# [Python], 13 bytes

<!-- language-all: lang-python -->

 [(0,0)].count

[Try it online!] Takes input as a tuple.

[Python]: https://docs.python.org/2/
[Try it online!]: https://tio.run/nexus/python2#S7ON@R@tYaBjoBmrl5xfmlfyv6AoM69EIVojUSdJUyEtv0ghUSEzT6EoMS89VcMUIpKEIpKZppCmAVauqWBrq5ConRT7HwA "Python 2 – TIO Nexus"

Using an [object method][1] for the function avoids the boilerplate of a `lambda`.

 lambda a,b:a-~b # 15 bytes

Here, the idea is to map `(0,0)` to `1` and everything else to `0`. Since only `0+0` gives a sum of `0` among natural numbers, that always avoids matching the sum.

If one could output a Boolean here, which I find shady, a byte could be saved as 

 (0,0).__ge__

This checks if the input tuple is at most `(0,0)`, which is only true for `(0,0)`. In Python, `True==1` and `False==0`. Even more shadily, outputting via exit code and treating that as a Python Boolen would save two bytes:

 [(0,0)].pop

If string I/O is allowed and leading zeroes are OK, there's the 8-byte solution

 '1'.join

This concatenates `a1b`, which is always bigger than `a+b`.

 [1]: http://codegolf.stackexchange.com/a/95156/20260