Apple releases AI models that could show the future of artificial intelligence on iPhones

Apple researchers have released a host of new AI models that offer a hint at how artificial intelligence might work on the iPhone of the future.

The company has largely been quiet on its plans for AI, as much of the rest of the tech industry has pivoted towards offering new artificial intelligence features and products. Initially, it appeared resistant to using the phrase at all.

In the time since, however, chief executive Tim Cook has teased that Apple is working on new generative AI features. He said earlier this year that Apple was investing “a tremendous amount of time and effort”, though without saying what those are.

Some of those updates might be revealed at Apple’s upcoming Worldwide Developers Conference, in June. Apple’s marketing chief, Greg Joswiak, announced that event by saying it would be “Absolutely Incredible”, using capital letters in an apparent hint that AI features are coming.

Now, however, Apple’s release of what it called “OpenELM” might offer a hint at how those features could work. Apple says the name stands for “Open-source Efficient Language Models”, and that they are relatively small models specifically calibrated for certain tasks.

Apple’s developers said the release was made public in the name of “reproducibility and transparency”, which is “crucial for advancing open research, ensuring the trustworthiness of results, and enabling investigations into data and model biases, as well as potential risks”. It did not give any indication that they would be used in future products.

However, the fact that Apple is focusing on small and efficient models might be a hint at the way they will appear in the iPhone. Such small models are built specifically to run on consumer devices, rather than the large cloud services that are required to power some other, less efficient AI tools.

Apple appears to be looking to build AI models that can run on the iPhone itself, without sending data to its servers. That would be in keeping with Apple’s commitment to privacy, as well as helping to ensure that any AI-powered features run more quickly and do not rely on having a fast internet connection.

The company already looks to ensure that as much processing happens on its devices as possible. The new Apple Watch, for instance, includes processer updates that allow more of the AI transcription and understanding for Siri to happen on the device itself.

Apple’s new releases are language models, of the kind that power ChatGPT. It already uses such technology in its iPhones to improve the automatic suggestions in the keyboard – but it might use similar technology to help write emails or improve Siri, for instance.

Earlier this year, Apple researchers also published a new model that allows users to edit photos by writing messages, and another that can understand what is happening on a smartphone and help people navigate them.