@ianRobinson Common sense would suggest that while the growth may be exponential for a bit, there would be an asymptote.
Also and unfairly, because I will not listen to the program, what were the metrics used to evaluate “better?” Were they suggesting 2n source data -> n^2 correctness? As I believe this is essentially applied probability theory, were they asserting the unsurprising result that larger sample sets increases confidence values?