The recently concluded Google I/O developer conference had so much to offer. But one thing that captured the attention of many was Duplex, a new AI that can call and talk on the phone like a real person. The unveiling of Duplex came as a huge surprise. But nonetheless, the demonstration of the AI calling and booking a haircut was quite impressive. What’s more futuristic though was the way the AI held the conversation. It asked the right follow up questions, and it even managed to throw in an odd “mmhmm” and paused at some points during the conversation just like a normal human would.
The crowd was obviously surprised. But that wasn’t the most important thing. What really seemed to get everybody on their feet was the fact that the person on the other side of the phone couldn’t even tell he was speaking with a machine. It was just a regular conversation for him. There’s no doubt that this is a huge tech leap for Google. However, there are still a number of unresolved issues when it comes to privacy and ethics.
For instance, if Google Assistant would call and make an appointment with the barber on your behalf, it would probably need access to some of your private info. What will Google do with that? There are also other social issues. Will Google have to notify people that they are in fact speaking to a machine and not a human? In addition to this, if we can build a technology that can mimic how humans speak, what does that do to regular, everyday conversions? Will people have to worry every time they get a call that perhaps it’s a machine on the other end and not who they think it is? These are all very important questions and Google hasn’t addressed them yet.
Nonetheless, the new AI is not as futuristic as it seems. It’s not really a technology that can have a conversation with humans because it can only work in the right context. For example, you can easily use Duplex to schedule an appointment or reserve a table in a restaurant. But the AI wouldn’t know anything to say beyond this context. AI experts say that developing an AI that can talk with humans in a random setting is simply impossible.
Google has also made it clear that Duplex is just an experiment. The company confirmed that the technology will be deployed under human supervision, and this is part of its training. The human element will be withdrawn only if Google thinks that the AI can hold conversations in the expected way. In addition to this, in cases where conversations don’t go well, there will be a human standing by to take over from the machine. The only challenge now for Google is on how to deploy the AI within the required ethical standards. The search engine giant is looking at AI as an integral part of its future operations.