here", "navigate to here and be stealthy" etc they amount of data the drone pings back would be very minimal. Now video that's a slightly different story, but the question occurs, what's the point of video? For recon work the data has to go somewhere, but not necessarily to the pilot/operator in real time. Compress it and kick it out to any capable receiver if you're not trying to be stealthy. WVR dog fighting and potentially CAS/A2G are really where video might be of use. In BVR, 1990's vector graphics would suffice, as it's far more about finding the target and guesstimating if they've found you, versus looking for them. A2G/CAS would be the main one I could see having real time video being particularly important, but with good enough ISR or ground marking this could also be not much of an issue. A non visible, but IR hot flare/strobe for the UCAV to lock onto, and "300 meters at bearing 035 from strobe" would be plenty for the on board AI to make a very precise attack.
Honestly being a hot shot drone operator in 2030 will probably resemble being a code jockey/professional RTS player more then being a pilot. The ability to quickly reconfigure and integrate AI scripting and hierarchic targeting priorities would be huge, and the ability to multitask a bunch of units against a bunch of units in a 3D fight from a non FPS view would be critical.
