Amazon is working on a paid subscription plan that would use an upgraded, AI-powered Alexa. However, the company is not satisfied with Alexa’s current performance internally, sources tell Business Insider.
New updates include the ability to control alarm and timer volume independently within the Alexa app. Plus, customers with speech or mobility disabilities can now gaze at their devices to play music and shows, control home smart-home systems, and call loved ones, hands-free.
It’s still not good enough
For a long time, Alexa seemed like the key to Amazon’s future. The voice assistant was a top priority for former CEO Jeff Bezos, and a big part of the company’s hardware team was dedicated to Alexa development. But that investment has never been able to translate into an adequate business model. The division has been plagued with layoffs and losses for years, and a new subscription plan is the latest attempt to boost revenue. But the move is also a major risk for Amazon’s most important technology.
The company wants to take the Alexa experience to a whole new level. Its new version, dubbed Remarkable Alexa, will be a more personalized AI assistant that can converse naturally with users. It will eliminate the need for constant wake words and will offer more empathetic responses to user inquiries. It will also be able to detect emotions and mimic their tone, making the experience more personal and natural.
During last week’s Alexa event, Amazon’s senior vice president of devices David Limp walked the audience through some of the upcoming features of Remarkable Alexa. He showed how the next-gen assistant will feel eerily close to talking with a real person, bantering back and forth, and even laughing or sounding sad. That’s largely due to the new Alexa Language Model (LLM).
A key aspect of the LLM is that it can recognize and interpret human emotion. It can sense if you’re happy or sad, and it will adjust its behavior accordingly. It can even make recommendations based on your mood, such as suggesting a good movie or music.
Another feature of the new Alexa is its improved synthesis capabilities. The system now uses a single natural language model to process interaction history and personalization signals, compared to the previous Alexa, which used multiple different models. It’s also more efficient, reducing memory usage by as much as 20x.
Another exciting new feature is the ability to send emails and calendar invites to Alexa. This will allow you to keep track of your family’s schedule without having to use your voice. And for those with vision impairments, Amazon is introducing an eye gaze mode that will allow them to control Alexa with just their gaze. This is set to debut this fall on Echo Show devices with screens and cameras, and later on Fire Max 11 tablets.
It’s too expensive
Amazon’s price tags are notoriously high, and many consumers feel that they can find better prices elsewhere. The company’s pricing strategy has been a huge factor in its success and growth, but it may also be a contributing factor to declining customer satisfaction levels. In a recent survey, 51% of respondents said that they are “never satisfied” with Amazon’s pricing. The company is well aware of this issue and has been working to improve the experience for its customers.
One example is the new Alexa LLM, a conversational AI digital assistant that uses voice and visual processing to communicate with users. The next generation of Alexa eliminates the need to constantly repeat the wake word and provides more natural conversations – similar to conversing with a friend. It can also identify emotions and maintain context throughout a conversation. In addition, it can automatically recognize when a user is talking to it and respond accordingly.
During last week’s hardware event, Amazon Senior Vice President David Limp demonstrated the new Alexa by talking to it about football and asking it to compose dinner invitations for his friends, while also requesting poems that it wrote on the spot. Backed by a large language model that is akin to technologies like OpenAI’s ChatGPT, the revamped Alexa sounds more human than its everyday counterpart. It can banter back and forth with the user, ask leading questions, sound happy or sad, and even laugh.
The upcoming version of Alexa will also be able to do more complex tasks, such as controlling smart home devices and scheduling appointments in a calendar. It will also be able to translate calls between two users of different languages in real time. It will be available on Echo devices with screens and video cameras later this year, and it will be a free update for existing Alexa customers.
Other features that are reportedly coming to the new Alexa include email to Alexa, which will allow users to send any type of document to their Alexa device and have it read aloud. Alexa will also be able to add events from an email to a user’s calendar. Another new feature is Eye Gaze Mode, which will allow customers with speech or mobility disabilities to control Alexa using their eyes.
It’s too complicated
Despite Alexa’s success as a consumer product, Amazon has struggled to turn it into a viable business. The company recently laid off employees and rolled out major cost-cutting measures. It’s now trying to revive the voice assistant with a new subscription service that will feature an upgraded AI technology. But the project is facing internal conflict, and could ultimately fail if it doesn’t succeed at the same level as the original Alexa.
The upcoming version of the virtual assistant is intended to be more conversational and personalized. It will be able to identify the user and adjust to their preferences, including their preferred voice, accent, and speaking rate. Additionally, it will be able to differentiate between multiple profiles, and will allow users to customize the assistant for specific devices. This means that users will be able to have their own personal Alexa, even when they use the device with someone else.
Amazon is working to create more natural conversations, but they also want the virtual assistant to be able to complete real-world tasks. To do this, the company is incorporating natural speech recognition and using large language models to understand nuances in context. For example, if you ask how your favorite team did in a game, Alexa will respond empathetically if they won, or provide more details if they lost.
In addition, the next-gen Alexa will be able to connect with Amazon’s line of “Amazon Basics” products. This will allow the company to better understand how customers are using their products, and make improvements based on that data. For instance, if you use the microwave a lot to pop popcorn, the Alexa app will notice this and automatically reorder your preferred brand of popcorn at a discount for you.
But this increased connectedness raises privacy concerns. With the reduced reliance on the wake word, it’s possible that the digital assistant will be listening to customers for extended periods of time without them knowing. This will require Amazon to address these concerns and build trust with consumers. Otherwise, users may not be willing to give up their privacy for the convenience of an Alexa-connected life.
It’s not fast enough
With Amazon poised to charge for a new version of Alexa, there’s internal tension over whether the feature will prove popular. “If people don’t pay for it, we don’t have a business model,” one person said. Despite its trailblazing status, Alexa has never been able to generate an ongoing revenue stream. In the latest quarter, Amazon’s devices division lost $3 billion. That’s double the losses of any other division and is leading to major layoffs.
Amazon execs showcased a revamped version of Alexa last week at its big hardware event. The new Alexa, which it calls ‘Remarkable,’ uses an upgraded speech-to-text LLM model and is designed to provide more natural conversations. It also has a more centralized architecture that allows for better reasoning over a larger set of context and interaction history, Amazon says. The new Alexa can also carry contextual clues across different parts of a conversation, and it will have a more engaging personality.
During the event, a demo showed Alexa gamely bantering back and forth with a human, asking leading questions, and even laughing or sounding cheerful or sad depending on the situation. It can also understand non-verbal cues like body language, making it more natural to interact with. And it can understand a wider range of vocabulary, making it easier to ask a question that would be difficult for Classic Alexa to answer, such as a request to set a timer or play music.
But despite all this, Alexa still struggles to match the performance of Google Assistant and Apple Siri in the voice assistant wars, with the former having more than 88 million users this year and the latter nearly as many. That’s in part due to the fact that they are both more accurate than Alexa and can handle a wide variety of requests.
During the event, Limp indicated that Echo owners will get a free preview of Remarkable Alexa this year. But it’s not clear how long that will last. If the company isn’t satisfied with the feature’s performance, it may delay a launch for a paid subscription service.
Leave a comment