I have attended Peyton Young's presidential address, "STRATEGIC LEARNING: Recent Advances and Open Problems", also Sergiu Hart's plenary talk addressing the question of when does there exist natural dynamics leading to Nash Equilibrium (Hart is president-elect). Also, yesterday I attended the "Nobel Panel", which is as you might guess, a panel discussion by Nobel laureates (Aumann, Maskin and Myerson). Notwithstanding Aumann's delightful rendition of "que sera, sera", there are some take-home messages.
Computer Science, and the notions that we in the CS community work with, are of great interest to mainstream game theory; a paper by Papadimitriou used the phrase "as the ice between game theory and computer science continues to melt..." -- well I can report that it is well and truly melting. This is a great time for computer scientists to continue to work at making contributions to game theory. Sergiu Hart himself is very conversant with the relevant terminology and concepts from CS, perhaps in part due to his work with Yishay Mansour (to appear in GEB).
It was claimed that the time is ripe for more emphasis on cooperative solution concepts, as opposed to non-cooperative ones. (I would note that work in the AI community is addressing this topic.)
Generally, the word "decentralized" is a big buzz-word around here. I did not foresee when I wrote my most recent research proposal "efficient decentralised algorithms for computational game theory" quite how mainstream are the general topics on which I was proposing to work. Indeed, if anything the situation is "too good", in that my research program is being pursued by some of the top people in the game theory field, and I have to take care in a future version of the proposal, to identify a distinctive agenda. (added 21.7.08.): By the way, this web page describes a long-running conference series on "decentralization", based in the economics community, that I just found out about.