AI News

几天前，我们刚刚聊过 Cline 3.1 的一些疯狂新特性。 说实话，我个人一直对 Cline 相当有好感，他真的是一个很不错的开源项目：Cline v3.1 疯狂更新！最佳自主的 AI 编程助手...

每个人心中都有一个电影梦——想要走入不同角色体验戏中人生，或成为导演编排每一组镜头，或作为编剧书写平行宇宙里的无限可能。 海螺AI是一台造梦机，让不同人都能拥有接近电影的方式。新年伊始，海螺AI为全球...

1月7日，西藏定日县发生6.8级地震，很多人关注着救援进展，祈祷灾区平安。同时，在人们的善意和关切中，一张“小男孩被埋废墟下”的图片却迅速在网上走红。 这张图配着“日喀则地震”等字样，戳中了无数人的泪...

我们推出了 vdr-2b-multi-v1 ，这是用于视觉文档检索的最佳多语言嵌入模型。我们还发布了它的纯英文版本 vdr-2b-v1 ，并开源了新的 vdr-multilingual-train 数...

✨ Little Red Book Nuggets Secrets Revealed! 🔥 Leek programs can be fun? There are also operation tutorials for free! Hello everyone, recently found that everyone is very interested in the small red book related projects, it seems that everyone wants to dig gold in the small red book this blue sea ah! 🚀 Don't worry, today I bring you...

Coze（扣子）自动化工作实战教程简介 在现代工作环境中，自动化技术正在以其高效、精准和可扩展性，成为企业提升生产力的重要工具。**Coze（扣子）**作为一款轻量化、高灵活性的自动化工具，为各行...

在AI的快速发展中，数字人（Digital Humans）已经逐渐成熟，可以低成本的快速生成。因广泛的商业应用场景，受到大家的关注。无论是在虚拟现实（VR）、增强现实（AR）还是影视制作、游戏开发、品...

英伟达 (NVDA) 、Alphabet 旗下 谷歌 (GOOGL) 和热门初创公司 OpenAI 正在转向“合成数据”工厂，以满足训练深度学习人工智能算法所需的大量数据需求。在本周的消费电子展上，这...

你是否希望手机上根据你兴趣推送的新闻信息流直接生成一条播客自动播放？Google 的新功能“每日聆听”或许正是答案。 Google 正在测试一项名为“每日聆听”（Daily Listen）的新功能...

AI 在教育领域的崛起 人工智能（Artificial Intelligence，简称 AI）是一门致力于研究、开发和应用能够模拟、延伸人类智能的理论、方法、技术及应用系统的新兴科学。自 20 世...

以下是一些在教育教学中比较实用的AI工具： 1. 言笔AI教学助手 ：这是一个国内领先的AI教学设计平台，提供全方位的功能支持系统。核心教学功能包括教案生成、课程设计、作业设计、教学素材自动生成以...

今早收到消息：商业新玩法！赞赏组件内测邀请！恭喜您，获得赞赏组件内测机会！一键开通即可躺赚！ 去年早些时候，首席AI分享圈介绍了 文心智能体平台，同时给大家分享一种引流技巧 文心智能体引流提示词：通过...

2025 has just begun, AI video generation is about to usher in a technological breakthrough? This morning, Ali's Tongyi Wanxiang video generation model announced a heavy upgrade of version 2.1. The newly released model has two versions, respectively, Tongyi Wanxiang 2.1 Extreme Edition and Professional Edition, the former focuses on efficient...

A rising AI trend in the US adult entertainment industry is attracting a lot of attention. This time it features a star who has been struggling in the industry for years.Chloe Amour, 33, has decided to hand over some of her work to a virtual double by selling her image to an AI company. This...

Phi-4，一个 140 亿参数的开源模型，其性能与 Microsoft 的 OpenAI GPT-4o-mini 相媲美，现已在 Ollama 上可用。 ollama run phi4 主要...

2025 年消费电子展在展示奇异科技方面没有让人失望。从机器人到金戒指，以下是对最古怪的创造的一些速览： 害羞的机器人 Mirumi： Yukai Engineering推出了Mirumi，这款迷人的...

Dify Plugin Introduction Dify v1.0.0-beta released, the core lies in the introduction of the plug-in system and Marketplace (component marketplace), through the modular components to achieve flexible expansion of AI applications, plug-ins contain models, tools, agent strategies, etc., support for unique...

什么是Dify插件？ 来自第三方的模型与工具是帮助开发者增强应用能力的重要单元。虽然 Dify 平台已内置多个由官方维护与社区贡献着开发的工具，但在此模式下，现有的工具难以全面覆盖各类细分场景的需求...

引言 DashInfer-VLM是一个针对于视觉多模态大模型VLM的推理架构，特别优化了Qwen VL模型的推理加速，DashInfer-VLM和其他的VLM的推理加速框架最大的区别是， 它把VIT部...

背景 自2023年以来，AI大模型在全球范围内掀起了有史以来规模最大的人工智能浪潮。进入2024年，全球大模型竞争态势日益加剧，随着Sora、GPT-4o、o1的发布，国内大模型在2024年进行了波澜...

Cline 的3.1版本，又来了。开发者们的劲头，真是疯狂。 实际上，他们一周前才刚刚才推出了 3.0 版本。我已经好久没讲 Cline 了，这次把几个版本的更新，统一讲了。 我还是要先简单解释一下 ...

一句话总结：能放在你桌面上的个人AI超级计算机 NVIDIA 今天在CES 2025大会上发布了 NVIDIA Project DIGITS，这是一款能放在你桌面上的个人AI超级计算机。 将传统上...

Industry-leading mapping AI comparison, subjective after extensive personal testing Note: All products here are directly available and offer a small daily usage quota. Composition hold is a modeling capability, use requires a technical foundation that may not be supported by the product.

今天，我们很高兴地推出 Dify v0.15.0 版本，带来了全新的「父子检索」功能。这是一种在检索增强生成（RAG）系统中实现的高级技术，旨在进一步提升信息获取与上下文理解能力。通过这一功能，Dif...

Yesterday, Sam Altman, CEO and co-founder of OpenAI, posted his latest in-depth article, Reflections, on his personal blog. The main review of OpenAI's founding in the past 9 years: from the initial unappreciated by the outside world, to the 20...

罗永浩这次又要进军AI行业了。此前报道，他的新公司“细红线”将在蛇年春节前后，发布自成立以来的首款新品。 而早在去年4月，罗永浩就在直播中首次预告，称将发布一款神秘产品，并形容其“具有颠覆性、破坏式的...

Making predictions, especially in a fast-moving field like data and AI, is notoriously difficult. Nonetheless, we, Rajesh Parikh and Sanjeev Mohan, released our 2024 trend predictions last year. With 2024 ...

The development of AI models is becoming more and more diversified. In addition to large-scale language models and small-scale language models, "world models", which are called world simulators, are being regarded as one of the next key development directions of AI. 2...

In the current wave of rapid development of AI technology, several open source projects provide developers with powerful features and flexibility, especially in the field of Multi-Agent Systems (Multi-Agent Systems). Here are five of the most popular Agent projects (GitHub...

UCLA plans that a comparative literature course at the school in 2025 will be powered by AI to generate textbooks, assignments and provide a teaching assistant ready to explore questions. Thanks to AI, the format and materials for this 2025 course will take on a whole new look. "This new technology will...

What is flowith 2.0? flowith is a next-generation AI authoring platform dedicated to exploring more natural and efficient interactions between humans and AI. In version 1.0, we gained hundreds of thousands of users through canvas-based interaction. In the upcoming...

In 2021, Microsoft launched GitHub Copilot, which jumped to become the most sought-after AI tool in the programming world. GitHub Copilot can automatically generate complete code functions based on contextual information provided by the user, such as function names, comments, code snippets, etc., and is known as a coding...

General Introduction Common Sense Education offers a series of free AI literacy courses designed for students in grades 6-12. These lessons are designed to help students understand the basic concepts of Artificial Intelligence (AI) and its social and ethical implications. The curriculum covers AI...

Our Plans for Model Context Protocol Evolution (H1 2025) Model Context Protocol is evolving rapidly. This page outlines our key priorities for the first half of 2025...

Experts generally believe that 2024 is the year of AGI. In this year, the big model industry has undergone radical changes: OpenAI's GPT-4 is no longer out of reach; image and video generation models are becoming more and more realistic; multimodal big language models, reasoning models, and intellectual...

I have found that there is quite a lot of interest and demand for digital people. Recently very many of you, because of the previous article written about digital people, private message me to chat about digital people. Here, I'm going to re-discuss them, pick 4 models and share them with you. These 5 models, mainly public modeling digital person mainly (public image). Such as...

Document image understanding technology aims to enable computers to understand the content in document images as well as humans do. It mainly involves analyzing, processing and understanding document images (e.g., paper contracts, book pages, invoices, etc.) obtained by scanning or photographing, extracting valuable information in them, such as text, tables, charts, etc., and...

Each of these knowledge points has different content for teachers and students. In 2024, the Massachusetts Institute of Technology (MIT) launched the Day of AI program to fire up the world...

FaceFusion has been updated to version 3.1.1. This update adds the batch feature, face swap model, and a new UI interface. The batch this time is different from the job workflow form of the previous version, and the operation is more convenient and simple. In this article, we will use a FaceFusion...

Retrieval-augmented generation (RAG) has emerged as a powerful technology for enhancing the capabilities of large language models. The RAG framework combines the advantages of retrieval-based systems and generative models to produce more accurate, context-aware, and timely responses. As the demand for sophisticated AI solutions grows...

The most popular AI product of 2024 is NotebookLM. It's been a hit since September, and it's remained so until the end of the year. In December, NotebookLM was updated to include new features. Users can also be a part of the podcast program...

1. What is a No Code/Low Code platform? Simply put, it allows people to create applications, websites or business processes without writing any code. Users can do this by simply clicking or dragging and dropping components. For beginners, creating technology projects...

With the rapid development of artificial intelligence technology, the reasoning ability of large language models on difficult scientific topics at the graduate level has become a hot topic of research. Taking OpenAI as an example, its new model OpenAI o1, officially released in early December, demonstrates strong scientific reasoning ability...

FastGPT is a knowledge base Q&A system based on the LLM large language model developed by the Ringworld Cloud team, providing out-of-the-box data processing, model invocation and other capabilities. Meanwhile, workflow orchestration can be performed through Flow visualization to realize complex Q&A scenarios.FastGP...

Comprehensive Introduction Xorbits Inference (or Xinference for short) is a powerful and versatile library focused on providing distributed deployment and services for language models, speech recognition models, and multimodal models. With Xorbits Inference...

关于主流闭源LLM的参数规模一直讨论不断，在2024年最后2天来自微软的一篇关于检测和纠正临床笔记中医疗错误的测试基准MEDEC的研究一不小心直接漏了它们的参数规模：o1-preview, GPT-4...

The copilot function in OneDrive is very powerful, it grasps all the files as a whole from one location, and summarizes and compares multiple files, and has completed complex work. Of course, the above features require a subscription to Microsoft 365 copilot business edition...

难以想象，如果不是Scaling Law放缓，2024年AI领域会发生哪些惊人变化，但你可能又会感到庆幸，正是由于Scaling Law放缓，它给了这个行业的后来者们追赶的机会，也给了更多普通人搭乘这...

About Free Model Rate Limit api call total consumption $0 - $50/month (not included) GLM-4-Flash: concurrent 200 GLM-4V-Flash: concurrent 10 Cogview-3-Flash: concurrent 5 CogVideoX...

英伟达，这个GPU领域的巨头，又搞出大新闻了。这次，他们以据说高达 7亿美金 的价格收购了以色列软件初创公司 Run:ai。不仅如此，他们还宣布将把 Run:ai 的软件开源！这个操作，直接炸翻了AI...

Highlights Analytics 1.58-bit FLUX, the first quantization model that reduces the parameters of the FLUX Visual Transformer (totaling 11.9 billion) by 99.5% to 1.58-bit, eliminating the need to rely on image data and drastically reducing storage requirements...

课程讲师: Pranav Rajpurkar 博士 (哈佛大学助理教授) 课程概述: 本课程将带您深入了解前沿的AI开发工具，如PyTorch、Lightning和Hugging Face，并使用VS...

Conclusion Domestic primary and secondary schools have issued documents to popularize AI education from top to bottom, and the mature stage of the "industry" is to get certificates, advancement, training, and finally become a rich man's game. It may be better to follow the example of the United States and directly enter the experimental stage of popularization of science or learn from Japan to give a clear guiding learning framework for the early stage of practice...

Recently, the speech team of Ali Tongyi Labs officially released the speech synthesis model CosyVoice2.The model supports bidirectional streaming of text and speech, supports multilingualism, mixed languages and dialects, and provides more accurate, stable, faster and better speech generation capabilities. Now, the silicon-based flow Sil...

Deep Research is a member feature of Gemini, following the synchronization of 2.0, currently unavailable to domestic users. As a content creator who often needs to do research and write reports, I recently tried Google's new Gemini...

本周五晚间，OpenAI 突然宣布了公司重组的消息，不仅让马斯克，也让我们有些措手不及。 根据 OpenAI 的最新声明，新一轮组织结构调整是围绕营利与非营利的矛盾展开的。 在 ChatGPT 推出之...

度小满开源全球首个金融行业推理大模型——轩辕-FinX1！ 该模型是金融领域首个类GPT-O1推理大模型，采用创新的“思维链+过程奖励+强化学习”训练范式，显著提升逻辑推理能力，并可展示O1模型未公开...

序言 最近与几位朋友讨论了O3（OpenAI o3），他们的反应基本可以概括为："天啊，这真的发生了吗？" 是的，这确实正在发生。接下来的几年将会疯狂。这是具有历史意义的时刻，甚至可以说是具有银河系级...

这是 Perplexity 继 2023 年收购 Spellwise 后的第二次收购，当时 Spellwise 的 CEO 负责开发 Perplexity 的手机应用。 Perplexity 此次...

DeepSeek-V3是一款强大的混合专家（Mixture-of-Experts, MoE）语言模型，拥有6710亿总参数，针对每个token激活37亿参数。该模型采用了一种创新的多头潜在注意力（Mu...

今天早些时候收到“寻光”内测申请通过的通知，睡前发个简单的评测。 该平台定位是达摩院的“视觉技术能力应用平台”，目前应用较少（对比发布会）期待逐步开放更多视觉类应用。 寻光分为两个地址： https...

General Description mymind is a revolutionary personal information collection and management tool that uses AI technology to simplify the process of storing and retrieving information. Unlike the traditional way of categorizing folders, mymind allows users to naturally save any form of inspiration, ideas, and web content, including diagrams...

Cursor is a good thing, but good things can go wrong in all sorts of ways. A conversation (be it Chat or Composer) that goes wrong can often be solved by opening another docking station, but the context is lost. We can try to fix this by bringing in external...

评估大模型在真实世界、长文本、多任务中的「深度理解与推理」能力 近年来，长文本大语言模型的研究取得了显著进展，模型的上下文窗口长度已经从最初的 8k 扩展到 128k 甚至 1M 个 tokens。然...

大家都在用AI工具，我们看着AI一步一步发展壮大，以前大部分是我们仅仅用文字跟它们聊天，有些时候果核就会在想：要是什么时候能对图片进行很好的思考就好了。 研究了一堆AI之后，后面用起Kimi，发现它的...

今天凌晨2点，OpenAI的12天直播，终于来到了最终章。OpenAI o3正式发布！ o3 是 o1 系列模型的继任者。这类模型的特点是让模型在回答问题之前花更多时间思考（推理），从而提高回答的...

屏幕上密密麻麻的代码中夹杂着各种模型API的配置信息，桌上的咖啡早已凉透。 这是许多开发者在尝试构建AI应用时的真实写照：环境配置繁琐、API成本高昂、文档支持不足...... "如果能有一个统一的平...

In the ever-evolving field of audio production, artificial intelligence is making significant strides, providing a set of tools that can revolutionize the way creators approach sound design. For podcast producers, musicians, and content creators, these advances mean more efficient workflows and higher quality audio effects...

If you're looking for affordable Artificial Intelligence (AI) tools to help you begin the journey of making AI everyday, from email to video production, it's easy to do. This quick start guide features 10 amazing AI services and platforms that will keep you from spending your hard-earned...

每个人都希望在工作中提升自己的生产力和效率。无论是快速处理 Excel 表格的小技巧，还是能够轻松同步到现有工作流程的工具，在竞争激烈且动态多变的职场中，每一点优势都至关重要。⚡ 这正是像 Glean...

Google 发布了一款被称为全新“推理”AI模型的产品——但目前还处于实验阶段，从我们的简短测试来看，该模型确实还有提升空间。 这款新模型名为 Gemini 2.0 Flash Thinking E...

The company's valuation has tripled since June. Perplexity AI Inc. is an artificial intelligence startup developing a search product to compete with Alphabet Inc.'s Google. According to people familiar with the matter...

Heavy news in the AI circle exploded. Alec Radford, the legendary OpenAI researcher known in the industry as the "father of GPT", announced that he was leaving the company to pursue independent research. As the chief designer of the GPT series, the core technology behind ChatGPT, Radford's decision...

GitHub has announced a free program for its AI programming assistant, GitHub Copilot, now available to all users in Visual Studio Code. All users need is a GitHub account to start using...

In large model applications, processing complex requests is often accompanied by high latency and cost, especially when there is a lot of repetition in the request content. This "slow request" problem is especially prominent in scenarios with long prompts and high-frequency interactions. To address this challenge, OpenAI recently ...

Clio: A Real-World AI Usage Insight System for Privacy What do people use AI models for? Despite the rapidly growing popularity of big language models, until now we've lacked insight into exactly how they're used. It's not just a matter of curiosity...

Small models can outperform larger models if they are given longer to think. In recent times, there has been an unprecedented amount of enthusiasm in the industry for small models, with a number of 'practical tricks' to allow them to outperform larger scale models in terms of performance. It can be argued that putting the spotlight on improving smaller...

Frameworks like LangChain, CrewAI, and AutoGen have become popular by providing high-level abstractions for building AI systems. However, many developers, including myself, have found that these tools do more harm than good, often adding unnecessary complexity and frustration to the development process...

Overview In the age of the information explosion, organizations have come to rely on search technology not just to find content, but to improve efficiency and productivity. However, traditional search models often struggle to truly understand user intent, resulting in inaccurate, irrelevant or even incomplete search results. This experience not only frustrates users...

Everyone can customize the "Research Knowledge Base Model" from 0 base. Model out of artificial customer service has become a foregone conclusion! [Openai released Project features] 1. Support for uploading files to Project, building a knowledge base for a specific field. 2. 2. Support networking search, real-time access to the latest ...

我们 R 系列中最小的模型提供顶级的速度、效率和质量，可在普通 GPU 和边缘设备上构建强大的 AI 应用程序。 今天，我们很高兴发布 Command R7B，这是我们专为企业开发的大语言模型（LLM...

GLM-4V Series The GLM-4V series contains 3 models, which are suitable for different application scenarios. GLM-4V-Plus:With excellent multimodal understanding capability, it can process up to 5 images simultaneously and supports video content understanding, which is suitable for complex multimedia analysis scenarios. ...

Earlier this year, Google launched its video generation model Veo and its newest image generation model Imagen 3. Since then, it's been exciting to see people bring their ideas to life with these models: YouTube creators are exploring the possibilities for YouTub...

Recently, GenmoAI open-sourced the video generation model mochi 1 preview (10B) with high-fidelity actions and robust cue following capabilities, currently supporting 480p resolution video generation. Today, SiliconCloud, a silicon based flow, went live with an inference accelerated version of mo...

In today's competitive e-commerce market, how to make your product stand out from the crowd of choices has become a challenge that every brand and business must face. The importance of visual marketing as one of the key factors for e-commerce success cannot be overstated. An attractive and professional product image display not only...

Four 10s! It's a rare occurrence, but in ICLR, where the average score is only 4.76, how can it not be considered as quite a blowout? The paper that has won over the reviewers is IC-Light, a new work by ControlNet author Lumin Zhang. we...

Reasoning is unpredictable, so we have to start with incredible, unpredictable AI systems. Ilya has finally shown up, and right off the bat, he's got something amazing to say. This Friday, Ilya Sutskever, former chief scientist at OpenAI, spoke at the Global ...

Phi-4 在仅有 140 亿（14B） 个参数的情况下，通过创新的训练方法和高质量的数据，展现出了媲美甚至超越一些更大规模模型的性能。本文将详细介绍 Phi-4 的架构、特点、训练方法，以及在实际应...

Learn how Rexera migrated to LangGraph to create powerful quality control intelligences for real estate business processes and significantly improve the accuracy of their Large Language Model (LLM) responses. Rexera is revolutionizing manual processes by leveraging AI to automate...

Are you still using Baidu to search for information? The efficiency is too low! In the era of information explosion, accurately obtaining the required information has become a scarce ability. The first step to improve your search ability is to change your search engine! The traditional search engine is not only full of advertisements and complicated information, but also inefficient, making it difficult for you to judge the search...

Preface As the saying goes, "A good tool for a good job is a good tool for a good job". Choosing a good tool is especially important for programmers. Whether it's for automating code or debugging, AI coding assistants can save you time and improve the quality of your code. In this article, we'll explore some of the most frequently used by programmers around the world...

作为 Devin 正式推出的一部分，我们将为选定的开源项目维护者提供 500 个免费 ACU 的 Devin 团队计划配额。立即访问 app.devin.ai 注册，并通过 osi@cogniti...

From Sundar Pichai, CEO of Google and Alphabet: Information is at the heart of human progress. That's why we've been working for 26 years to organize the world's information and make it accessible and useful. It's also why I...

On November 20, 2024, OpenAI launched a free online course designed to help K-12 teachers introduce ChatGPT into their teaching. The course is called ChatGPT Founda...

Following the free open language model in August 2024 Wisdom Spectrum supports 128K context/4K output length bigmodel GLM-4-Flash full volume free open! After the August 2024 release of GLM-4-Flash with 128K contexts and 4K output lengths, we are now open to the public on the bigmodel...

自从 Chatgpt 诞生以来，LLM（大语言模型）的参数量似乎就成为了各个公司的竞赛指标。GPT-1 参数量为 1.17 亿（117M），而它的第四代 GPT-4 参数量已经刷新到了 1.8 万...

Meta has introduced Llama 3.3, a large language model with 70 billion parameters that rivals the performance of its predecessor, 405B-parameter Llama 3.1, and reduces input cost by a factor of 10 over Llama 3.1 405B! Instructions...

从我们专为推理和训练打造的 Inferentia 和 Trainium 芯片，到提供业界最佳的基础模型（Foundation Models）于 Amazon Bedrock ，以及由 AI 驱动的...

If you have used Conch APP, you may have used his voice cloning, the effect is very good, but you can only use it as a voice assistant in the APP. This time, Conch AI video international version official website, new "Audio" channel, you can easily clone your own voice. Back to the main interface of text-to-speech...