Google's AI voice calling assistant raises controversy
In May 2018, Google announced an AI system to carry out tasks such as scheduling appointments over the phone using natural language. A Duplex user wanting to make a restaurant booking, for example, could hand the task off to Duplex, which would make the phone call and negotiate times and numbers. In announcing the service, Google stressed its use of "speech dysfluencies" - that is, non-verbal syllables such as "um" and "er" to make the interaction sound more natural.
The system almost immediately met controversy because the company's initial demonstrations focused on the surprise people on the receiving end of these calls displayed when they discovered they were talking to an AI agent. Google quickly promised that such calls would always be identified as coming from AI assistants.
Privacy concerns were raised by the fact that processing the receiving human's replies and formulating responses would of necessity be done on Google's servers. This would appear to violate the law in areas where recording calls requires the consent of both parties. In addition, it's unclear when an answering human might have the chance to find out whether and for how long Google is retaining this data. A final concern is the impact on human service workers and their employers of AI systems that sever the human link between customer and supplier.
Publication: Google blog, The Verge, TechCrunch
Writer: Google, Natt Garun, Devin Coldewey