X.ai develops AI assistants that sound enough like human beings to trick users into thinking they’re messaging with a fellow person.

CEO Dennis Mortensen gave insights to Google Cloud editorial head Quentin Hardy about building realistic bots and what they mean for human-machine interactions going forward.

Referring to the previous talk about the ethical dilemma of AI sex robots, Hardy said that there is no evidence that real sentient beings with a will can be created, but the fact they appear deeper and more emotionally relevant is not limited to these creations.

Looking at virtual assistants, Mortensen discussed how they have to talk and respond to the “crazy” way in which real people converse. “I think a very large amount of the apps in the app store shouldn’t be there,” he commented. “For every pain point we have, an app is not a solution.”

Looking at why virtual assistants are often young, educated women, this is a response to what focus groups are most comfortable with. However, it poses the dilemma of making money versus reaffirming prejudices.

Also, despite people knowing they are talking to a machine, Mortensen said that 11% of users express gratitude to virtual assistants for their help. Assistants are being developed to reflect sadness when they inconvenience people (e.g. rearranging meetings), as well as replicate small human mistakes, to make them appear more trustworthy.

“Voice is very primal, people sense this,” said Mortensen, discussing working with drama majors, schooled in human relationships when developing assistants. “We need to reflect society back on itself, with different designs for different people.”