Generating tests that match the implementation doesn't tell you the implementation is doing the right thing. If it isn't, changing the implementation will break the tests, which in the best case wastes time and in the worst means the bugfix is abandoned.

I deeply hate "regression tests" that turn red when the implementation changes, so you regenerate the tests to match the new implementation and maybe glance at the diff, but the diff is thousands of lines long so really it's not telling you anything other than "something changed".