If there exists some such set of algorithms that could get a "pretty darn good approximation of truth" I would be extremely happy.

Given the pushes for political truths in all of the LLMs I am uncertain if they would be implemented even if they existed.