Social RoboticsAmazon rebuilds Alexa around a large language model
Amazon said the new Alexa would hold more natural conversations, read non-verbal cues and drop the need to repeat the wake word.
At its annual devices event Amazon announced that Alexa would run on a new custom large language model optimised for voice. The company said Alexa would personalise responses, use Echo sensors to pick up non-verbal cues, and allow ongoing conversation without repeating its wake word, while admitting it would sometimes make mistakes. Amazon also described a planned speech-to-speech model with more expressive tones. The world's most widespread home voice assistant moving to an LLM raised expectations for every conversational device, robots included.
Source: Voicebot.ai
This archive entry summarises reporting published elsewhere; follow the source link for the full story. The summary was drafted with AI from that reporting.