16% effectiveness for complex tasks vs 88% is a big gap to close. I wonder what the magic number would be on effectiveness to make the AI more cost efficient.

16% effectiveness for complex tasks vs 88% is a big gap to close. I wonder what the magic number would be on effectiveness to make the AI more cost efficient. Will we all become AI quality control supervisors, reviewing work and stepping in to finish random incomplete tasks that the llms get hung up on? Would be interesting to ponder.