IDSIA / University of Lugano in Switzerland is where e.g. Schmidhuber is. His research is quite neural network-focused, but also AGI-focused. Also Shane Legg (now at DeepMind, one of the hottest AGI-ish companies around) graduated from Lugano with a PhD thesis on machine superintelligence.

"AGI but not machine learning or neuroscience or anything with neural nets" sounds a little odd to me, since the things you listed under the "not" seem like the components you'll need to understand if you want to ever build an AGI. (Though maybe you meant that you don't want to do research focusing only on neuroscience or ML without an AGI component?)

Zoubin Gharhamani / Carl Rassmussen (Cambridge)

Michael Osborne / Yee Whye Teh (Oxford)

Microsoft Research Cambridge

Just wondering why you don't want to do machine learning? Many ML labs have at least some people who care about AI, and you'll get to learn a lot of useful technical material.