Study tests cocktail-party audio as reverse Turing test

A small study used 40 synthetic clips with background chatter to test whether speech models could serve as an audio CAPTCHA. Humans averaged 0.71 word accuracy and a 77% pass rate, versus 0.54 and 55% for instructed Gemini 3.8 Flash and 0.55 for Qwen3-ASR—too narrow a gap for a reliable test.

• Google opens Gemini and Search to third-party apps2026-10-28
• Chrome AI Autofill can now read Gmail to fill forms2026-10-28
• AI use doubles to 55% among rural Nebraskans, poll shows2026-10-28
• DeepMind unveils Genie 4 world model2026-10-27
• AI art is creeping into haunted attractions2026-10-27