Zoom Image

Amazon manager Dave Limp at a presentation in 2019

Photo: Andrej Sokolov / dpa

It may have been the last time Dave Limp was on the big Amazon stage. The prominent hardware boss will leave the group at the end of the year. But before that, he really let it rip again in Arlington, in the US state of Virginia, where Amazon has set up a second headquarters. A new Echo Show, tablets for children and several other gadgets were on the program. But above all, a new version of Alexa was enthroned, which is supposed to become more understanding and human.

Amazon has developed its own new language model for this, similar to those used by chatbots such as ChatGPT and Bard. Unlike its competitors, however, Amazon's variant of this technology is specially optimized for voice control, Limp said. And to provide real-time access to information, control smart homes and access entertainment content. The focus of the development was on making Alexa look more human with the help of the new technology.

Talk to Alexa without saying "Alexa"

This includes the fact that in the future you will not have to pronounce the introduction "Alexa" again and again during a conversation with Alexa. Rather, by saying "Let's chat," you start a conversation to German, "Let's chat." Dave Limp showed that in the further course of such a conversation you can talk to Alexa as if you were talking to a real person by using artificial intelligence (AI) to conduct a conversation that is safely rehearsed, but still natural-looking.

For example, he asked how his favorite football team was doing and received a correct answer that referred to the right team. Alexa had apparently previously learned which sports results the manager was particularly eager to follow. A little later in the conversation, Alexa was also able to formulate a natural-looking invitation to a barbecue party with friends. Remarkably, during the demo, which lasted several minutes, Dave Limp actually didn't have to say "Alexa" once to keep the conversation going.

However, he had to repeat himself a few times to make his inputs clear to the system. In addition, Alexa's answers were always correct and often enriched with images, but sometimes came with a noticeable delay. It is unclear whether this was due to the WLAN in the event hall or to weaknesses in the software.

Alexa is supposed to stumble linguistically

Alexa's new generative AI will also be used on Fire TV sticks and boxes in the future to make it easier for users to search for movies and series with voice input. During a demonstration of the new system, the system also provided inaccurate information such as "Show me movies with the guy, you know, who made this movie..." correctly interpreted.

A little later, a new version of Alexa's voice acting was also demonstrated on stage, which is less technical, but instead much more human because it is more emotional. This is exactly the goal: In the future, AI should sound more emotional, even integrate typical human interruptions such as ahhs, ahhs and ohs.

Free for now

However, not everywhere and for everyone. The new version of Alexa is expected to launch in the coming months and will work with all Echo devices, including the first model from 2014. However, as is so often the case with Amazon, this only applies in the USA for the time being. It remained unclear when an introduction in other countries is being considered. In addition, Limp said in an interview with the tech magazine »The Verge« that the new Alexa version will remain free for the time being, but it doesn't always have to stay that way.

Such an advanced digital assistant "that can improve your smart home and also perform complex tasks for you could be so useful that we will end up charging for it," the Amazon executive explained.