Amazon's Alexa+ Becomes a Smarter AI Assistant

A Glimpse into Amazon’s Secret Apartment

Around 1 p.m., local time, on the 5th, at Amazon’s headquarters in Seattle, Washington State. After stepping off the elevator on the 30th floor of the Day One Building and passing through a door marked “Restricted Area,” a “household” space appeared. This large studio-like area was furnished with appliances such as a refrigerator and TV, as well as furniture including a sofa, arranged as if someone actually lived there. Amazon smart speakers “Echo” of various designs were installed throughout the space. Daniel Rausch, Amazon’s vice president of the Alexa and Echo division, who was standing behind the kitchen island, asked the device questions like, “I want to buy a coffee machine. Can you recommend one?” and “Take a photo of the space in front of you and share your thoughts.”

This is a laboratory for developing Amazon’s voice assistant “Alexa” and the Echo devices that house it. The residential space was recreated to test the technology in an environment identical to where real customers use the products, allowing for direct testing and improvements. On that day, two speakers and two devices with displays, similar to tablet PCs, were placed in the kitchen area. Amazon, which calls this lab the “Amazon Secret Apartment,” first revealed it to an international press delegation, and among domestic media, this newspaper was the only one present.

Alexa Evolves into a Smarter AI Assistant

Alexa, the voice assistant built into the voice-recognizing smart speaker “Echo,” was first released in 2015 as a voice assistant that played music and set alarms. At the time, there were reactions that “a speaker that understands and responds to speech” had turned science fiction into reality. Starting with 13 functions, Alexa later evolved into a home manager that helped control smart homes. However, as time passed, most users only utilized basic functions like music, weather, and alarms, leading to a slump in early 2020. According to U.S. Fortune and other foreign media, over 2,000 employees were laid off from the Alexa and Devices division between 2022 and 2023. At the time, some analysts called Alexa “Amazon’s failed innovation.”

However, Alexa has recently evolved into a much smarter assistant. In March, Amazon released “Alexa Plus (+),” which adds AI agent capabilities to the existing Alexa. Like ChatGPT, it can engage in deep conversations on any topic, manage smart homes, and perform actions like shopping and reservations on behalf of users. It remembers past conversations to provide personalized responses and actions. The Echo devices, which now come with Alexa+, have gradually evolved to include models with displays, becoming assistants that can see, hear, and speak. Vice President Rausch said, “The Alexa and Echo combination can interact with people instantly using eyes and ears.”

At the site that day, when asked, “Can you tell me about the ice hockey game in Seattle tonight?” Alexa responded, “There is a game against the San Jose team at 7 p.m.” When asked, “I want to buy tickets,” it found the tickets and replied, “They start at $45.6.” Then, when asked, “I want two seats near the ice,” it recommended seats in Section 24, Row AA, and asked if the user wanted to purchase them. If the user agrees, Alexa can handle the ticket purchase.

There was almost no delay in responses. Alexa answered most questions within 5 seconds, except for the query, “Tell me a joke using the word ‘Press,’” which took 7 seconds. There were no restrictions on conversation topics. It could engage in discussions not only about general knowledge but also academic subjects like science and geology. An Amazon source said, “The most challenging part was preventing delays, as in human conversations,” adding, “It can handle any topic in any category.”

Amazon’s Strengths in the AI Agent Era

Alexa+’s strength lies in its “multi-model” strategy. Over 70 large language models (LLMs) are connected to Alexa+. When a user gives a specific task, the system automatically selects the most suitable model. Alongside this, it has an “AI Orchestration” function that connects with thousands of system APIs and external systems to handle complex commands effectively. For example, a command like “Summarize the meeting and add it to my schedule, then find related news” is executed simultaneously, much like an orchestra.

Through this, Amazon aims to strengthen its competitiveness in the AI agent era. The AI technology race has evolved from LLM competition to multimodal model competition, combining images and videos, and recently shifted focus to AI agent competition, where agents predict and act directly. Although Amazon has its own AI models, including LLMs, it has sometimes been evaluated as less competitive compared to OpenAI’s ChatGPT and Google’s Gemini.

Beyond developing Alexa+, Amazon Web Services (AWS), Amazon’s cloud service subsidiary, is preparing for the full-fledged AI agent era by operating the “Bedrock Agent Core,” an AI agent platform that collaborates with “AI colleagues” possessing various abilities and features. Swami Sivasubramanian, AWS’s vice president of Data and AI, said, “AWS’s goal is to be the best place for everyone to build and deploy agents in a trustworthy, secure, and accurate manner.”

In particular, AWS, as the world’s leading cloud service provider, claims that its expertise in securely storing data and operating services for numerous companies will be a strength in developing and operating AI agent systems. Continuous large-scale infrastructure investments and the development of custom chips are also strengths. Matt Garman, AWS CEO, said, “AWS is continuing large-scale investments, planning capital investments of $125 billion over 12 months,” adding, “After introducing AI agents, we saw remarkable results with team efficiency improving by 4 to 5 times.”