Apple in iOS 18 plans to use language models to make Siri smarter, according to The Information. Apple wants Siri to be able to help users automate complex, multi-step tasks with voice-based commands.
As an example, Siri might be able to do something like take a series of five photos, turn them into a GIF, and send the resulting GIF to a friend, all as one single action. Automatically converting photos into a GIF is possible right now with the Shortcuts app, but it has to be set up manually.
Siri's functionality will be similar to what's possible with the Shortcuts app, and it is likely that it will see Siri integrating more deeply with Shortcuts. Shortcuts have become a key part of the iPhone operating system and can be used for some powerful automated actions. Apple maintains a gallery of Shortcut options for everything from removing backgrounds from images to creating chore lists.
The Information says that Apple is aiming to release the new Siri capabilities as part of the iPhone's operating system next year, which suggests that it will be part of iOS 18. Apple competitor Google is also working to upgrade Google Assistant with language model software able to handle more complex tasks, so Apple may want to beat Google to the punch or at least launch similar functionality around the same time.
Apple's Siri personal assistant has long been criticized for being notably behind voice assistants from other companies like Google and Alexa, primarily due to Apple's focus on security and privacy. Many Siri tasks are performed on-device rather than through a cloud server, so Apple will need to find a balance between privacy and functionality with the adoption of language models for Siri.
The iOS 17 update that is set to see a public release in just a couple of weeks has very little in the way of new Siri functionality, but there is key functionality that would be a precursor to multi-step tasks. Siri in iOS 17 able to recognize and respond to multiple back-to-back requests without needing to be reactivated, something not possible in iOS 16. You can, for example, ask Siri the time and follow that up with a request to text your friend that you're going to be late without two separate requests.
With iOS 17, Apple is also eliminating the "Hey" wake word, so you are now able to just say "Siri" to activate the personal assistant. Eliminating "Hey" allows for more natural Siri interactions.
Apple follows a strict schedule for its software releases, so we can expect to see iOS 18's feature set unveiled at the 2024 Worldwide Developers Conference.
Top Rated Comments
It's about time Siri is finally getting major improvements. In addition, I have a feeling Apple is going to take Siri seriously due to VisionOS for Apple Vision Pro. It will be heavily integrated.
I think Apple's problem with a consumer-facing LLM is control. Executive management is terrified of the possibility that it could say (or do) something not explicitly approved by The Company. That's why Siri is stuck in 2011, with capabilities and all potential responses literally entered by hand by a small team. It's so they always have 100% control over the image, or "personality" (bland) projected by Siri.
It will be interesting to see them come to terms with the need for a much more powerful assistant, which due to its necessary LLM underpinnings, may not always toe the company line.
Personally, I think their fear is mostly unfounded. Apple's LLM will say some weird or unexpected things which will go viral. People will have a laugh about it and then largely forget it. Just like when Bing's ChatGPT integration could be nudged to start sounding really insane early on. We all laughed and they fixed it. Big whoop. Everyone knows Siri is already the biggest joke in personal assistants, it can't be any worse than it already is from a public image perspective!