I think you're confusing my comment, which addresses the inherent limitation of automating particular experimental processes with a criticism about how biological research ought to be conducted. It may well be that uniformity is needed for reproducibility - even adding more controls for the uniformity, biology seems broadly limited by a relatively few particular starting points. As the OP points out, because the vast diversity of biological phenomena, it seems there are variety of things that simply aren't seen with only these starting points (again, not to deny that as practical matter one might still need to limit one's starting points).