4 ms·

I'm under impression that current AI cannot handle 10% of edge cases that human can. Worse AI feels like to hallucinate so it's even harder to identify whether

I'm under impression that current AI cannot handle 10% of edge cases that human can. Worse AI feels like to hallucinate so it's even harder to identify whether the edge cases happen. How do companies plan to work around or mitigate that? Or they're betting on them being acceptable soon?

rar00 2y ago
with human supervisors is the plan. Instead of X workers, they now have X agents + Y<<X supervisors to unblock (and collect additional data for) agents