[frankinstien]

Here's a company that has a digital assistant with a photo-realistic human face and facial expression:

Workforce.ai

I've been trying to find a tool that can produce this kind of visual realism in a virtual agent. The conversation is OK as what is usually seen today but it becomes obvious that her responses are based on key words or combinations of key words where the A.I. just iterates a scripted response. But the real question becomes is this approach that much better than just hearing a voice? Her facial expressions are limited and I asked her if she could smile and it did smile which surprised me, but when I ask if she could frown she directed me to her documentation. So the facial expressions, which we humans use either to convey an emphasis or deliberately fool (aka poker face), doesn't seem applicable in this kind of application.

For me the greatest test of an A.I. is if it can lie to suit its best interests or those it has responsibility for. I mean how often do we humans tell little white lies, just to be polite or avoid an escalation of a conflict?...