To the blind spots/edge cases objection; the article about SPRs linked in the post shows that even when human experts are given the results of an algorithm and asked to fix their errors, they still do worse than just the algorithm on it’s own.
Because even though the human may recognize an obvious mistake, they will also recognize ten good predictions as “mistakes”.
To the blind spots/edge cases objection; the article about SPRs linked in the post shows that even when human experts are given the results of an algorithm and asked to fix their errors, they still do worse than just the algorithm on it’s own.
Because even though the human may recognize an obvious mistake, they will also recognize ten good predictions as “mistakes”.