Mixer, Agitator mill, V-type mixer manufacturer
It is a professional manufacturer of mixers,producers of a series of products such as V-type mixers
PRODUCT LIST
Is the intelligent assistant Siri too easily awakened? Apple is considering using AI for improvement
Is the intelligent assistant Siri too easily awakened? Apple is considering using AI for improvement
News Source:  | Release Time: 2026-01-22

  Tencent Technology News, April 17th.According to foreign media reports, this year, problems caused by Apple's intelligent assistant Siri have been continuous.It has always held a dominant position in the media.

  Apple explained that iOS devices are now constantly listening for prompts and pointed out that advancements in artificial intelligence (AI) can eliminate the need for initial training for Siri. In "Personalized Siri", the Siri team pointed out that the biggest challenge faced by this feature is accidental activation and accidental missed activation. Currently, Apple is attempting to prevent these issues by using the voice of the device owner to set up Siri, requiring users to conduct brief training on Siri and create user profile features with five types of conversations. Then Siri quietly added the user's 35 "accepted" remarks to help improve the profile.

  Interestingly, the trigger point for "HeySiri" is that this profile does not merely attempt to match a single voice output, but ultimately includes a baseline version of this phrase, with 40 variants acceptable. Siri will only be activated when the given "HeySiri" message is equal to or higher than the average score. The device also stores 40 "HeySiri" request records, allowing users' personal profiles to be rebuilt without being retrained by the users, while Apple improves the system through software updates.

  In the future, Apple hopes that Siri will not have to undergo pre-training. Users' profiles will be cleared, and then with users' permission, as more requests are added, Siri can grow and update itself. The company is also studying how to block "false acceptance", that is, Siri being awakened by something other than the user. By using deep neural networks, Apple believes it can reduce the "false acceptance" rate by 50%, lower the false rejection rate (when the device does not respond when the user says "HeySiri") by about 40%, and decrease the chance of Siri being activated by others by 75%.

  Another issue is to make Siri perform better in larger and noisier environments, such as cars or windy outdoor conditions. Apple's team said that they are currently studying how to handle environments that are seriously incompatible with the existing "HeySiri" recordings in users' profiles, but have achieved success in what is called "multi-style training". In this kind of training, a subset of the training data is enhanced by different types of noise and reverberation.

  Because Apple is constantly improving Siri and may not necessarily change all the adjustments, it is unclear when the improvements mentioned above will be implemented. At the Worldwide Developers Conference in June, Apple frequently emphasized the significant changes to Siri. Given the recent discussions about Siri's shortcomings, we expect to hear more important news about Siri in the near future.