Proponents of the growing field argue that the early weirdness of AI chatbots can be avoided by a human giving the machine all the right instructions.
