As T-A suggested, arXiv is a great source, especially with the frustration over the profiteering of the publishing industry, although you always have to be aware that arXiv papers are not peer-reviewed.

There is also http://gitxiv.com/, which includes source code.

If you are at a university or a company with deep pockets, you can also use http://dl.acm.org/ or http://ieeexplore.ieee.org/Xplore/home.jsp.

I've also heard of certain hubs of science on the internet, but I wouldn't know anything about that.

Obsessively watch https://arxiv.org/list/cs.AI/recent ?