@Shadowman311
cc @dave
This os the exact kind of thing I meant, when I said, that parity with doctors diagnosis is not a goal you should be aiming for.
@Shadowman311 @dave
Or in other way:
Every single smart guy on the planet knew the humans cannot fly, until one idiot thought different.
I do want an AI to make its own deductions, not the same one as the experts.
@rysertio @dave @Shadowman311 
I have no idea what kind of data do doctors have available.
I am also not an AI scientist. All I know is, that AI data needs to be as clear as possible, and if the data is trained against expert diagnosis, then it cannot possibly surpass it.
@jmw150 @dave @Shadowman311
Yes.
But that is, because you can train AI easilly on scans, and the "alive" field 3 years later 
(or, a binary and quite measurable result)
@jmw150 @dave @Shadowman311
The problem I am taking about with expert data is, that all 3 doctors I mentioned in my previous post believe that they did a good job.
How are you going to get data for training it. AI is not magic. It just find patterns in data.