Amazon Rebuilt Alexa Using a ‘Staggering’ Amount of AI Tools

Daniel Rausch, Amazon vice-president of Alexa And Echois in the middle of a major transition. More than a decade beyond the launch of Alexa of Amazon, he was responsible for creating a new version of the brand vocal assistant, which is fueled by Great language models. As he said in my interview with him, this new assistant, nicknamed Alexa +, is “a complete reconstruction of architecture”.

How did her team approach Amazon’s largest overhaul never condemned to his voice assistant? They used IA To build AI, of course.

“The rate with which we use tools on AI through the construction process is quite amazing,” explains Rausch. When creating the new Alexa, Amazon used AI at each stage of the construction. And yes, this includes the generation of parts of the code.

The Alexa team also introduced a generative AI in the test process. The engineers used “a large language model as a judge on the responses” during the learning processes of strengthening where the AI ​​selected what it considered the best answers between two Alexa +outings.

“People get the lever effect and can move more quickly, better thanks to AI tools,” explains Rausch. Amazon’s emphasis on the use of internal generative AI is part of a greater wave of disturbances for software engineers at work, because new tools, such as the Anyspirre cursor, modify the way the work is done, as well as the expected workload.

If these types of workflow The engineer will fundamentally change. “We will need fewer people who are working today, and more people who make other types of jobs,” said Amazon Andy Jassy CEO A memo This week to employees. “It is difficult to know exactly where it takes place over time, but in the coming years, we expect it to reduce our total workforce of the company as we obtain efficiency gains by largely using AI throughout the business.”

For the moment, Rausch mainly focuses on the deployment of the genetive version of Alexa to more Amazon users. “We really didn’t want to leave customers’ customers,” he said. “And that means hundreds of millions of different devices that you should support.”

The new Alexa + discussions more conversational with users. It is a more personalized experience that remembers your preferences and which is able to complete the online tasks you give it, such as the search for concert tickets or the purchase of grocery stores.

Amazon announced Alexa + during a company event in February and deployed early access to a few public users in March, although it is without the full list of announced features. Now the company says that more than a million people have access to the updated voice assistant, which is still a small percentage of potential users; Finally, hundreds of millions of Alexa users will have access to the AI ​​tool. A wider release of Alexa + is potentially planned Later this summer.

Amazon faces the competition from several directions because it works on a more dynamic vocal assistant. Advanced vocal mode of OpenaiLaunched in 2024, was popular with users who found the voice of engaging AI. Also, Apple announced a overhaul of its native vocal assistant, SiriDuring the conference of developers last year – with many contextual and personalization features similar to what Amazon works with Alexa +. Apple has not yet launched the Siri rebuilt, even in early access, and the new voice assistant is expected next year.

Amazon has refused to give Alexa + precocious access for practical tests (voice-on?), And the new assistant has not yet been deployed on my personal Amazon account. Similar to the way we approached the advanced Openai vocal mode which was launched last year, the wired plans to test Alexa + and provide an experiential context for readers as it becomes more widely available.