That’s it, many months after the competition, Apple has entered the AI race. WWDC 2024 was an opportunity for the company to detail how this new tool works on iOS, iPadOS and macOS.
It was expected, but it remains a big step forward for Apple: generative artificial intelligence finally arrives at the heart of the firm’s operating systems. “ This is all about more than artificial intelligence, it’s personal intelligence, and it’s the next big step for Apple», Tim Cook immediately admitted on stage during WWDC 2024, even allowing himself to play with the acronym “AI” to make it become “ Apple Intelligence“.
After observing what was happening on the market with ChatGPT, Google Gemini and others, the late Steve Jobs’ firm detailed its own vision of AI, with some interesting biases. Little everything from the owner.
There is not one, but AI at Apple
Whether on iOS, iPadOS or macOS, artificial intelligence at Apple will take several forms. The most basic tasks will be handled by “small” language models (or SLM forSmall Language Model) which will run locally. The latter are numerous and each time specialized in a particular task. So, summarizing your notifications or analyzing items in your calendar will not require an Internet connection.
More complex tasks such as analyzing photos or writing text will be delegated to more complex models (LLM forLarge Language Model) which will run on the cloud. Apple promises that the choice of models will be made fluidly and intelligently according to the user’s requests.
ChatGPT as a model of excellence
In some cases, iPhones, iPads and Mac computers may use ChatGPT to analyze or generate certain content. These use cases will always be clearly labeled as such for issues of transparency in data management. Those who have a paid account or subscription will also be able to repatriate their history, if that is their wish.
By default, however, the IP addresses of machines accessing OpenAI’s servers will be masked and the data will not be retained by the company. The firm opted for this language model because it wanted “start with the best of all“, confessed Craig Federighi, head of software at Apple.
Google integration soon?
Even though ChatGPT was widely highlighted during Apple’s conference, the company is not holding back from integrating other AI models into its machines in the future. In an interview, Craig Federighi explains that Google Gemini could well carve out a small place for itself on the brand’s iPhone, Mac and iPad.
“We think people may have a preference for several models […] for example a model that is perfectly suited to writing and a model that is more suited to coding. So we want to allow users to choose the model that suits them best» detailed the manager during an interview.
Apple wants to put privacy at the center of its project
Even for requests processed in the cloud, Apple wants to ensure that its users’ data is secure. To do this, the company launched the “Private Cloud Compute” initiative. Behind this somewhat obscure name are in fact servers running on Apple Silicon chips (the same as in Macs) and capable of processing complex requests in complete security.
Apple promises that data shared with these servers “will never be stored or exploited» and that only those “useful for accomplishing a task» are sent to the company’s cloud.
A little taste of open Source
To convince the general public that Apple’s services are indeed secure and that the firm’s AI does not pose any risk to your personal data, the company announced that the code underlying its platform ” Cloud Compute” will be publicly available. So “independent experts will be able to inspect the code» to assess the merits of the promises made by Apple.
According to the company, the system was designed so that Apple devices “cannot connect to a server whose code has not been publicly audited“.
