What is the Drosophila of AI now? In the mid 1960s, researchers have famously referred to chess as the "Drosophila of AI": like the fruit fly, the game of chess was accessible and relatively simple problem to experiment on, which yet produced important knowledge more complex problems.
Now, people seem to say "chess is just a search problem" and "chess methods will yield little of further interest to the AI community".
So, what is the Drosophila of AI now?
 A: I just googled and a lot of people quote John McCarthy calling Go "the new Drosophila of AI" - although I haven't found his original saying.
There's also an interesting paper "THE DROSOPHILA REVISITED" (pdf) which, in particular, reads:

After the match DEEP BLUE - Kasparov (New York, 1997) in which the machine proved its superiority, a slow  transition was observed in the games world from chess to other games, with Go as the current frontrunner. The  ICCA changed its name to ICGA, and the question arose: Is Go the new Drosophila of AI?  
Some would agree with this statement and others would vigorously oppose it. In more balanced terms one would say: for such a change of paradigm, a paradigm shift is a prerequisite, a shift of focus is not sufficient. At this moment (2010), we may state that the conditions are fulfilled, since MCTS can be considered as a paradigm shift. 

A: How about Robotics (specifically, humanoid robots)?
Specifically I think the challenge in robotics is to combine a set of technologies that in themselves are quite well developed:


*

*Computer vision: the robots need fast processing of the visual world

*Internal modelling of the world: they also need to know how they can affect the world, and how to connect the visual landscape with their movement

*Speech recognition: we want to be able to talk to them, right?

*Speech synthesis: and we want to hear what they have to say!

*Reinforcement Learning: they should be able to learn through trial and error, etc.

*Bayesian reasoning: at some point they will probably need to have probabilistic notions of objects in the world in order to facilitate decision making


It would be easy enough to give them chess- or go-playing capabilities as well ;-)
I think the only trouble with this, from the Drosophila point of view, is that there is a significant cost in terms of hardware. However there's no reason why the robot couldn't live in a simulated world. 
And perhaps there is something in the gaming world like this, where you can create your own AI bot that can interact with the physics engine using multiple modalities?
