Wouldn't there be an obvious trade off there, requiring an extra heap allocation and indirection every time a slice is used in a data structure? I've always thought of Go as lower-level than Java (since it has explicit pointers, more control over memory layout, etc), which I think makes it a more useful language overall even if not the best fit for every application (since if you do want Java, it already exists and is mature).
In other words, I think both Java's and Go's decisions make sense here.