Innovating Alexa amid the Rise of Large Language Models
Niklas Strüver · transcript Verlag eBooks · 2024
For about a decade, Amazon's Alexa was a pioneer in automatic speech processing; now, however, new Large Language Models (LLMs) are posing challenges for Amazon.One attempt to confront these challenges is by integrating technologies developed for Alexa by university research teams in the Alexa Prize Competitions (APCs).This chapter examines how participants in these contests deal with the conditions set and the resources provided by Amazon for the competition, and offers a snapshot of the practical development processes of the voice assistant at a time of technological transition.It then outlines some of the path dependencies, risks, benefits, and aspects of structuration that are encountered by the participants in their attempts to innovate Alexa.Technically speaking, Alexa is the voice interface for Amazon's cloud products Alexa Voice Service (AVS) and Amazon Web Services (AWS), where all requests are processed by various machine learning algorithms (Crawford and Joler 2018), which are constantly optimized based on the incoming usage data.This service is embedded in the Echo devices produced by Amazon.