Welcome to the AI Daily section! This is your daily guide to exploring the world of artificial intelligence. Each day, we bring you the latest in AI, focusing on developers and helping you understand technology trends and innovative AI product applications.
Discover fresh AI products by clicking here: https://top.aibase.com/
1. KREA AI introduces video extension feature, adding magical effects to real videos
KREA AI's latest Video Extend feature has sparked innovation in the video creation field by skillfully merging real videos with AI-generated content, offering creators an unprecedented video special effects experience. Its core highlight is the unique mechanism of utilizing the last frame of the video to extend and generate up to 5 seconds of continuous video content, achieving seamless visual transitions. The technology is fully integrated, supporting full model compatibility, and is easy to operate, even for beginners in video production.
AiBase Summary:
✨ Innovation: Video Extend feature skillfully merges real videos with AI-generated content, offering an unprecedented video special effects experience.
🌟 Visual Effects: Unique mechanism extends continuous video content, achieving seamless visual transitions.
💡 Technology Integration: Successfully integrates with major mainstream AI video model APIs, supporting full model compatibility.
2. Midjourney releases significant update, introducing a new external image editor and image re-texturing mode
Midjourney recently released a major update, introducing a new external image editor and image re-texturing mode, further enhancing the flexibility and detail of image creation. The update includes an external image editor and an image re-texturing mode, allowing users to directly edit images on the platform, improving lighting and texture effects, making images more vivid.
AiBase Summary:
🎨 External Image Editor brings creative freedom, allowing users to directly modify images on the platform without external software.
🌟 Image Re-texturing Mode optimizes details and texture, intelligently recognizing scene shapes and redefining lighting, material, and surfaces.
🔍 V2AI audit system comprehensively checks content security, real-time analysis of user input to ensure compliant content generation.
3. Apple to launch private AI cloud service, offering $1 million bounty for AI cloud security vulnerabilities!
Apple is set to launch a private AI cloud service, offering a $1 million bounty for potential vulnerabilities that could harm its cloud service security. This move will further enhance the security of Apple services and provide an opportunity for security researchers to showcase their skills.
AiBase Summary:
💰 Apple offers a $1 million bounty for private AI cloud service security vulnerabilities.
🔒 Apple's Bug Bounty program encourages private reporting of security issues, enhancing the security of customer devices and accounts.
📱 Apple launches a researcher-exclusive iPhone for more effective security testing and vulnerability discovery.
4. Meitu's WonderModel image generation capabilities upgraded again: generating more delicate and natural textures
Meitu Inc. announced that its WonderModel has achieved another upgrade in image generation capabilities, further enhancing its comprehensive strength. At the same time, it launched a one-stop AI short film creation tool, MOKI, providing users with a new visual experience. This upgrade specifically strengthens image generation capabilities, achieving precise visual expression and a story-like atmosphere presentation.
AiBase Summary:
🚀 WonderModel achieves another upgrade in image generation capabilities, enhancing comprehensive strength.
💡 Launches AI short film creation tool MOKI, gradually covering Meitu's product ecosystem.
🎨 The upgrade specifically strengthens image generation capabilities, integrating diverse aesthetic concepts, showcasing film-level visuals.
5. OpenAI macOS app receives significant update: voice interface finally supports image upload!
OpenAI recently made an important update to its macOS app's advanced voice mode interface, introducing a new image upload feature, allowing users to upload and discuss images through the voice UI, enhancing interaction experience. In addition to the image upload feature, users can also directly use the laptop camera to take photos and share them, but there is no video sharing feature yet. Future prospects point to the upcoming launch of the Canvas editor, and the full release of the voice mode may be delayed.
AiBase Summary:
🌟 New image upload feature: Users can now upload and discuss images through the voice UI, enhancing interaction experience.
📸 Direct photo sharing: Users can use the laptop camera to take photos, but there is no video sharing feature yet.
🔍 Future prospects: Canvas editor is即将推出, full release of voice mode may be delayed.
6. Xpeng AI Tianqi 5.4.0 global premiere, P7+全系标配高阶智驾
Xpeng Motors held an AI intelligent driving technology sharing session in Guangzhou, announcing that P7+ and subsequent models will be equipped with advanced AI intelligent driving as standard, without the need for optional, subscription, or paid services. The company emphasized that the cloud-based large model is the key to winning the competition in intelligent driving, building a powerful cloud-based large model using the same approach as OpenAI. The AI Tianqi 5.4.0 version brings multiple upgrades, including the AI Eagle Eye visual solution, parking capability improvement, and the Space-Time Light Shadow display system.
AiBase Summary:
🚗 Xpeng P7+ and subsequent models will be equipped with advanced AI intelligent driving as standard, without the need for optional, subscription, or paid services.
🔑 The cloud-based large model is the key to winning the competition in intelligent driving, Xpeng using the same approach as OpenAI to build a powerful cloud-based large model.
🔮 AI Tianqi 5.4.0 version brings multiple upgrades, including the AI Eagle Eye visual solution, parking capability improvement, and the Space-Time Light Shadow display system.
7. Meta AI's new quantized version Llama 3.2: speed doubled, runs on phones
Meta AI's new quantized Llama 3.2 model has made significant improvements in size and computational resource requirements, increasing the speed of model operation, suitable for various devices and real-time application scenarios. This technical progress is of great significance for promoting the sustainable development and popularization of artificial intelligence.
AiBase Summary:
🌟 Quantized Llama 3.2 model includes 1B and 3B versions, size reduced by 56%, computational resource requirements lowered.
⚡️ Model inference speed increased by 2-4 times, suitable for consumer-grade hardware, ideal for real-time applications.
🌍 Quantized Llama 3.2 performs similarly to the original version in natural language processing, helping businesses and researchers achieve AI applications.
Details link: https://www.llama.com/
8. Apple releases iOS18.2 developer beta, adding Siri integration with ChatGPT
Apple has released the first developer beta of iOS18.2, introducing integration with ChatGPT, allowing Siri to answer detailed questions about screen content. Users can ask Siri about the content in videos or photos, and Siri will take a screenshot of the screen and upload it to ChatGPT to get the answer. iOS18.2 also brings Image Playground, Genmoji, and a redesigned Mail application.
AiBase Summary:
📱 Siri integrated with ChatGPT, users can ask about screen content details.
🔍 Siri takes screenshots and uploads to ChatGPT to get answers, ensuring privacy permissions.
🚀 iOS18.2 new features include Image Playground, Genmoji, and Mail application redesign.
9. Farewell to the "black box"! Peking University develops new AI framework FakeShield, making image forgery无处遁形!
With the rapid development of AIGC technology, image editing tools have become more powerful but also easier to tamper with and harder to detect. The research team at Peking University has proposed an interpretable IFDL task, designed the FakeShield framework, and implemented image authenticity assessment and tampered area mask generation through a multi-modal large language model, addressing the shortcomings of traditional IFDL methods. FakeShield has strong generalization capabilities, can detect and locate various tampering techniques, and provides an interpretable solution, which is of great significance to digital content manipulation, generative artificial intelligence, and other fields.
AiBase Summary:
🔍 Interpretable IFDL task and FakeShield framework address the shortcomings of traditional methods, providing an interpretable tampering detection and localization solution.
🛡️ FakeShield uses a multi-modal large language model to assess image authenticity and generate tampered area masks, with strong generalization capabilities.
💡 FakeShield becomes a versatile practical tool, suitable for various real-world applications, helping to improve regulations, guide generative artificial intelligence development, and enhance the reliability of the online environment.
Details link: https://zhipeixu.github.io/projects/FakeShield/
10. Another OpenAI executive departs! 6-year security advisor and AGI team leader to leave
