Amazon CEO Andy Jassy Can’t Stop Talking About Generative A.I.

"I could write an entire letter on LLMs and Generative AI as I think they will be that transformative," Andy Jassy wrote.

Andy Jassy in a dark navy suit.
Amazon CEO Andy Jassy. Thos Robinson/Getty Images for The New York Times

Under the leadership of its new CEO Andy Jassy, Amazon is finally joining the big tech arms race of generative artificial intelligence, the latest breakthrough in machine learning and popularized by the success of OpenAI’s ChatGPT. In his annual shareholder letter published yesterday (April 13), Jassy said Amazon has been developing its own large language models (LLMs)—the core technology behind text generators like ChatGPT—and promises continuing investment in the area despite the company facing ongoing pressure to cut costs elsewhere.

“We have been working on our own LLMs for a while now, believe it will transform and improve virtually every customer experience,” Jassy wrote.

Amazon has been using machine learning for more than 20 years in a wide array of products and services, Jassy noted in the letter, from recommending products on to training the Alexa voice assistant. But the onset of generative A.I. “promises to significantly accelerate machine learning adoption,” he wrote.

Unlike Google and Microsoft, which recently introduced consumer-facing A.I. chatbots, Amazon’s generative A.I. strategy will first focus on corporate customers through the Amazon Web Services, its most profitable division. And, in addition to developing its own model, Amazon will serve as a platform for other A.I. companies to run their models on.

Yesterday, Amazon introduced Bedrock, a cloud platform that allows corporate customers to build custom applications using pre-trained language models from top A.I. startups including AI21, Anthropic and Stability AI. Both AI21 and Anthropic make text generators that compete with ChatGPT, while Stability AI is the company behind Stable Diffusion, an image generator similar to OpenAI’s DALL-E.

Bedrock also provides access to Amazon’s own foundation model called Titan. Foundation model (FM) is another term describing an A.I. neural network trained on a large amount of unlabeled data.

“Most companies want to use these large language models, but the really good ones take billions of dollars to train and many years and most companies don’t want to go through that,” Jassy told CNBC in an interview yesterday. “So what they want to do is work off of a foundational model that’s big and great already and then have the ability to customize it for their own purposes. That’s what Bedrock is.”

Bedrock is currently in a limited preview mode and its first customers will include software company Pegasystems and consulting firms Deloitte and Accenture, Amazon said.

“I could write an entire letter on LLMs and generative A.I. as I think they will be that transformative,” Jassy wrote in his letter to shareholders. “But I’ll leave that for a future letter. Let’s just say that LLMs and generative A.I. are going to be a big deal for customers, our shareholders, and Amazon.” Amazon CEO Andy Jassy Can’t Stop Talking About Generative A.I.