And little is known about how this AI system actually works.
The robots on the other side of the customer support line could soon start to sound a lot more human.
Google is reportedly shopping its Duplex AI system around as a tool for call centers, according to The Information, including a large insurance company.
Duplex would handle simple calls for the insurance company, and if the customer started asking complex questions the bot can’t handle a human would step in, according to the report. However, it’s unlikely that AI research will cease after mastering simple conversations, meaning call centers could one day be largely automated using this technology.
The AI system was first debuted at Google’s I/O developer conference in May, where it was demonstrated making calls to local businesses to place reservations on behalf of Google Assistant users. After public outcry at the implication of people in the future not knowing whether they were talking to humans or machines, Google adapted the bot’s introduction so it clearly explains it’s not a human.
Little is known about how Duplex actually works. Unlike many other AI systems Google has shown off, like evolutionary algorithms and voice-generating AI, the company hasn’t released a research paper detailing how conversational AI works.
The Information report, coupled with a flurry of media coverage when Google let reporters test the technology, shows Google’s confidence in the new technology as a potential offering to surpass its cloud competitors selling AI tools. Typically these tools range from image recognition to speech-to-text software that other companies can build into their own apps or websites.
Update: A Google spokesperson reiterated that Duplex is only being tested as a consumer technology for now, and that the company isn’t testing it for enterprise. The entire statement is below:
We’re currently focused on consumer use cases for the Duplex technology and we aren’t testing Duplex with any enterprise clients. As we shared last week, Duplex is designed to operate in very specific use cases, and currently we’re focused on testing with restaurant reservations, hair salon booking, and holiday hours with a limited set of trusted testers. It’s important that we get the experience right and we’re taking a slow and measured approach as we incorporate learnings and feedback from our tests.