It’s no longer possible to completely avoid generative artificial intelligence when dealing with technology, but Apple has a little less of it. It’s not entirely by choice, though. The iPhone maker has repeatedly delayed AI-powered Siri after first promising it for 2024, but Agreement with Google will pair the iconic assistant with the Twins later this year. As WWDC approaches, Apple is working to bring its big AI brains into the humble processing environment of a smartphone. Apple fans may not like the result.
Apple has long been hyped about the value of privacy runs AI nativelybut a new report suggests that despite Apple’s best efforts, the iPhone’s Gemini makeup will rely more on Google and Nvidia in the cloud. Information reports That Apple’s Gemini-infused Siri will work both on-device and in the cloud is a clear reversal of its privacy-focused preference for native AI.
With every new chip announcement, we hear how silicon is being optimized for AI – even Apple is doing it with a focus on Neural Engine improvements. You might think from the fancy language that smartphones are equipped to handle powerful AI models, but this is not necessarily the case. In fact, GPUs in most phones can process more AI tokens than AI-focused NPUs. Components like the Apple Neural Engine are designed for contextual, efficient AI processing. Even if phones have faster AI, they don’t have the RAM to store huge models.
Even the greatest AI models are still mediocre assistants, making native AI very difficult. AI models running on phones are physically smaller and have at most a few billion parameters. Compare that to Google’s latest Gemini models, which have trillions of settings, The Information reports. On-device AI models are also “quantized” to operate with lower precision, making them faster but affecting the accuracy of token generation. All of this adds up to AIs that feel less intelligent than their cloud brethren, and even great cloud-based models can be pretty dumb at times.
Amazing, tapering Gemini
Google has mobile-optimized versions of Gemini Gemini Nano. However, these are intended to enhance contextual features such as Magic Cue and audio summaries. On the other hand, Siri should be a conversational assistant – you talk to her and she does everything. It’s a different experience that requires a different model. On Android, Google doesn’t even bother trying to do this natively. Talking to Gemini always goes directly to the cloud.
