Main problem is that most new commercial AI systems usually aren't really designed to let the human take over. There's a closed loop with a front-up designed ex

Main problem is that most new commercial AI systems usually aren't really designed to let the human take over. There's a closed loop with a front-up designed experience, and the only option in case of failure is to stop using the system. There's no recourse to manual adjustment of system behavior by the user. Research articles should insist on including the user in the loop from the beginning.