Products You May Like
Any company that is even tangentially connected to technology has an AI strategy right now. They’re not all upfront about it, though. Ever since ChatGPT kicked off the new AI race, Apple has been biding its time. A new report claims that Cupertino is set to reveal its AI strategy, and it’ll be all about on-device processing.
That would be a stark departure from how Google, Microsoft, and others run large language models (LLMs) in the cloud. Local AI could be a boon for privacy, but dropping the cloud has its own thorny issues.
This is not the first time we’ve heard that Apple was working on a large language model for its phones. However, Bloomberg’s Mark Gurman now says that Apple will launch its AI model without a cloud component, which is how models like Google Gemini and ChatGPT operate. For most companies, the cloud is just where the AI lives.
Today’s latest generative AI systems require substantial computational power to tokenize and process inputs to generate new content on the other end. When you access ChatGPT or Gemini, your data is sent to a massive data center with row after row of expensive AI accelerator cards. Even with all that power, it takes a few seconds for the chatbot to generate a reply.
Apple has created some of the fastest Arm chips in the mobile space, but that’s still a far cry from the power of the cloud. Samsung recently released its Galaxy AI features for smartphones, almost all requiring an internet connection. You can force some Galaxy AI features to run locally, but they take longer, and the results will be lower quality even on the AI-optimized Snapdragon 8 Gen 3. Apple may be in a similar place with its local AI, but the company has released some research to suggest it’s made progress in adapting LLMs to run more efficiently on smartphones.
OpenAI uses data centers filled with AI accelerators to run ChatGPT. An iPhone is somewhat smaller.
Credit: OpenAI
Apple will reportedly focus on specific use cases where it believes generative AI can help users get things done. Still, users may find that Apple’s LLM doesn’t do as much as the systems they’ve used over the past few years. The upshot is that your data is secure on your device rather than bouncing around in the cloud. Apple likes to draw a distinction between it and other big tech firms when it comes to privacy, and this would be another opportunity to trot out that old chestnut.
There may still be hope for a robust cloud AI on the iPhone. A previous report claimed that Apple was in talks with Google to integrate Gemini into iOS 18. This could serve as a general AI option for less sensitive data. This is all speculation until Apple spills the beans, which we expect to happen in June at the annual WWDC event.