Nedim Fresko of Amazon presented a bracingly simple picture of the company’s consumer-facing AI strategy: one agent to rule them all, and that agent will be Alexa.
The voice-activated speaker system, and all the tech behind it, are a potentially formidable leg up for Amazon when it comes to consumer AI. Fresko said the company is now integrating its own large language model into the Alexa system, allowing longer, unbounded conversations, more personalization, better speech recognition and other new features. The general idea is “ambient computing,” where the machines do things without people even noticing, and with hundreds of millions of Alexa devices already in use, it’s possible to see that future.
Fresko was dismissive of the idea that people might ultimately have multiple intelligent agents performing specific tasks in the real world. Consumers, he insisted, would not be served by having to switch from one agent to another.
“We’re after a single, all-purpose AI that’s personalized to you,” he said. “We’ll figure it out behind the scenes.”
He did allow, though, that “latency constraints are punishing,” leading the company to deploy “edge models” for simple tasks.
Add comment