just finished feynman's QED.
1. virtual particles are particles that, as Jonathan said pop in and out of existence without you seing them. let's say you have one photon which is supposed to go from A to B. most of the times, meaning the greatest probability amplitude is that it goes straight from A to B. But that's not all, experiments counting photons going from A to B say. It can also go from A to B but passing through any C point in space, even though C is not on the direct path from A to B. These paths have a smaller probability, but they happen. Also on the road the photon may spontaneously break into an electron and a positron which later recombine to recreate the photon reaching B. It can also break into a moun-antimuon pair or other particle-antiparticle pair. These things can also happen, with an even smaller probability. If the distance from A to B is big it doesn't actually matter that the photon can go in other ways than straight from A to B. Only if you want to get all the decimals right. But if the distance from A to B is quite small, like a few wavelengths it does matter. When stuff becomes "quantum stuff" that's when it starts to matter.
2. negative mass is the same thing as negative energy, right? and both are something you may end up with by trying to determine the mass of a real particle from its QED behaviour (coupling constant). You usually get infinities. That's why they invented that thing called renormalisation.