Apple's AI Strategy: Data Protection Also Causes Delays
Even after the cooperation with Google for Gemini-based language models began, little has changed in Apple's strategy. The directive comes from the very top.
Logo and icon of Apple Intelligence: Apple has been working on it for years.
(Image: Apple)
There was already talk of Apple's Gordian AI knot bursting – thanks to the comprehensive cooperation with Google as a supplier of Gemini for new foundational models at the iPhone manufacturer. However, internally, the cautious stance on artificial intelligence has apparently not changed. At an all-hands meeting of the group last week, more and more details of which are leaking out, this was made clear by the top management. This also apparently means that the new Siri, overdue since 2024, is being delayed once again.
New Class of Products and Services
Initially, Apple will continue to rely on its data centers based on its Apple Silicon chips, which are “tailor-made for our devices,” as CEO Tim Cook stated at the meeting. With this, he hopes, they will enable “a completely new class of products and services.” It is well known that Apple has been working on AI chips for servers for a long time and is partly already using them.
Videos by heise
However, there is a problem with the switch to Gemini: The search giant's large models are optimized for its Tensor Processing Units (TPUs), which also run in Google's own data centers. Based on these, according to at least credible reports, iOS 27 will work with a real Siri chatbot.
How data protection-friendly will Gemini usage be?
The long-awaited context-sensitive Siri is still being held back by technical hurdles stemming from the high degree of data protection. Software chief Craig Federighi, who was long considered an AI skeptic, said at the same all-hands meeting that it is “super important that the request to a model remains private.” The industry standard, he continued, is still to send data to a server where it is logged, available to the provider, and also used for training.
Apple, on the other hand, aims for a “personal and super powerful user experience” where this is not the case. How Apple intends to implement this remains unclear. Initial approaches like Private Cloud Compute (PCC) are considered a good path, but these in turn clash with Gemini usage. However, Google is also working on such data protection-friendly AI cloud queries. Apple is likely to be among the first customers.
Empfohlener redaktioneller Inhalt
Mit Ihrer Zustimmung wird hier ein externer Preisvergleich (heise Preisvergleich) geladen.
Ich bin damit einverstanden, dass mir externe Inhalte angezeigt werden. Damit können personenbezogene Daten an Drittplattformen (heise Preisvergleich) übermittelt werden. Mehr dazu in unserer Datenschutzerklärung.
(bsc)
