I find this quite scary, main reason being that 1) depression is not easily diagnosed and therefore you cannot train machines with the right information and 2) voice is not reliable. I’m not saying the AI won’t pick up on depression, it probably will, but there will be lots of false negatives i.e. missed diagnoses that do not make it to the next stage which I hope involves a real human. We are all so complex, we have different types of depressions. On the other hand, false positives may alert us to the fact that depression, like all mental disease, is a continuum of severity as well as type and often correlating and coexisting with other mental illness. It may help the evolution of psychology. I guess we need to dive in to find out, but I can already see too many ways in which this can go wrong and ruin people’s lives.