This month on our blog we have been talking about internet law related legislation and development. This week will continue the topic with a discussion of artificial intelligence and privacy issues. Specifically, we will talk about Google’s new Duplex technology, which allows users to make appointments, book reservations, and call businesses via an artificial intelligence robot.
Last month, Google introduced Google Duplex at their annual conference. Google’s CEO presented the technology on stage, and played a recorded phone call in which the digital assistant booked a hair appointment for a client. A second conversation of the digital assistant making a restaurant reservation was also played. What was immediately striking about these conversations is how realistic the robot sounded. Upon listening, the robot voice and interaction is nearly indistinguishable from a human conversation. The digital assistant uses phrases such as “uh” and “mmhmm,” and also pauses before responding as if it is “thinking” of a response. More often than not, the party on the other end of the line likely has no idea they are talking with a robot, or even further that they are being recorded. This may pose a liability for Google.
Google has been recording the conversations that the robots have in order to improve and modify the technology. By recording the conversation, engineers at Google can see exactly what phrases, or questions, make the robot function in a less than optimal way. When engineers and programmers are able to recognize these challenging variants in conversation, they are more able to program an appropriate response for the robot, so as to make the technology more user-friendly and successful. Problems arise in the legal world, however, when Google developers fail to tell the party on the other end of the line that they are being recorded.