Interrupting is the future
summary
Financial Times Labs turns articles into voice with experimental Amy using Amazon Polly, runs an A/B test of full-article narration, and publishes a Google Home game called Maker Connections. Main claim: making voice truly conversational requires interruptions—devices must accept midstream utterances (no wake word) and return precise interrupt timing so the system can resolve context. Found problems: TTS suits sentences not 15‑minute reads, messy editorial formatting and narrator‑ready data, and brittle name bugs like Michel Barnier crashing the app.
AI-written summary · about
chapters (9)
- 0:00Introduction and background
- 1:45FT Labs and audio strategy
- 3:10Early audio experiments
- 5:19Voice device experiments
- 7:02Key challenges in audio
- 9:26Problems with voice and text
- 12:00The case for interruptions
- 16:38Data and GDPR implications
- 18:07Summary and call to action
related talks
16mWeak signals of a vocal revolution in UIs
19mAlexa, show me a good voice UI
13mVOICE Summit 2019 | VED Talks | Dylan Zwick | July 23
33mStrategies for Conversational Product Experiences
36mEmbodied AI: A Cognitive Shift, Not a Cosmetic Change
44mDelivering On The Promise Of Voice With Alexa
50mDesigning Experiences For Voice Interactions
20mVoice Interfaces in Everyday Life