Google Duplex, the tech giant’s experimental voice technology, which can carry out human-like phone conversations, sure sounds like something banks could put to good use. But if you ask bankers or security experts, it’s not likely to happen anytime soon. If at all.
Google’s AI-based experimental voice assistant wowed the entire world last month. The new software sounded exactly like a human when it rang a hair salon to set up an appointment and booked a table at a restaurant. It’s the quality of the voice and its natural-sounding conversation style that sets Duplex apart from other voice assistants, like Amazon’s Alexa, Apple’s Siri or Microsoft’s Cortana, which use synthetized voices.
As Google put it, “even with today’s state of the art systems, it is often frustrating having to talk to stilted computerized voices that don’t understand natural language”. To make things worse, automated phone systems are still struggling to recognize the simplest of words and commands. “They don’t engage in a conversation flow and force the caller to adjust to the system instead of the system adjusting to the caller”, the tech giant explained.
So how does Duplex work? At its core, there’s a recurrent neural network, which was trained to speak, understand and interact just like a human, using a wealth of anonymized phone conversation data. It uses Google’s automatic speech recognition technology as well as contextual information like the history and the parameters of the conversation. Google has trained its understanding model separately for each and every task.
Banking: testing the waters
OK, that’s very well, but what can Duplex actually be used for? Testing will start this summer within Google Assistant to help users make restaurant reservations, book hair salon appointments and get holiday hours over the phone. But businesses are expected to benefit from Duplex, too: by allowing customers to book through Google Assistant, there will be no need for them to change any day-to-day practices or train employees. Using Duplex could also reduce no-shows to appointments.
But is there a chance that Duplex will also be able to tackle more complex tasks, and be used in financial services? Some banking executives have already envisioned how they will embed Duplex in partially automated processes, like collecting information for mortgage applications. “Having a bot to walk you through that process could be very helpful,” Eric Smith, chief data and analytics officer at Texas-based financial services group USAA, told The American Banker. And why couldn’t the bot become a personalized assistant to help with all kinds of things in the future?
The most obvious use case would be getting Duplex to answer calls from customers to slash the operation costs of call centers. But that may remain a pipe dream for a while. As Jacob Jegher, senior vice president of banking and head of strategy at Javelin Strategy & Research, pointed out, the industry has already cut down on call center volume successfully, having invested in improving mobile and online banking, for example.
The race is on
In any case, Duplex can only carry out natural conversations after being thoroughly trained in the given domain. In other words, it can’t just strike up a conversation. Not to mention that existing voice assistant systems, including Amazon Echo and Google Home, haven’t lived up to expectations of their possible uses in banking, either. According to a report by Tearsheet, bankers agree that voice will be the biggest and most important channel for their business after mobile but they first need to figure out how people use Alexa, for example.
It’s also yet to be seen how the fierce competition among tech giants to roll out more and more sophisticated voice assistants will change the picture. TechRepublic believes Microsoft could give Google a run for its money with its social chatbot XiaoIce, which can also pull off human-sounding phone calls and has already talked with over 600,000 people since its launch last August. While Xiaolce is only available in China, Microsoft has similar chatbots in other markets, including Ruuh in India, Rinna in Japan and Indonesia, and Zo.ai in the US.
Users also seem to be on board with using voice assistants. In the US alone, the number of such devices on all platforms, including smartphones, tablets, PCs, speakers, TVs, cars and wearables, will reach 870 million by 2022. This is a whopping 95% growth from 450 million estimated for 2017, Juniper Research predicts. And more than half of consumers asked would use Amazon’s Alexa to make routine banking requests, such as checking account balances (47%) and reviewing recent transactions (43%), according to Javelin Strategy & Research. So it might be a good idea for large banks to support emerging voice capabilities from Amazon and Google, and for smaller banks to follow suit.
Google Duplex: call security is critical
Per usual, the biggest obstacle to using Duplex in financial services, or in any other industry for that matter, is security and authentication concerns. Eric Smith at USAA has asked the tough questions: How can banks make sure that the conversation is with the right person? How would data flow? And last but not least, how much of that would Google need access to? Until these issues get sorted out, USAA will most probably test out Duplex in inquiries about balances and other types of information that are not transactions. Next up might be advice interaction on spending or saving.
Let’s be honest: these doubts are not without reason. Fraudsters could use the technology to assess conversation data and determine how best to get a call center agent to do their bidding, Vijay Balasubramaniyan, chief executive at Pindrop Security, told The American Banker. As long as the use of the new technology is restricted to scheduling appointments, there’s not much to worry about. But if hackers get their hands on voice recordings, they can synthesize the voice of the person speaking and possibly fool a human or a bot, experts warn.
And there’s also the ethical angle, of course. Using a machine to copy a person’s voice could potentially deceive the listener, the Financial Review says. “This technology is amazing, and a big step forward, but I don’t think the main goal of AI should be to mimic humans,” Erik Brynjolfsson, a Massachusetts Institute of Technology professor, reckons. Google, too, seems to share his opinion: shortly after their much talked-about demonstration of Duplex, the tech giant was quick to tell CNET that its new AI tool would identify itself when talking to people and disclosures would be promptly included in the feature.
