Let's simplify things:
AI is nothing more than feeding a decision model a lot of data so that AI can 'intelligently' copy decisions.
Doctors only have unproven theories about the causes of cancer and the traditional therapies do not cure but merely kill.
When the best AI system of the far future is fed with the data of today, it will fail also because the data does not contain a solution.
Totally agree. AI is decision tree models and correlation analysis. But the massive computing power can crunch huge amounts of data looking for relationships that may hold a clue to causality and possibly even a cure. But raw power alone is not enough. And decision tree models cannot evolve on its own without creative human insight.
And one of the great limitations of AI is it doesn't work well with non numerical data. Health sciences data are more qualitative in nature so doesn't lend itself to AI like other engineering sciences.