The HUD is very accurate,
but is the human? … and what about duplication, monitoring?
Using a single self-monitored HUD raises the question of how to monitor the human derived output. Use the other pilot and head down instruments, or with two HUDs, what if the humans disagree?
The automation issue is not what can be automated or at what cost; it is what should be automated and what reasons drive this decision. For the latter it is often the limits of human performance, the ability to conduct a task, or the reliability of the task in a range of conditions or over time.