People working in probabilistic programming can be seen as working towards AGI, in some ways.
I'd take a look at the work done by Tenenbaum at MIT and Goodman at Stanford. There are also lots of academics related to them in different places. You can reconstruct this graph by looking at Google Scholar citations and co-authorships.
As AGI isn't something that is widely agreed upon to be currently extant you'll want to search for Computer Science and more specifically Cognitive Science instead.
Read a bunch of papers. Find some authors that think like you. Reach out to them.
I'd take a look at the work done by Tenenbaum at MIT and Goodman at Stanford. There are also lots of academics related to them in different places. You can reconstruct this graph by looking at Google Scholar citations and co-authorships.
A great starting point is: http://probmods.org/