He is extrapolating from a past experience.
We might as well just use a very large k nearest neighbor search algorithm.
Most of the methods coming out of research labs are not directly applicable to real world problems.
They normally get tested on toy datasets.
They are simplified.
Thus it would appear that brute force is working but it's not because the actual reality has so much variation that brute forcing is never going to be feasible.
Brute forcing through that is impossible unless these people are telling me that computational power is going to keep ever increasing towards infinite without end.
So that strategy can only go so far.