Apple's New Siri AI Improves in 10-Test Trial as Accuracy and Conversation Still Lag
Updated
Updated · ZDNet · Jun 17
Apple's New Siri AI Improves in 10-Test Trial as Accuracy and Conversation Still Lag
2 articles · Updated · ZDNet · Jun 17
Summary
A 10-round ZDNET test on a Mac found Apple’s new Siri AI more useful than the old Siri, but still prone to wrong answers and awkward back-and-forth exchanges.
On general questions, summaries and advice, Siri often produced concise, helpful responses and cited sources, though some tasks needed extra prompting before it would summarize or offer an opinion.
Accuracy broke down on retrieval and image tasks: Siri found only 3 of 6 matching Lincoln photos in a library and misidentified two paintings before correctly naming a Van Gogh work.
Conversation flow also remained clumsy, with the tester needing to click the microphone again after each reply instead of getting a fluid voice exchange.
The feature is currently gated behind a waitlist on supported Apple Intelligence devices running the version 27 developer beta, with Apple expected to refine it before a public release anticipated in September.