Amazon’s revamped Alexa, set to launch in October ahead of the U.S. holiday season, will be based primarily on Anthropic’s Claude AI models rather than its own AI, five people familiar with the matter told Reuters.
As Reuters reported in June, Amazon plans to charge $5 to $10 per month for the new “Remarkable” version of Alexa, as it will use advanced generative AI to answer complex queries, while the “Classic” voice assistant will be offered for free.
But the first versions of the new Alexa, using internal software, simply had trouble finding the words and sometimes took six or seven seconds to confirm a command and respond, one of the people said.
That’s why Amazon turned to Claude, an AI-powered chatbot developed by startup Anthropic, because it proved to be more effective than the online retail giant’s own AI models, the authors say.
Reuters based this story on interviews with five people with direct knowledge of Alex’s strategy, all of whom declined to be named because they were not authorized to discuss nonpublic matters.
Alexa, available primarily through Amazon TVs and Echo devices, lets you set timers, play music, act as a central control hub for your smart home, and answer one-time questions.
Yet Amazon’s attempts to convince users to shop through Alexa and thereby increase revenue have largely been unsuccessful, and the division continues to lose money.
As a result, senior executives have stressed that 2024 is a critical year for Alexa, where the company must finally show it can generate significant sales. The revamped paid version is seen as a way to achieve that goal and keep up with the competition.
“Amazon uses a variety of technologies to power Alexa,” a company spokeswoman said in a statement responding to detailed questions from Reuters for this article.
“When it comes to machine learning models, we are starting with those created by Amazon, but we have used and will continue to use a variety of models – including (Amazon AI model) Titan and future Amazon models, as well as models from partners – to ensure we provide the best customer experience,” the spokeswoman said.
Anthropic, in which Amazon has a minority stake, declined to comment for this article.
AI Partnerships
Amazon generally avoids using technologies it has not developed itself, so it can have full control over how the site is used, how data is collected and how direct relationships are with customers.
But it wouldn’t be alone in turning to a partner to improve its AI products. Microsoft and Apple, for example, have partnered with OpenAI to use ChatGPT to power some of their products.
Remarkable Alexa, as the device is called internally, is expected to launch in October, with an announcement of the new service to be made at Amazon’s annual event dedicated to devices and services, which usually takes place in September, the sources said.
Amazon has not yet announced a date for the planned conference, which will be the first major public appearance by its new devices chief, Panos Panay, who was hired last year to succeed longtime executive David Limp.
The launch in late 2022 of ChatGPT, which provides near-instant answers to complex queries in full sentences, has sparked a frenzy of corporate investment and maneuvering to develop better AI software for a variety of functions, including image, video and voice services.
Amazon employees told Reuters that by comparison, Amazon’s decade-old Alexa seems outdated.
While Amazon has a motto of “working from start to finish, starting with the customer” to invent new services, some people say the Alexa group has been focused on keeping up with competitors in the AI race over the past year.
Amazon employees also expressed skepticism about whether customers would be willing to pay $60 to $120 a year for a service that is currently free — on top of the $139 many of them pay for a Prime membership.
Alexa Improvements
The paid version of the Alexa app will allow users to conduct conversations based on previous questions and answers, according to people familiar with Alexa’s strategy.
The improved Alexa is designed to let users seek out shopping advice, such as what clothes to buy for vacation, and aggregate news, the people said. And it is expected to perform more complicated requests, such as ordering food or composing emails, all from a single prompt.
Amazon hopes that the new Alexa will also become an advanced home automation hub, remembering user preferences, so that, for example, you can set a morning alarm, and the TV will know when to record your favorite programs even if you forget to do so.
The company’s plans for Alexa could be delayed or changed if the technology fails to meet certain internal criteria, the people said, without providing further details.
Bank of America analyst Justin Post estimated in June that there are about 100 million active Alexa users and that about 10% of them might opt for a paid version of Alexa. Assuming the lower end of the monthly price range, that would generate at least $600 million in annual sales.
Amazon says it has sold 500 million Alexa devices, but doesn’t disclose how many active users they have.
When announcing a deal to invest $4 billion in Anthropic last September, Amazon said its customers would get early access to its technology. Reuters could not determine whether Amazon would have to pay Anthropic extra to use Claude on Alexa.
Amazon declined to discuss details of its deals with the startup. Alphabet’s Google has also invested at least $2 billion in Anthropic.
The retailer, like Google, is facing a formal investigation by the U.K. antitrust regulator into the Anthropic deal and its impact on competition. It announced a preliminary investigation in August and said it had 40 business days to decide whether to move it to a more heightened stage of scrutiny.
The Washington Post previously reported that the new Alexa is set to launch in October.
© Thomson Reuters 2024
(This story has not been edited by NDTV staff and is auto-generated from a feed.)
