Apple unveiled its proprietary generative AI model, "Apple Intelligence," at WWDC 2024, marking a significant push into AI technology. The model, which will be integrated across iOS, iPadOS, and macOS, can generate images based on users' photo libraries using styles such as Sketch, Illustration, and Animation. It can also execute tasks and actions based on natural commands and user context. Apple's emphasis on privacy ensures data security by using cloud-based models on special servers. Siri receives notable upgrades, including text interaction, contextual understanding, and the ability to integrate with OpenAI’s ChatGPT for enhanced responses. New AI-powered writing tools will also be introduced, providing features like email summarization and improved note-taking. These updates aim to provide a more personalized and secure user experience across Apple devices.

Apple announces 'Apple Intelligence', its own generative AI model - The Tech Portal The Tech Portal

At WWDC 2024, Apple introduced Apple Intelligence, a comprehensive initiative integrating generative AI across its ecosystem, including iOS, macOS, and visionOS. Built with user privacy and personalization at its core, Apple Intelligence focuses on understanding individual routines, relationships, and communications. It leverages large language models to process data locally on Apple silicon, with some tasks offloaded to Private Cloud Compute for enhanced privacy. The update significantly overhauls Siri, enabling text-based queries and deeper app integrations, and adds features like Smart Replies, customized emojis (Genmoji), and the Image Playground tool. Apple also announced a partnership with OpenAI to bring ChatGPT functionalities to Siri without requiring additional sign-ups or fees. The feature will be available on iPhone 15 Pro and M1 Mac and iPad devices later this year.

Apple Intelligence is the company's new generative AI offering | TechCrunch TechCrunch

Researchers highlight a looming challenge in the AI industry: the scarcity of human-written training data for machine learning models. Due to the exponential pace at which computing capabilities are growing—outstripping the availability of new textual data—AI tools from companies like OpenAI and Google risk a future where they can't improve their models effectively. This potential shortfall could force a shift towards using AI-generated data for training, which may degrade the quality of AI outputs. The debate continues about whether advancements in AI efficiency could mitigate this issue, preventing an "AI ouroboros" scenario where models recursively train on their own generated content.

AI Appears to Rapidly Be Approaching Brick Wall Where It Can't Get Smarter Futurism

Microsoft’s new AI-enabled Recall feature, which captures and stores periodic screenshots of a user's computer desktop, has sparked significant cybersecurity concerns. For instance, security researchers revealed that hackers could access this historical data without needing administrator privileges, effectively bypassing the tool’s sole protection mechanism. The vulnerability highlighted by Google Project Zero's James Forshaw demonstrates that control lists can be easily manipulated, leading to unrestrained access to users' data. The discovery has led to Microsoft announcing changes, including disabling Recall by default and promising additional security measures following intense backlash from the cybersecurity community.

Microsoft’s Recall Feature Is Even More Hackable Than You Thought wired.com