Amazon is set to unleash a revamped version of its Alexa voice assistant, and it’s generating a lot of interest. During their recent fall hardware event, the company introduced an all-new Alexa, powered by their new Alexa large language model. According to Dave Limp, Amazon’s SVP of devices and services, this upgraded Alexa boasts the ability to understand conversational phrases, provide more context-aware responses, and handle multiple requests in a single command.
Voice assistants have been in need of a fresh approach for some time. Their progress has been somewhat stagnant, with incremental improvements in comprehension that have left them feeling more like basic tools than the groundbreaking technological wonders we initially envisioned over a decade ago.
Generative AI has long appeared to be the lifeline for these voice assistants. While they’ve always had elements of AI, they’ve lacked the intricate processing capabilities and human-like interactions that generative AI can offer. This development marks a significant milestone for smart homes, potentially elevating home automation from a mere remote control experience to a genuinely intelligent and responsive household.
Dave Limp explained that the fresh Alexa LLM is “a genuinely specialized large language model, finely tuned for the specific needs of Alexa. It’s not akin to something like Bard or ChatGPT.”
However, this brand-new Alexa won’t be an immediate, widespread release. Amazon is taking a gradual approach by introducing it through a preview program “in the coming months,” and initially, it will only be available in the United States. It’s clear that Amazon is treading carefully, likely drawing lessons from the missteps of Microsoft and Google in their AI endeavors.
Dave Limp emphasizes the importance of minimizing “hallucinations” when connecting a Large Language Model (LLM) to the real world. He acknowledges that while they believe they have robust systems in place, there’s no substitute for real-world testing. To express your interest in joining the preview, simply instruct your Echo device, saying, “Alexa, let’s chat,” and your request will be noted.
It’s hardly surprising that this advanced version of Alexa might come with a price tag at some point. Limp clarified that while the current version of Alexa will remain free, the concept of a supercharged assistant capable of enhancing your smart home experience and handling complex tasks might eventually lead to a fee for its use.
The upcoming changes to Alexa bring a major improvement in the form of a more conversational assistant. This means you won’t need to be as specific or use particular terminology to get things done, addressing a common source of frustration with voice assistants. No more having to repeat yourself when you ask Alexa to adjust the thermostat, or dealing with responses like, “A few things share the name ‘lights.’ Please choose unique names and run discovery again,” for the umpteenth time when all you want is to locate your remote control.
With the new Alexa, you can simply say something like, “Alexa, I’m feeling chilly,” and it should take care of raising the temperature on your connected thermostat. As Dave Limp explained, you can even say, “Alexa, make this room feel like the Seahawk colors,” and it will understand which room you’re in and what the Seahawk’s colors are, seamlessly bridging the gap between various APIs to fulfill your request.
According to Dave Limp, the crucial element here lies in the Application Programming Interfaces (APIs). He mentioned, “We’ve integrated a vast array of smart home APIs, over 200 of them, into our Large Language Model (LLM).” This wealth of data, coupled with Alexa’s awareness of the devices in your home and the specific room you’re in (determined by the Echo speaker you’re interacting with), equips Alexa with the context required to effectively and seamlessly manage your smart home.
This contextual understanding goes beyond merely identifying which connected devices you want to control; it extends to tasks like recognizing changes in your home setup. Limp elaborated, “If you introduce a new device into your home, you can simply say, ‘Alexa, turn on the new light,’ and it will discern which light you mean. It excels at disambiguating, making it more straightforward to control newly added smart plugs or lights.”
Another impressive feature is Alexa’s ability to handle multiple requests simultaneously, taking it beyond the basics it could already manage to some extent. As Limp puts it, “You can say something like, ‘Alexa, turn on the sprinklers and open my garage door, and turn off the outside lights,’ and it will intelligently sort out all those tasks.”
This remarkable capability will also extend to creating Routines using voice commands alone, without any need for manual programming in the Alexa app. As Dave Limp described it, “I set up one this morning for my child simply by saying, ‘Alexa, every morning at 8AM, turn up the light, play wake-up music for my kid in his bedroom, and start the coffeemaker.’ And just like that, as complex or vague as you want it to be, it will instantly appear in your app as a Routine.”
Initially, this multi-command feature will be compatible with a select group of device types, including lights and smart plugs, among others, according to Limp. However, the team is actively working to expand its compatibility to encompass a broader range of devices in the future.
Developers will have the opportunity to harness Alexa’s new cognitive capabilities. Amazon is introducing two tools known as Dynamic Controller and Action Controller, which empower the new Alexa to manage unique features of third-party products, even those not typically found in Amazon’s smart home ecosystem toolkit.
Dynamic Controller, for instance, will make it easier to use features like predefined lighting scenes. For instance, if you have GE Cync-colored light bulbs and you say, “Alexa, make it look spooky in here,” Alexa will comprehend your request without the need for you to create a Routine or import scenes into the Alexa app.
Similarly, Action Controller allows developers to incorporate straightforward actions that Alexa can execute. For instance, if you say, “Alexa, the floor is dirty,” the assistant will understand that you want the robot vacuum to start cleaning.
Amazon has already begun collaborating with companies like GE Cync, Philips, GE Appliances, iRobot, Roborock, and Xiaomi on implementing these features, and they plan to extend the program to more developers. While Amazon hasn’t provided specific details on how or when developers can access these tools, they have been contacted for further clarification.
Dave Limp envisions that this marks just the beginning of Alexa’s exciting new path. He explains, “We’ve developed a fresh generative AI Large Language Model (LLM) that will gradually power various aspects of Alexa, including a range of innovative smart home experiences. The initial focus is on simplifying everyday tasks.” Where Alexa’s journey will lead next promises to be a captivating evolution.
The new Alexa, driven by the Large Language Model (LLM), will kick off with a preview launch in the United States, available to anyone with an Echo device. While Amazon hasn’t disclosed a specific date for the preview, they have noted that the new smart home features powered by Alexa LLM will be part of an exclusive invitation-only preview. Once you’re part of the preview, you can request an invitation when it becomes available at a later date.