Welcome to the AI Daily section! This is your daily guide to exploring the world of artificial intelligence. Each day, we bring you the hottest topics in the AI field, focusing on developers, helping you stay ahead of technological trends and understand innovative AI product applications.

Explore fresh AI products by clicking here: https://top.aibase.com/

1. Text-to-Image Revolution! Zhipu's Open-Source Image Generation Model CogView3 is 10 Times Faster than SDXL!

Zhipu AI's latest open-source models, CogView3 and CogView-3Plus-3B, have injected new vitality into the text-to-image field. CogView3 employs a unique cascaded diffusion method, surpassing the quality of SDXL with a tenfold increase in inference speed. CogView-3Plus-3B is based on the DiT framework, optimizing Zero-SNR diffusion noise scheduling and joint text-image attention mechanisms, supporting a wide range of image resolutions. Zhipu AI offers suggestions and tools to help users optimize prompts, lowering the barrier to entry.

image.png

AiBase Highlights:

🚀 CogView3 uses a cascaded diffusion method, generating higher quality images than SDXL at ten times the speed.

💡 CogView-3Plus-3B introduces Zero-SNR diffusion noise scheduling and joint text-image attention mechanisms based on the DiT framework.

🔧 Zhipu AI provides suggestions and example scripts to help users optimize prompts, making it easier to use.

Details: https://github.com/THUDM/CogView3

2. AI Voice Cloning Scams Escalate! A Phone Call Almost Cost a Lawyer's Father $30,000

Recent AI voice cloning scams have garnered widespread attention, highlighting the significant risks posed by this technology. The incidents reveal the vulnerability of humans when faced with highly realistic AI voice clones, calling for stronger regulation of the AI industry to protect consumers. The article also discusses the potential risks and positive uses of voice cloning technology, as well as recommendations for combating high-tech fraud.

image.png

AiBase Highlights:

🚨 High-tech fraud: AI voice cloning technology raises alarms, with scammers using AI to mimic voices for fraud, achieving alarming success rates.

💡 Security challenges: Humans have a high error rate in identifying AI-generated voices, urgently needing deepfake detection tools.

⚠️ Risk warning: New attack methods, such as audio hijacking combined with multiple AI technologies, could lead to funds being transferred to fake accounts.

3. Disrupting 3D Modeling! GGHead: Achieving Ultra-Fast and High-Quality, Consistent 3D Head Generation

GGHead is an innovative 3D modeling technology developed by Tobias Kirschstein et al. at the Technical University of Munich. It can generate high-quality, view-consistent 3D head models at ultra-fast speeds, lowering the barrier to 3D modeling and opening new possibilities for future human modeling research.

image.png

AiBase Highlights:

🌟 Generates high-resolution 3D head models in real-time using a standard GPU.

💡 Utilizes 3D Gaussian point cloud representation and 2D CNN to generate Gaussian attributes, improving modeling efficiency.

🔧 Introduces "total variation loss" technology to enhance geometric precision, ensuring image quality and consistency.

Details: https://tobias-kirschstein.github.io/gghead/

4. New Visual Recognition Technology: Revisit Anything

This article introduces an innovative visual location recognition technology called Revisit Anything, which can identify locations from input images. The technology combines SAM and DINO techniques to enhance image segment retrieval, achieving more accurate location re-identification. The research team tested multiple datasets and recommends users to start with the 17places dataset for a quick start. Through feature extraction, cluster center generation, and SegVLAD pipeline operation, users can obtain efficient recognition results. This research advances the field of visual location recognition.

image.png

AiBase Highlights:

🌟 Combines SAM and DINO technologies to introduce a novel visual location recognition method.

📊 Users can quickly get started and run experiments, obtaining efficient results through detailed steps and scripts.

🔍 Provides a new visual location recognition solution, leveraging modern deep learning models for image analysis, advancing the field.

Details: https://github.com/AnyLoc/Revisit-Anything

5. Hisense Launches AI TV E7N with Built-in AI Chip

Hisense's E7N Ultra-HD Mini LED TV features Hisense's self-developed Mini LED backlight control chip, with a brightness of up to 4000nits, precise backlight control, and clear, high-contrast images. The built-in XinCard AI chip optimizes image quality through deep learning, offering an excellent viewing experience. With powerful performance, it supports 4 HDMI 2.1 ports, making it ideal for gamers.

image.png

AiBase Highlights:

🌟 The E7N features Hisense's self-developed Mini LED backlight control chip, with each chip controlling 8 LEDs, achieving 4224 zones and a brightness of up to 4000nits.

🤖 Equipped with Hisense's first domestic high-end image quality chip, the XinCard AI chip, which optimizes color, contrast, and clarity through deep learning, delivering member-level image quality for various programs.

🎮 The E7N is the first to feature the MT9655 SOC chip, with performance improvements of 164% and 219%, supporting 4 HDMI 2.1 ports, significantly enhancing gaming experiences.

6. PearAI, Claiming to Be an Open-Source Version of Cursor, Recently Raised $3.5 Million and Was Accused of Plagiarism

PearAI, self-proclaimed as an open-source version of Cursor, recently received $500,000 in funding from YCombinator, sparking debates about innovation, borrowing, and commercialization. Founder Pan claims the product outperforms C0pilot, but it has been accused of plagiarism from other open-source projects. The controversy exposes the boundaries of innovation and borrowing in the open-source world, as well as the potential negative impacts of exaggerating a project's uniqueness in tech entrepreneurship.

image.png

AiBase Highlights:

🔍 PearAI received funding from YCombinator, sparking debates about innovation, borrowing, and commercialization.

💡 Pan claims the product outperforms C0pilot, but it has been accused of plagiarism from other open-source projects.

🌐 The open-source community faces challenges in maintaining innovation and showcasing project value.

Details: https://x.com/CodeFryingPan/status/1840464744626675719

7. California's AI Regulation Bill Vetoed: Newsom Says Good Intentions but Inappropriate Approach

California Governor Newsom vetoed the highly anticipated SB1047 bill, sparking widespread discussion in the tech and political spheres. Newsom believes the bill, though well-intentioned, failed to consider the specific application environments and risk levels of AI systems, not being the best way to protect the public. This decision highlights the complexity of AI regulation, challenging the balance between promoting innovation and ensuring safety.

image.png

AiBase Highlights:

🔍 SB1047 aimed to regulate the development of large AI models, requiring companies to implement safety protocols, but faced opposition from Silicon Valley.

🚫 Newsom vetoed the bill, believing it did not consider the specific application environments and risk levels of AI systems, not being the best way to protect the public.

⚖️ Newsom emphasizes that he has signed 17 AI-related bills and invites experts to help formulate feasible AI regulatory measures, highlighting the complexity of AI regulation.

8. Science Fiction Magazine Faces AI-Generated Work Surge

Against the backdrop of the rapid development of artificial intelligence technology in recent years, online science fiction and fantasy magazine Clarkesworld has faced unprecedented challenges. Editors feel helpless and frustrated, with the creative environment deteriorating, but the volunteer-built spam filters are only temporary measures, and long-term solutions still need to be explored.

image.png

AiBase Highlights:

🌊 Editors feel frustrated, with the magazine founder closing the submission channel as a last resort.

🛑 AI proliferation has worsened the creative environment, with volunteers helping to build spam filters.

🔧 Long-term solutions still need to be explored, with science fiction magazines facing the survival threat of AI-generated work surges.

9. Say Goodbye to Tedious Tasks! ScreenPipe Keeps Your Screen "Working Automatically" 24/7!

Tired of endless meeting notes and switching between different software? The open-source AI screen recording tool ScreenPipe can solve your troubles. It can continuously record screen activity and audio 24/7 and automatically complete work tasks, saving time and energy.

image.png

AiBase Highlights:

⏰ 24/7 continuous recording: Fully records screen activity and audio data.

🤖 Intelligent assistance services: Automatically updates notes, generates reports, fills data, etc.

🔍 Instant content overview: Quickly extracts key information, saves time, and focuses on the essentials.

Details: https://github.com/mediar-ai/screenpipe

10. Zhipu AI National Day Sale: 90% Off Usage + 100 Million Free Tokens

Zhipu AI and Qingyan Technology have separately launched National Day special activities and a new membership system, offering users more benefits and choices. The Zhipu AI open platform bigmodel.cn offers 90% off AI model usage and 100 million free tokens, reducing the cost for businesses and developers to use advanced AI services. Qingyan Technology has introduced a membership system, starting at 19 yuan/month, allowing a wider user base to afford advanced AI features.

image.png

AiBase Highlights:

🎉 Zhipu AI National Day sale, 90% off AI model usage, reducing costs for businesses and developers.

💰 The event offers 100 million free tokens, allowing more people to experience advanced AI technology.

🚀 Qingyan Technology launches a membership system starting at 19 yuan, enabling a wider user base to enjoy advanced AI features.