BY TODD BISHOP on August 30, 2024 at 9:21 am

In-depth Amazon coverage from the tech giant’s hometown, including e-commerce, AWS, Amazon Prime, Alexa, logistics, devices, and more.

Dario Amodei, CEO of Anthropic, right, speaks about the company’s partnership with Amazon at the AWS re:Invent conference in November 2023. With him is Adam Selipsky, who has since left his role as AWS CEO. (GeekWire File Photo / Todd Bishop)

Amazon will make its Alexa voice assistant more conversational with help from Anthropic’s Claude artificial intelligence technology, Reuters reported Friday morning, citing interviews with five people familiar with the matter.

The Seattle-based tech giant is attempting to catch up to the capabilities of OpenAI’s ChatGPT, which has expanded beyond text-based interactions via an increasingly sophisticated voice chat feature.

Reuters reported previously that Amazon plans to charge a monthly fee of $5 to $10 for users to access the more conversational version of Alexa, while still offering a more basic version of its voice assistant for free.

The decision to leverage Anthropic’s technology came after homegrown versions of Alexa’s generative AI responses “simply struggled for words, sometimes taking six or seven seconds to acknowledge a prompt and reply,” the news agency reports.

Amazon has invested $4 billion in Anthropic for a minority stake in the San Francisco-based AI company as part of a broader partnership that includes offering Anthropic’s AI models via Amazon Web Services.

On the surface, the move could be seen as a setback for Amazon in its two-year quest to dispel the notion that it has been caught flat-footed by the debut of ChatGPT and the broader rise of generative artificial intelligence.

In a statement following the report, Amazon offered a more nuanced explanation.

“Amazon uses many different technologies to power Alexa,” a spokesperson told GeekWire. “When it comes to machine learning models, we start with those built by Amazon, but we have used, and will continue to use, a variety of different models—including Titan and future Amazon models, as well as those from partners—to build the best experience for customers.”

Transitioning into a pitch for Amazon’s managed service for AI models, the statement continued, “That’s one of the key advantages of Amazon Bedrock—it makes it easy for application developers to seamlessly use the world’s highest-performing models, and multiple models, through a single API.”

As reflected in a recent talk by Swami Sivasubramanian, the AWS vice president of AI and data, it’s becoming more common for companies to use a variety of AI models for single applications.

It’s not as simple as turning Alexa’s voice over to Claude. Alexa uses a variety of foundational AI technologies, including speech recognition and text-to-speech, in addition to language models.

Amazon last fall previewed a new feature, dubbed “Let’s Chat,” driven by generative artificial intelligence, designed to make its Alexa voice assistant more conversational on its Echo devices. It has yet to be released publicly.

Since then, there’s been a changing of the guard in the Amazon Devices & Services business, with former Microsoft executive Panos Panay joining Amazon to lead the division, replacing Dave Limp, who became Blue Origin CEO.

Reuters and the Wall Street Journal have reported previously that the Amazon devices division is under internal pressure to improve its bottom line after years of financial losses.

Amazon typically holds a launch event unveiling new devices and services each fall. A date for this year’s event has yet to be announced.

GeekWire co-founder Todd Bishop is a business and technology journalist who covers topics including Amazon, Microsoft, and startups. He also hosts GeekWire's weekly podcast. Twitter: @toddbishop. Email: todd@geekwire.com. Contact him via phone, ?(530) 230-3439?, or encrypted messaging app Signal, 1-206-300-0265.