Apple Unveils Apple Intelligence: A Glimpse into Next-Gen AI Features in iOS 18.1
After over a month of anticipation, Apple has officially introduced its artificial intelligence (AI) system, Apple Intelligence, on its devices. On July 29, Eastern Time, the first version for iPhone AI was launched, now available in the iOS 18.1 developer beta, exclusively for registered developers. This update brings several new features, including enhancements to Siri, text processing tools, and photo management, though some functionalities like screen sensing and AI image generation are temporarily unavailable.
Apple plans to roll out additional features next year. Currently, the update supports only devices equipped with M-series and A17 Pro chips, meaning only the iPhone 15 Pro and iPhone 15 Pro Max are compatible, with similar requirements for iPads and Macs. Developers must set their devices to the U.S. region and use English as the primary language.
The initial version of Apple Intelligence includes an upgraded Siri, writing tools, email summarization, and photo search capabilities. Apple has also released a report detailing its proprietary large model technology, highlighting the AFM-on-device model with 3 billion parameters and the cloud-based AFM-server model, both of which outperformed GPT-4 in instruction handling and text summarization tests. Here’s an overview of the features:
1. Enhanced Siri Experience
The revamped Siri boasts a significant visual and performance upgrade. No longer confined to a spherical icon, the new Siri now features vibrant visual effects around the screen, enhancing the user experience. Users can access a text input interface by double-tapping the bottom of their iPhone screens for text-based interactions. The updated Siri understands fragmented commands and addresses device functionalities more efficiently, though it is yet to fully meet the high expectations set during its announcements.
2. Powerful Text Processing Tools
A standout feature of this update is the new text processing tools, compatible with virtually all native and third-party text input applications. Key functionalities include proofreading, rewriting, and summarizing, allowing users to easily check for spelling and grammatical errors, enhance written content, and efficiently summarize email communications while providing smart reply options. While these tools do not create entirely new text, they significantly boost user productivity in text management.
3. Photo and Call Recording Features
A new Focus mode filters out unimportant notifications to improve the user experience. The photo functionality allows users to create slideshows and search for images using natural language, along with video content support. Additionally, call recording is now live, letting users record calls with a simple button in the upper left corner of the screen, storing content directly in their notes. Note, however, that the summarization feature for recordings is not yet supported on devices sold in China.
Apple's foundational model training utilizes Google's TPU (Tensor Processing Unit) technology instead of conventional Nvidia GPUs. According to the technical report, the cloud system employs 8,192 TPUv4 chips, while the on-device setup uses 2,048 TPUv5p chips. Due to the strong demand for Nvidia GPUs causing supply shortages and price increases, tech companies like Apple are exploring alternative solutions. Designed specifically for machine learning tasks, Google's TPUs offer advantages in pricing and interconnectivity.
Compared to standalone Nvidia chips, Google TPUs operate on a cloud platform, allowing Apple to access computational resources without substantial hardware investments. Integrated into Google's infrastructure since 2015 and made available to external developers in 2017, TPU pricing remains competitive.
Analysts suggest that Google TPUs, with their high interconnectivity and competitive pricing, may serve as a viable alternative to Nvidia GPUs. The release of the iOS 18.1 Beta has allowed registered developers to explore some functionalities of Apple Intelligence. Many developers have shared their experiences on social media, praising Apple Intelligence's performance in writing, conversation, and image search applications.
Experts emphasize that the ability of Apple Intelligence to reshape the industry lies in delivering truly personalized AI, integrating device information with services to provide meaningful answers to users. As users evaluate the strengths and weaknesses of various AI products, they should remain patient to find the AI solutions that best fit their needs.
Ranking
