Go tries to be safer and more automatic, like Java. While it's more organized and prescriptive than C or C++, it's less religious than Java about everything being a class. So it encourages the pragmatism and low-levelness of closer-to-the-metal languages while also abstracting out a lot of the details which get in the way of just writing systems.
Because they want both speed and OO programming? :)
C++ is a multi-paradigm language. It has more abstractions than C, while retaining the ability to get down to the metal when you need to. Arguably, these higher level features are desirable when working on larger code bases. C++ also has features that make it easier to write clean and correct code (such as RAII). Templates and operator overloading makes for writing expressive libraries. I remember reading and article that compared C and C++ in the context of a matrix library. The C++ code was much more concise and readable, and was easier to get right compared to C, mainly due to those two features.
Furthermore, C++ can actually be faster than C. The canonical example is the templated std::sort() vs C's quicksort(). The C++ compiler is able to inline the comparator argument to the former, whereas qsort() takes a function pointer.
http://commandcenter.blogspot.fi/2012/06/less-is-exponential...
Apparently the remaining C++ users are really fixated on something called zero cost abstractions, and it's not a goal of Go: "Zero cost isn't a goal, at least not zero CPU cost. Go's claim is that minimizing programmer effort is a more important consideration."
(though I don't think C++ succeeds in "zero cost abstractions")