Researchers from Apple are probing whether it’s possible to make use of artificial intelligence to detect when a user is chatting with a tool like an iPhone, thereby eliminating the technical need for a trigger phrase like “Siri,” in keeping with a paper published on Friday.
In a study, which was uploaded to Arxiv and has not been peer-reviewed, researchers trained a big language model using each speech captured by smartphones in addition to acoustic data from background noise to search for patterns that would indicate once they want help from the device. The model was inbuilt part with a version of OpenAI’s GPT-2, “because it is comparatively lightweight and may potentially run on devices comparable to smartphones,” the researchers wrote. The paper describes over 129 hours of knowledge and extra text data used to coach the model, but didn’t specify the source of the recordings that went into the training set. Six of the seven authors list their affiliation as Apple, and three of them work on the corporate’s Siri team in keeping with their LinkedIn profiles. (The seventh creator did work related to the paper during an Apple internship.)
The outcomes were promising, in keeping with the paper. The model was capable of make more accurate predictions than audio-only or text-only models, and improved further as the dimensions of the models grew larger. Beyond exploring the research query, it’s unclear if Apple plans to eliminate the “Hey Siri” trigger phrase.
Neither Apple, nor the paper’s researchers immediately returned requests for comment.
Currently, Siri functions by holding small amounts of audio and doesn’t begin recording or preparing to reply user prompts until it hears the trigger phrase. Eliminating that “Hey Siri” prompt could increase concerns about our devices “all the time listening”, said Jen King, a privacy and data policy fellow on the Stanford Institute for Human-Centered Artificial Intelligence.
The way in which Apple handles audio data has previously come under scrutiny by privacy advocates. In 2019, reporting from The Guardian revealed that Apple’s quality control contractors often heard private audio collected from iPhones while they worked with Siri data, including sensitive conversations between doctors and patients. Two years later, Apple responded with policy changes, including storing more data on devices and allowing users to opt-out of allowing their recordings for use to enhance Siri. A category motion suit was brought against the corporate in California in 2021 that alleged Siri is being turned on even when not activated.
The “Hey Siri” prompt can serve a crucial purpose for users, in keeping with King. The phrases provide a method to know when the device is listening, and eliminating which may mean more convenience, but less transparency from the device, King told MIT Technology Review. The research didn’t detail if the trigger phrase would get replaced by some other signal that the AI assistant is engaged.
“I’m skeptical that an organization should mandate that type of interaction,” King says.
The paper is one among quite a few recent signals that Apple, which is perceived to be lagging behind other tech giants like Amazon, Google, and Facebook in the substitute intelligence race, is planning to include more AI into its products. In response to news first reported by VentureBeat, Apple is constructing a generative AI model called MM1 that may work in text and pictures, which can be the corporate’s answer to Open AI’s ChatGPT and a number of other chatbots by leading tech giants. Meanwhile, Bloomberg reported that Apple is in talks with Google about using the corporate’s AI model Gemini in iPhones, and on Friday the Wall Street Journal reported that it had engaged in talks with Baidu about using that company’s AI products.