Welcome to the [AI Daily] column! This is your daily guide to exploring the world of artificial intelligence, where we present the hot topics in the AI field every day, focusing on developers to help you understand technical trends and learn about innovative AI product applications.
Fresh AI products click to learn more: https://top.aibase.com/
1. Claude Introduces Projects Feature: Create Personalized Robots for Specific Projects
Claude's latest innovation, the "Projects" feature, offers users customized chatbot services with unique characteristics and advantages. With Projects, users can create personalized robots tailored to specific projects, build comprehensive knowledge bases, avoid cold start issues, customize instructions, and collaborate more closely with Claude.
【AiBase Summary:】
🚀 Projects feature provides enhanced knowledge bases, allowing users to upload documents, code, and insights to improve the effectiveness of robot responses.
🔥 Avoids cold start issues, solving the challenge of AI lacking context information and providing more professional support.
💡 Artifacts feature enhances collaboration with Claude, offering larger code windows and real-time preview functions to support developers' coding.
Product entry: https://top.aibase.com/tool/claude-3-5-sonnet
2. Figma Redesigns Interface and Introduces AI Design Tools: Initial Design Drafts in Seconds
Figma announced significant feature updates at the Config conference, including a new UI design, generative AI tools, and built-in slide functionality, aiming to provide smarter and more efficient design tools. The new design focuses on user work, meeting the needs of both new users and experts. The generative AI tool helps users quickly start design projects, speeding up the design process. The new slide functionality built into Figma makes it easy for users to create and present their design work.
【AiBase Summary:】
⭐ Figma introduces a new UI design focusing on user work, meeting the needs of both new users and experts.
⭐ Generative AI tools help users quickly start design projects, speeding up the design process.
⭐ New Figma slide functionality makes it easy for users to create and present their design work.
3. Viggle Introduces Move Feature: Preserves Original Backgrounds Without Extra Editing
Viggle recently introduced a new feature called "Move" that offers a simple and fun way to add animation effects to static photos. This innovative feature preserves the original background of the photo, is easy to access, and requires no complex editing, providing users with more vivid and interesting visual effects. Viggle has taken an important step in the field of image processing technology, providing powerful tools for users and content creators.
【AiBase Summary:】
✨ Preserves original backgrounds: The "Move" feature differs from previous functions limited to green and white backgrounds, preserving the original background without extra editing.
🎨 Easy to access: Users can simply visit https://viggle.ai to use the new feature.
⚙️ No complex editing: Directly upload photos and easily add animation effects without cumbersome post-processing.
Details link: https://viggle.ai
4. Alibaba's Qwen-2 Tops Global Open Source Model Rankings
Alibaba's latest open-source Qwen-2-72B instruction tuning version has topped the global open-source model rankings, demonstrating China's leadership in the open-source model field. The re-evaluated data from the rankings show that Qwen-2 has surpassed Meta's Llama-3 and France's Mistralai's Mixtral, attracting attention and discussion within the industry.
【AiBase Summary:】
⭐ Alibaba's Qwen-2-72B instruction tuning version tops the global open-source model rankings
⭐ China's open-source models occupy four spots in the top ten rankings, demonstrating a solid leading position
⭐ Showcases competitive strength comparable to internationally renowned closed-source model platforms
5. Zero-One Universe Launches 80% Off GPT Alternative Plan
I am impressed by Zero-One Universe's introduction of the "Yi API 80% Off GPT Alternative Plan." They offer a high-value alternative solution to help former OpenAI users migrate seamlessly, with performance comparable to GPT and a cost reduction of up to 91%. This initiative is significant for the AI industry.
【AiBase Summary:】
🚀 High-value alternative solution helps users seamlessly migrate to Yi series large models, with performance comparable to GPT and a 91% cost reduction.
💰 Offers a 100 yuan gift amount and a 50% top-up gift, with RPM/TPM speed limit directly upgraded to Tier3 service to ensure high-level service quality for users.
🌐 Zero-One Universe's Yi series large models perform excellently, surpassing GPT-4, with costs far lower than competitors, gaining recognition from multiple enterprises.
Details link: https://platform.lingyiwanwu.com/Yi-Large
6. Video Editing App Captions Introduces AI Editing Features
Captions, a video editing app, has recently introduced new AI editing features that can add special effects to existing unedited videos, offering more transition and effect options. Users can use AI characters to create videos in different styles without shooting the original video, saving time and effort. This feature may present discoverability issues for video creators but also provides higher quality services for sales, marketing, and communication channels.
【AiBase Summary:】
✨ Captions introduces AI editing features to add special effects to videos
🎥 Use AI characters to create videos in different styles
💡 AI tools may present discoverability issues
Details link: https://top.aibase.com/tool/captionsAI
7. AuraSR, an Open-Source Image High-Definition Restoration Model, Can Enlarge Images by 4x and Add Details
AuraSR is a giant upscaling model with 600 million parameters, developed from the GigaGAN paper, capable of enlarging images by four times and adding details, with fast processing speed and excellent results. As a super-resolution image enhancement model, AuraSR focuses on improving the resolution of generated images and is simple to use. It represents the forefront of artificial intelligence technology and promotes the democratization of AI development.
【AiBase Summary:】
✨ Can enlarge images by four times and add details, with fast processing speed and excellent results
⚡ Through multiple enlargements, the details become more abundant
🎨 Can handle realistic and non-realistic content, enhancing the overall visual experience
Details link: https://top.aibase.com/tool/aurasr
8. Hugging Face Updates Leaderboard Evaluation Rules, Entering a New Phase of AI Evaluation
Hugging Face's latest update to the Open LLM Leaderboard has a significant impact on AI development. The update provides a more rigorous and detailed evaluation, addressing the issue of stagnant performance improvements in large language models. The AI community realizes that evaluating models based solely on performance numbers is insufficient, and the update introduces more complex evaluation metrics.
【AiBase Summary:】
⭐ Open LLM Leaderboard update provides a more rigorous and detailed evaluation, addressing the issue of stagnant performance improvements in large language models.
⭐ The update includes introducing more challenging datasets, implementing multi-round dialogue evaluations, etc., aiming to create a more comprehensive and challenging benchmark.
⭐ LMSYS Chatbot Arena launched to complement Open LLM Leaderboard, emphasizing real-time, dynamic evaluation methods, bringing new ideas to AI evaluation.
9. What Was Once a Treasure Trove Is Now a Pitfall! Rabbit Exposed with Severe Code Vulnerabilities
From a first-person perspective, I comment on the situation where Rabbit has been exposed to have severe vulnerabilities. The article reveals the security risks and trust crisis Rabbit is facing, as well as the issues with its product R1 in actual performance.
【AiBase Summary:】
🔑 The discovery of Rabbitude reveals hard-coded API keys in the company's code, exposing sensitive information to risks.
🔒 Although Rabbit has taken measures to limit access, security concerns remain, and rebuilding public trust is difficult.
🔋 R1 has issues with battery life and rudimentary features, and software updates have not addressed the core issue of overpromising and underdelivering.
10. Solving the AI Video Forgery Problem? DIVID Detects AI-Generated Videos with 93.7% Accuracy
AI-generated videos are becoming increasingly realistic and difficult to distinguish from real ones. Researchers from Columbia University's School of Engineering have developed the DIVID tool, which can detect AI-generated videos with 93.7% accuracy. DIVID improves upon old methods, identifying videos generated by diffusion models, which gradually transform random noise into high-quality, realistic video images. The researchers extend Raidar's AI-generated text insights to video, using language models to rephrase or modify text or video, and judging authenticity by the number of edits.
【AiBase Summary:】
⚙️ DIVID tool can detect AI-generated videos with 93.7% accuracy
🔍 DIVID identifies videos generated by diffusion models, gradually transforming into high-quality, realistic images
🛠️ Researchers extend Raidar's AI-generated text insights to video, judging authenticity by the number of edits
Details link: https://arxiv.org/abs/2406.09601
