Amazon is developing an improved LLM to power Alexa

Amazon is building a more “generalized and capable” large language model (LLM) to power Alexa, Amazon CEO Andy Jassy said yesterday during the company’s first-quarter earnings call. An LLM, like ChatGPT, is a deep learning algorithm that can recognize, summarize, and generate text and other knowledge-based content from huge amounts of text data.

Jassy said that while Amazon has had an LLM that powers Alexa, the tech giant is working on one that is more capable than the current one. The Amazon executive believes the addition of an enhanced LLM will help Amazon work toward its goal of building “the world’s best personal assistant,” but acknowledged that it will be difficult to do in many domains.

“I think when people often ask us about Alexa, what we often share is that if we just made a smart speaker, it would be a much smaller investment,” Jassy said on the call. “But we have a vision, that we have the conviction that we want to build the best personal assistant in the world. And doing it is difficult. It is in many domains and it is a very wide area. However, if you think about the advent of big language models and generative AI, it makes the underlying models much more effective, so I think it really accelerates the possibility of building the best personal assistant in the world.

Jassy said he thinks Amazon has a good starting point with Alexa, as it has “a couple hundred million endpoints” that are used in entertainment, shopping and smart homes. He also noted that there is a lot of involvement from third-party ecosystem partners.

“We’ve had a great language model underneath, but we’re building one that’s much bigger and much more pervasive and capable,” Jassy said. “And I think that will very quickly accelerate our vision of becoming the best personal assistant in the world. I think there’s a significant business model under it.”

On the call, Jassy emphasized that Amazon has invested in AI and LLM for years, and that while it has the ability to invest heavily in building LLM, small businesses don’t, which is why the company launched Bedrock early of this month Bedrock offers a way to build generative AI-based applications using pre-trained models from startups such as AI21 Labs, Anthropic, and Stability AI. Available in “limited preview,” Bedrock also provides access to Titan FM (base models), a family of models trained internally by AWS.

Since its launch last year, ChatGPT has taken over the internet and became more and more popular. With all the hype surrounding ChatGPT, it’s no surprise that major tech companies are looking to incorporate LLM-based enhancements into their own offerings to keep up with the fast-paced AI space. For example, The Information reported yesterday that Apple is developing LLM-based enhancements for Siri. It’s worth nothing that Google will probably do something similar for Assistant.

Amazon wasn’t the only company to bring up AI during its quarterly call with investors, as Alphabet, Microsoft and Meta also emphasized their investments in big language models. Alphabet CEO Sundar Pichai said Google would continue to incorporate AI to advance search, while Microsoft CEO Satya Nadella said the company would continue to invest in AI, noting that Microsoft already has seen an increase in Bing usage after the search engine was updated with a ChatGPT integration. Additionally, Meta CEO Mark Zuckerberg said the company will invest in AI and introduce new AI-related updates to its apps.

Amazon reported first-quarter earnings that beat expectations and initially sent shares higher, but later reversed course after executives expressed concern about continued weakness in cloud growth. Revenue for the quarter rose 9.4% to $127.4 billion, while operating income came in at $4.8 billion.