Will AI mean the end for "Hey Siri?" - Interesting Engineering (2024)

Researchers affiliated with Apple have been exploring doing away with the “Hey Siri” phrase to wake the virtual assistant, using artificial intelligence instead. Can this mean a seamless, person-like future for virtual assistants?

Virtual assistants such as Alexa and Siri depend on triggers to wake up and listen to the user’s prompts. Currently, these triggers are in the form of actions, such as pressing a button or swiping across the screen, or using keyphrases such as “Alexa,” or “Hey Siri.” These assistants listen for these phrases and only begin recording when triggered.

Featured Video

The big idea

In a study published in the pre-print archive,Arxiv, the team reveals three stages to their research. The first involved training classifiers using only acoustic information obtained from the audio waveform. The second involved taking the decoder outputs of an automatic speech recognition (ASR) system and passing them as input features to a large language model (LLM). The final stage involved a multimodal system combining acoustic and lexical features and the ASR decoder signals passed to the LLM.

According toMIT Technology Review, “six of the seven authors list their affiliation as Apple,” of which three “work on the company’s Siri team according to their LinkedIn profiles.”

Despite promising results, and predictions improving as the size of the model grew, it remains unknown if Apple plans to do away with “Hey Siri.”

Looming concerns

While an always-listening Siri might imply a smooth and seamless experience on paper, the larger looming question remains regarding privacy. Right now, Siri only starts listening after it hears “Hey Siri,” which helps alleviate concerns about constant surveillance. But if Apple ditches this wake word, this could mean our devices constantly listen in on our conversations.

Given the criticism the likes of Amazon and Apple have faced in the past over how it handles audio data, privacy and trust are certainly a big issue here.

Speaking toMIT Technology Review, Jen King, a privacy and data policy fellow at the Stanford Institute for Human-Centered Artificial Intelligence, expressed concern that getting rid of the phrase could imply less transparency from the device.

“I’m skeptical that a company should mandate that form of interaction,” King added.

Rocky times ahead

While Apple enjoyed large periods of innovation and grand success in the past, the company has recently found itself in the limelight for the wrong reasons. Despite its reluctance, the company was forced to change tact and replace lightning ports with USB-C on its iPhones.

Earlier this year, the company met further criticism for its malicious compliance with the European Union’s Digital Markets Act. While the law intended the likes of Apple to open its platform and allow alternative app stores on the platform, the Cupertino-based company decided to slap any blooming alternative app stores withhidden fees. These fees take back most of the savings for popular apps, sending profits to Apple’s treasury instead.

Although Apple teased users with a glimpse of itsgenerative AI research, news emerged that the company was working withGoogleandBaiduto provide cloud-based services to power AI features for its upcoming devices.

To add to the fire, the US Department of Justice (DOJ) announced a massiveantitrust lawsuitagainst Apple, accusing the company of building an illegal monopoly in the smartphone market.

To top the company’s woes, researchers found acritical security weaknessin Apple’s custom M chips, bringing the company’s big talk on security to question. The vulnerability, dubbed GoFetch, allows hackers to acquire cryptographic keys and decrypt data. According to the researchers, the vulnerability cannot be fixed with a software patch.