Welcome to the [AI Daily] column! Here is your daily guide to exploring the world of artificial intelligence, where we present the hottest content in the AI field every day, focusing on developers to help you understand technical trends and innovative AI product applications.
Fresh AI products click to learn: https://top.aibase.com/
1. Former Baidu executive Jing Kui's AI search venture valued at $1.8 billion, launches first product Genspark
As a former Baidu executive, Jing Kui founded the new company MainFunc and launched the first product Genspark, aiming to provide high-quality search experiences through AI technology. The company received a $60 million seed round of financing, with a valuation reaching $260 million, showing huge market potential. Jing Kui's outstanding resume now leads the AI search field again, expecting continued innovation and development.
【AiBase Summary:】
🚀 MainFunc is an AI innovation product company founded by Jing Kui and former Xiaodu CTO Zhu Kaihua, launching the first AI Agent search product Genspark.
💰 The company completed a $60 million seed round of financing, with a valuation reaching $260 million, showing huge market potential.
🔍 Genspark is positioned as an AI Agent engine, focusing on providing search services, generating custom pages called "Sparkpages" through AI technology, saving users time and providing reliable information.
Official website: https://mainfunc.ai/
Search product entry: https://top.aibase.com/tool/sparkpage
2. Kimi Open Platform to launch Context Caching internal testing
The Kimi Open Platform recently announced that the highly anticipated Context Caching feature will soon begin internal testing. This innovative feature will support long-text large models and provide an unprecedented experience for users through an efficient context caching mechanism. Context Caching is an advanced technology that significantly reduces the cost for users when requesting the same content by caching repeated Tokens content.
【AiBase Summary:】
🔑 Context Caching feature supports long-text large models and provides an unprecedented experience through an efficient context caching mechanism.
🚀 Context Caching can intelligently identify and store processed text fragments, greatly improving API interface response speed.
💡 Suitable for scenarios with high scale and repetition, improving processing efficiency and reducing costs by reusing cached content.
3. TikTok launches AI suite Symphony, covering scriptwriting, video editing, and digital avatars
TikTok's Symphony AI content toolkit has completely changed the way content is created and shared, allowing everyone to become a creative master. Symphony Assistant provides thoughtful assistance, discovering trends, providing creative guidance, inspiring ideas, scriptwriting, and optimization suggestions. Symphony Creative Studio can generate multiple TikTok video previews in 60 seconds, supporting multilingual translation and video editing. Symphony Digital Avatars help brands expand creative strategies, providing realistic character avatars.
【AiBase Summary:】
🚀 Symphony Assistant provides comprehensive creative assistance, from trend discovery to scriptwriting, making the creative process simpler and more efficient.
💡 Symphony Creative Studio supports rapid generation of diverse TikTok video previews, multilingual translation, and video editing features to make content more attractive.
👤 Symphony Digital Avatars create realistic character avatars through generative AI, helping brands expand global creative strategies.
Details link: https://www.tiktok.com/business/en-US/blog/tiktok-symphony-ai-creative-suite
4. Baidu Wenku: AI product "Cheng Pian" supports 100,000-word long text generation
Baidu Wenku's latest AI product "Cheng Pian" has achieved significant breakthroughs in long text generation and multi-modal editing, providing users with comprehensive creation and editing functions. The product is supported by powerful AI technology, allowing users to easily access professional academic resources, create ultra-long graphic content, and achieve one-stop multi-format editing and adjustment.
【AiBase Summary:】
🚀 "Cheng Pian" supports 100,000-word long text generation and multi-modal editing capabilities, meeting users' full-chain needs in the professional field.
💡 Break the barriers of academic resources, allowing users to easily access global professional academic site materials and literature.
✨ Supports ultra-long graphic understanding and generation, uploading multiple format files at once and achieving rapid summary, Q&A, and creation.
Details link: https://top.aibase.com/tool/chengpianai
5. First AI college entrance exam evaluation results released, GPT-4o takes second place
In this unique AI college entrance exam evaluation, multiple AI models underwent comprehensive ability testing in Chinese, mathematics, and English, demonstrating potential and limitations in the academic field. Although performing well in Chinese and English subjects, there is still room for improvement in mathematical reasoning. With the advancement of technology, AI will become smarter and better serve human society.
【AiBase Summary:】
🧠 AI models participated in comprehensive ability testing, demonstrating academic potential and limitations.
📚 Performed well in Chinese and English subjects, with room for improvement in mathematical reasoning.
🚀 Technological progress will make AI smarter and better serve human society.
6. Flash Diffusion is applicable to any diffusion model, achieving image generation in a few steps
Flash Diffusion method brings revolutionary breakthroughs to image generation technology, accelerating the generation process of pre-trained diffusion models, performing excellently and efficiently versatile. Researchers adopt innovative means such as adjustable distribution and adversarial objectives to improve the positioning and computational efficiency of prediction models. The method adapts to different backbone networks, significantly reducing sampling steps while maintaining high-quality generation. Flash Diffusion injects new vitality, enhancing image generation efficiency and versatility, with the potential to have a profound impact in various fields.
【AiBase Summary:】
⚡ Accelerates the generation process of pre-trained diffusion models, performing excellently and efficiently versatile.
🔍 Adopts adjustable distribution and adversarial objectives, improving prediction model positioning and computational efficiency.
🌟 Adapts to different backbone networks, significantly reducing sampling steps while maintaining high-quality generation.
Details link: https://top.aibase.com/tool/flash-diffusion
7. AI-generated images can be "tailor-made"! Huawei and Tsinghua jointly launch personalized generation technology PMG
In an era where personalization is paramount, Huawei and Tsinghua University have jointly launched personalized generation technology called PMG. This technology utilizes users' historical behaviors and preferences to generate multi-modal content that meets user needs, such as emoticons, T-shirt design drawings, and movie posters. Through experimental verification, PMG technology has demonstrated great potential and commercial value, bringing users a richer and more personalized experience.
【AiBase Summary:】
⚙️ PMG technology utilizes users' historical behaviors and preferences to generate personalized multi-modal content.
💡 PMG extracts user preferences through keyword generation and latent vector generation to achieve multi-modal content generation.
📈 PMG technology has been validated in e-commerce clothing image generation, movie poster scenarios, and emoticon generation, demonstrating effective generation results.
Details link: https://github.com/mindspore-lab/models/tree/master/research/huawei-noah/PMG
8. Gboard revolutionizes typing experience, Google achieves one-click correction of all errors with large models
Gboard is Google's smart keyboard for mobile devices, and the latest "Proofread" feature uses large language models to achieve one-click correction of entire sentences and paragraphs, completely changing the traditional experience of correcting errors word by word. The feature has been launched on Pixel8 devices, benefiting many users. The research team optimizes model performance through supervised learning and reinforcement learning techniques, combined with a complex error synthesis framework to generate simulated datasets, demonstrating the huge potential of large models in enhancing mobile input interaction experiences.
【AiBase Summary:】
🔍 Uses large language models to achieve one-click correction of entire sentences and paragraphs, changing the traditional experience of correcting errors word by word.
🚀 Through a complex error synthesis framework to generate simulated datasets, combined with supervised learning and reinforcement learning techniques to optimize model performance.
💡 Deployed on cloud TPU V5, improving user input efficiency through optimized latency, etc.
Details link: https://arxiv.org/abs/2406.04523
9. NVIDIA's Lumina-T2X image generation can be used in Confyui
NVIDIA's Lumina-T2X image generation model can be used in Confyui, as an open-source model, it performs almost on par with the industry-leading MJ V6 in aesthetic expression and image quality, a particularly commendable achievement in the open-source field.
【AiBase Summary:】
🌟 Lumina-T2X uses a unified DiT architecture, capable of generating various media content, expanding the application scope of AI in content creation.
💡 Lumina-T2I image generation model improves generation quality and reduces training costs, demonstrating the economic potential of AI technology.
🔑 The success of Lumina-T2I lies in the model backbone using Large-DiT, the text encoding model using Llama2-7B, and the VAE using SDXL, laying the foundation for high-quality image generation.
Interested parties can use this plugin in Confyui: https://github.com/kijai/ComfyUI-LuminaWrapper
10. OpenAI's soul figure Ilya founds new AI company SSI, aiming for safe superintelligence
After leaving, Ilya Sutskever founded Safe Superintelligence Inc. to focus on solving the safety issues of superintelligent AI systems. The company is dedicated to researching the control and limitation of AI that surpasses human intelligence and plans to solve safety problems through engineering and scientific breakthroughs. SSI has been a for-profit entity from the beginning, unaffected by short-term commercial pressures, and is currently recruiting technical talents.
【AiBase Summary:】
🔒 SSI focuses on solving the safety issues of superintelligent AI systems, dedicated to researching the control and limitation of AI that surpasses human intelligence.
🚀 SSI plans to solve AI safety problems through engineering and scientific breakthroughs, improving AI capabilities and safety.
💼 SSI is a for-profit entity, unaffected by short-term commercial pressures, and is currently recruiting technical talents.
11. GPT-4 passes the Turing test, more than half of people cannot distinguish GPT-4 from humans
One of the important milestones in the field of artificial intelligence is the Turing test, and a recent experiment showed that GPT-4 was mistaken for a human with a probability of 54% in the interactive two-person Turing test, drawing attention to the realistic performance of AI systems. Participants were more inclined to use language style and social emotional factors to judge whether the other party was human, which had a profound impact on the discussion of machine intelligence.
【AiBase Summary:】
🤖 GPT-4 was mistaken for a human with a probability of 54% in the Turing test, showing realistic performance.
🔍 Participants were more inclined to use language style and social emotional factors to judge whether the other party was human.
💡 The results imply that AI systems may deceive humans in practical applications, raising new ethical, privacy, and security challenges.
12. AI design tool Kittl: Input text prompts to generate icons, clip art, etc.
Kittl is an AI-driven design platform that uses advanced algorithms and machine learning, allowing users to create high-quality design elements through simple text prompts without complex skills and software operation. It provides vector logo icons, stunning images, and clip art, advanced text editing, magic coloring, ready-to-use templates. Users can access unlimited content such as illustrations, fonts, photos, icons, textures, etc., easily drag and drop, and customize.
【AiBase Summary:】
⭐ Utilizes AI technology and machine learning, allowing users to create high-quality design elements through simple text prompts.
⭐ Provides thousands of professional design templates, without the need for complex design skills and software operation.
