‘AI system should know its own limits’
Every artificial intelligent system makes mistakes, said Professor Inald Lagendijk during his lecture on Engineering Meaningful Human Control for the Bataafsch Society last Monday. For an AI system that has to distinguish apples from pears, mistakes might be amusing, but for an HR system that selects candidates for a job, they are a lot less so.
It is therefore important that AI systems are designed to be aware of their own limitations. They should then transfer the decision-making to a responsible person as soon as a task is outside its domain. Does this ability already exist? No, says Lagendijk, but that is where we need to go. Delft AiTech and the NL AI coalition are working on integrating moral values such as privacy, honesty, safety and transparency into AI applications.
Comments are closed.