Google’s I/O 2018 conference kicked off with a lot of talk and demos of AI and machine learning. It painted a future of wonderful convenience but left audiences and readers wondering about the ramifications of an AI-powered world. This dichotomy, confusion, and fear was perhaps best shown in the demonstration of Duplex, an AI that sounded so human that it could book a restaurant reservation without the human on other side being the wiser. And thus the AI disclosure debates were born.
Google Duplex seems to have reached the uncanny valley of AI voice technology. It sounded so authentically human that it actually managed to worry even tech pundits and AI supporters. Should Duplex identify itself as non-human or would that significantly diminish its chances of success and its efficacy?
More than just ethical consideration of not knowing that you’re speaking to a bot, there’s also the element of privacy. Like with anything involving Google, you send a tiny bit of yourself or your information when you interact with its products. The person on the other end of the line might not know that and will therefore be sending data to Google without his or her knowledge much less consent.
Doing damage control, Google assures the public that Duplex will disclose its true nature. Even if Sundar Pichai’s I/O demo never included that aspect. The company is just still in the process of figuring out how to best make that disclosure, probably in a way that won’t make the other party hang up in fear.