This post was originally published on this site
Since it was unveiled in 2018, Google Duplex has been making waves in the world of artificial intelligence and voice assistants, steadily gaining attention and popularity.
So, what exactly is this technology and what it can do?
Simply put, Duplex is a voice assistant that uses natural language processing and deep learning algorithms to carry out complex tasks and conversations in a way that is indistinguishable from a human. It can make phone calls, book appointments and even engage in casual conversation with people on the other end of the line.
Read: Microsoft will benefit from ChatGPT, OpenAI in multiple ways — potentially at Google’s expense
One of the key features that makes Google’s AI so special is its ability to mimic the rhythm, tone and inflections of human speech. Duplex sounds incredibly natural and lifelike when interacting with people over the phone, which can be especially useful for making restaurant reservations or scheduling appointments. It is also able to handle interruptions and digressions in a conversation, and it can even use fillers like “um” and “ah” to sound more like a real person.
Another key aspect of Google Duplex is its ability to understand and interpret the nuances of human speech. It can understand the context and meaning of words and phrases, even colloquialisms and slang. This enables it to carry out complex tasks like booking a hotel room or scheduling a haircut, even if the person on the other end speaks in a way that is not perfectly clear or formal.
The technology behind Google Duplex is truly impressive, and it has the potential to revolutionize the way we interact with machines. Google
GOOG,
is well aware of AI’s capabilities but still hasn’t found a way to truly monetize its potential. Case in point: Duplex on the Web.
Duplex on the Web was a set of artificial-intelligence-powered services developed by Google that aimed to simplify the process of ordering food, purchase movie tickets and more by navigating websites on behalf of the user. It was introduced at the 2019 Google I/O developer conference and initially focused on a few narrow use cases, such as filling out information on a movie theater’s website to book tickets.
However, it has later expanded to include features such as helping users automatically change passwords exposed in data breaches, assisting with checkout for e-commerce retailers and finding discounts on flight check-ins and other sites. At first, Duplex on the Web could only be used on Android devices; in late 2019, the service became available to Chrome for Android as “Assistant in Chrome.”
However, in late 2022, Google announced that it is shutting down Duplex on the Web and any automation features enabled by it as of December 2022. A Google spokesperson said the company is “fully focus[ing] on making AI advancements to the Duplex voice technology that helps people most every day.”
It is unclear why exactly Duplex on the Web was discontinued, but it is possible that the technical lift required to maintain the service and the resource-intensive nature of its AI models may have been too much for the company to justify.
Additionally, some brands may have been uncomfortable with Google inserting itself between them and their customers, and recent cuts to the Assistant side of Google’s business may have also played a role in the decision to terminate the service.
It seems as if the tech giant is still trying to find a killer use case for Duplex. But as people are warming up to AI, it’s only a matter of time before it becomes an important building block in making AI indistinguishable from an average human.
An important step on that journey could be integration with OpenAI’s ChatGPT. (Microsoft
MSFT,
is considering investing $10 billion in OpenAI, deepening ties with the company.) Google Duplex would provide perfect articulation and vocal recognition, while ChatGPT would be able to hold complex and coherent conversations as well as provide detailed information about a variety of topics, adding even more value to daily human-AI interactions.
Another no-brainer would be a semantic, AI-powered, natural-sounding search engine. It wouldn’t simply be a Google search engine that could talk, but an AI model that could use real-time data to train itself to understand current events, read website content and provide information just like a human.
It would be your personal data broker with a pleasant voice and superhuman search capabilities. Instead of typing a string to search for, you would be able to ask it to combine existing resources and create diagrams, infographics, answer complex questions and even create meta studies! If this sounds to you a bit like Duplex on the Web, you’re not wrong.
As we saw from that failed experiment, a service like that would likely present a significant cost for Google, as it wouldn’t simply rely on an existing pipeline, but augment or even replace it with a different way of processing and serving information.
Advertising in this new mode would also be something Google would need to rethink. Risking its main revenue stream just to improve the user experience isn’t something Google would voluntarily do, so it made sense for the company to scrap Duplex on the Web. However, other competitors in the space will most likely force the tech giant to revisit the project sooner rather than later.
ChatGPT has opened Pandora’s box and it cannot be closed. So in the near future, we can expect to see AI assistants capable of holding natural language conversations related to current events and real time data become an increasingly common way of searching for information on the world wide web. All they need is access to the internet and an appropriate training criterion. Once that’s resolved — and it’s no small task — all bets are off.
Whether Google will lead the way or (scramble not to) fall behind remains to be seen.
What about you? What is your experience with Google Duplex? Let me know in the comment section below.
