As new forms of generative AI spread across the tech world, it’s becoming tough for Alexa users to accept rudimentary responses from Amazon’s voice assistant, compared with sophisticated text-based conversations with OpenAI’s ChatGPT.
In one simple but illuminating example from a couple months ago, tech journalist Ryan J. Morrison of Tech Monitor, suffering from a sore throat at the time, posted a TikTok of himself asking Alexa to provide a recipe to help him feel better.
“Sure,” Amazon’s voice assistant answered, giving a standard response to recipe inquiries. “I can search by dish name and ingredients. What would you like?”
“Something to make me feel better when I’m not well,” Morrison responded.
Alexa asked him again to provide a dish name and ingredients.
His eyes widened in frustration. “I’ve got a sore throat,” he said. “Something to help with that.”
Alexa gave the same response a third straight time.
Cut to ChatGPT. “I’m not feeling well,” Morrison typed. “Give me a recipe to make me feel better.”
OpenAI’s large language model not only expressed sympathy for his illness, but offered “a simple recipe for a comforting soup that might help you feel better,” listing the ingredients and cooking instructions for what appeared to be a very tasty chicken noodle soup.
Why can’t Alexa do the same? The short answer is, Amazon is working on it.
That’s what Amazon CEO Andy Jassy said, essentially, on the company’s conference call Thursday afternoon about its first-quarter earnings. Jassy was addressing a question from analyst Brian Nowak of Morgan Stanley about how the new era of AI is influencing Amazon’s investment priorities for the Alexa voice assistant and Echo smart speakers.
Jassy started by saying that Amazon’s vision isn’t just to build a smart speaker but to create “the world’s best personal assistant. He said the rise of large language models and generative AI “really accelerates the possibility” of doing that.
Amazon is starting from a “pretty good spot with Alexa,” he said, with an installed base of “a couple hundred million endpoints” for applications including entertainment, shopping, smart-home technologies, and information.
“We’ve had a large language model underneath it. But we’re building one that’s much larger, and much more generalized, and capable,” Jassy said. “I think that’s going to really rapidly accelerate our vision of becoming the world’s best personal assistant. I think there’s a significant business model underneath it.”
We asked Amazon to elaborate on Jassy’s comment, including how long this more advanced large language model has been in development, and when it will be released publicly. A spokesperson responded with this statement:
“Alexa already benefits from a large language model called the Alexa Teacher Model, which has been enhancing ML systems behind the service for the a number of years. As Andy said, we’re also building new models that are much larger and much more generalized and capable, which will take what already is the world’s best personal assistant and accelerate our path to being even more proactive and conversational.”
On the call, Jassy also reiterated Amazon’s commitment to generative AI in its AWS cloud unit, which he highlighted in his recent letter to shareholders.
The continued investment in Alexa is notable in part because Amazon’s Devices & Services division, which includes Echo and Alexa, was one part of the company impacted by Amazon’s initial wave of job cuts starting in January.
