资讯列表
Meta Collaborates with Georgia Tech to Launch CATransformers Framework to Reduce AI Carbon Footprint
Recently, Meta's FAIR team collaborated with the Georgia Institute of Technology to develop the CATransformers framework, which aims to make carbon emissions a core consideration in the design of AI systems. This new framework significantly reduces the total carbon footprint of AI technologies by jointly optimizing model architectures and hardware performance, marking an important step toward sustainable AI development.With the rapid popularization of machine learning technology, applications in areas such as recommendation systems and autonomous driving are increasing, but their environmental costs cannot be ignored. Many AI systems require substantial computational resources and often rely on custom hardware accelerators for operations.
5/15/2025 10:01:53 AM
AI在线
Media Tensions Escalate in Meta Antitrust Trial
In the recent antitrust trial of Meta conducted by the Federal Trade Commission (FTC), the long-standing tension between Silicon Valley and the media once again surfaced.During the trial, Mark Hansen, Meta's chief lawyer, mentioned in a fierce cross-examination of FTC’s key economic expert Scott Hemphill that Hemphill had proposed an anti-monopoly investigation into Meta alongside Facebook co-founder Chris Hughes and former Biden official Tim Wu in 2019.Hansen displayed a slide of this investigation proposal in court, which mentioned the "public recognition" of two journalists, Kara Swisher and Om Malik, who had reported on Meta’s aggressive acquisition strategy. Hansen attempted to undermine Hemphill’s credibility by referring to Malik as a “failed blogger,” implying a personal grudge against Meta.
5/15/2025 10:01:53 AM
AI在线
The Next Generation Open Source 3D Model Step1X-3D Debuts, AI Industry Trend Draws Attention
Recently, the technology sector welcomed a brand-new open-source 3D large model called "Step1X-3D." The release of this model marks another significant advancement in AI technology, particularly in 3D modeling and reasoning capabilities. Not only is this model open-source, but it also provides developers with various practical features, greatly promoting innovation and research possibilities.At the same time, Xiaomi is continuously expanding its presence in the AI field. It has recently applied for the "MiMo" trademark, which is intended to be used for inference large models.
5/15/2025 10:01:53 AM
AI在线
Alibaba Qianwen Wan2.1-VACE Open Source Claims to Be the First Open-source Unified Video Editing Model
Wanxiang "Wan2.1-VACE" has been announced as open-source, marking a major technological revolution in the video editing field. The 1.3B version of Wan2.1-VACE supports 480P resolution, while the 14B version supports both 480P and 720P resolutions. The emergence of VACE brings users a one-stop video creation experience, allowing them to complete various tasks such as text-to-video generation, image reference generation, local editing, and video extension without frequently switching between different models or tools, greatly improving their creative efficiency and flexibility..
5/15/2025 10:01:53 AM
AI在线
Alibaba Open Sources All-in-one Video Foundation Model to Empower Video Generation and Editing
On the evening of May 14th, Alibaba officially launched Tongyi Wanxiang Wan2.1-VACE, which is currently the most comprehensive video generation and editing model in the industry. The highlight of this model lies in its multiple powerful capabilities, enabling it to simultaneously achieve text-to-video generation, image-based video generation, video retouching, local editing, background extension, duration extension, and other foundational generation and editing functions. This innovative product further lowers the threshold for video production, allowing more creators to easily get started..
5/15/2025 10:01:52 AM
AI在线
AI+数据智能体的三大支点:数据治理、知识库和大模型
当销售部喊出"业绩增长15%",财务部却坚称"只有8%"。 会议室里争论不休,时间流逝,竞争对手已经抢占先机。 你不禁自问:明明砸了千万建设数据系统,为何企业依然深陷数据内耗?
5/15/2025 9:56:32 AM
大数据AI智能圈
今天起全员免费!GPT-4.1上线ChatGPT,首波实测:又快又听话,油腻感没了
今天凌晨开始,GPT-4.1可以直接在ChatGPT中使用了! 而且是不管付费的没付费的,所有用户均可使用那种~官方介绍,GPT-4.1是一款专门针对编码任务和指令执行的模型,推理效率非常高。 看看这张网友们自制的表格,它的能力一目了然:这家伙一个月前被OpenAI公开,当时声明专供API使用。
5/15/2025 9:34:39 AM
速度最快:Stable Audio Open Small 端侧音频模型登场,手机上 8 秒内 AI 生成 11 秒音频
AI 初创公司 Stability AI 推出 Stable Audio Open Small,号称是市场上速度最快的“立体声”音频生成 AI 模型,可在智能手机上运行。
5/15/2025 9:18:20 AM
故渊
重磅!谷歌DeepMind发布AlphaEvolve:AI界的“算法设计进化大师”诞生
谷歌DeepMind刚刚又往前拱了一大步,宣布推出 AlphaEvolve智能体 ,目标直指更上游,用于通用算法的设计发现和优化简单说,AlphaEvolve就像个AI界的“算法育种大师”。 它把自家Gemini大模型(Gemini Flash负责广撒网,洞察力强的Gemini Pro负责深挖)和一套“自动化考官”(负责验证算法靠不靠谱、效率高不高)结合起来,再套上一个“进化论”的框架,让好算法能一代更比一代强AlphaEvolve工作流程:工程师设定框架,AI通过“提示采样器”给LLM喂招,LLM出新招(程序),“考官”打分,好招进“兵器谱”,并用来启发下一轮出招。 去年DeepMind就秀过肌肉,证明LLM能生成代码函数来搞定科学问题。
5/15/2025 9:17:00 AM
刚刚,OpenAI开放GPT-4.1,100万上下文、代码能力超强
今天凌晨1点30,OpenAI宣布开放GPT-4.1,从今天开始可以在ChatGPT中使用。 GPT-4.1是一款专门针对编码任务和指令执行的模型,推理效率非常高,对于日常编码需求来说,是替代o3和o4-mini非常好的选择。 GPT-4.1是OpenAI发布的最新模型,其最大亮点之一就是支持100万tokens上下文,这也是OpenAI首次发布长窗口模型。
5/15/2025 9:16:00 AM
GPT-4o不敌Qwen,无一模型及格!UC伯克利/港大等联合团队提出多模态新基准:考察多视图理解能力
多视图理解推理有新的评判标准了! 什么是多视图理解? 也就是从不同视角整合视觉信息进而实现理解决策。
5/15/2025 9:10:00 AM
破解300年数学难题,智能体大突破!谷歌发布超强AI Agent
今天凌晨,谷歌Deepmind在官网发布了,用于设计高级算法的编程AI Agent——AlphaEvolve。 AlphaEvolve与谷歌的大模型Gemini实现深度集成,用于自动评估通用算法的发现与优化,可以帮助开发人员快速设计出最好、高效的矩阵算法。 简单来说,大模型擅长生成各种想法和算法,但是没人知道这些到底行不行,而AlphaEvolve相当于“质检员”,能够按照特定标准来衡量这些想法是否可行。
5/15/2025 9:08:00 AM
ICML25 | 让耳朵「看见」方向!仅依靠360°全景视频,就能生成3D空间音频
空间音频,作为一种能够模拟真实听觉环境的技术,正逐渐成为提升沉浸式体验的关键。 然而,现有的技术大多基于固定的视角视频,缺乏对360°全景视频中空间信息的充分利用。 在这样的背景下,一项在空间音频生成领域具有里程碑意义的研究应运而生——OmniAudio:它能够直接从360°视频生成空间音频,为虚拟现实和沉浸式娱乐带来了全新的可能性。
5/15/2025 9:05:00 AM
DanceGRPO:首个统一视觉生成的强化学习框架
本文由字节跳动 Seed 和香港大学联合完成。 第一作者薛泽岳为香港大学 MMLab@HKU 在读博士生,在 CVPR、NeurIPS 等国际顶级会议上发表多篇研究成果。 项目通讯作者为黄伟林博士和罗平教授。
5/15/2025 9:04:00 AM
25岁MIT辍学天才一战成名!3年成为90亿美金公司CEO
硅谷又出现了一位新的天才。 AI浪潮中,一位年仅25岁的远见者正以惊人的速度改写着软件开发的未来,他就是Michael Truell,AI代码编辑器Cursor背后的母公司Anysphere的CEO。 Cursor仅仅用了12个月,ARR就达到了一亿美元,多篇业内分析认定Cursor是 「SaaS史上最快到$100M ARR的初创公司」。
5/15/2025 9:02:00 AM
Meta 推出 CATransformers 框架 助力AI行业实现减排目标
在人工智能迅猛发展的今天,Meta 的 FAIR 团队与佐治亚理工学院联合研发了一款名为 CATransformers 的全新框架。 该框架以降低碳排放为核心设计理念,旨在通过优化模型架构与硬件性能,显著减少 AI 技术在运营中的碳足迹,为可持续的 AI 发展奠定基础。 随着机器学习技术在各个领域的广泛应用,从推荐系统到自动驾驶,其背后的计算需求不断增加。
5/15/2025 9:01:02 AM
AI在线
阿里通义万相Wan2.1-VACE开源 号称首个开源的视频编辑统一模型
通义万相宣布VACE开源,这标志着视频编辑领域迎来了一次重大的技术革新。 此次开源的Wan2.1-VACE-1.3B支持480P分辨率,而Wan2.1-VACE-14B则支持480P和720P两种分辨率。 VACE的出现,为用户带来了一站式的视频创作体验,用户无需在不同模型或工具之间频繁切换,即可完成文生视频、图像参考生成、局部编辑与视频扩展等多种任务,极大地提高了创作效率和灵活性。
5/15/2025 9:01:02 AM
AI在线
阿里巴巴开源全能视频大模型,赋能视频生成与编辑
5月14日晚,阿里巴巴正式推出了通义万相 Wan2.1-VACE,这是当前行业中功能最为全面的视频生成与编辑模型。 该模型的亮点在于它具备多种强大的能力,可以同时实现文生视频、图像参考视频生成、视频重绘、局部编辑、背景延展和时长延展等多项基础生成和编辑功能。 这一开创性的产品标志着视频制作的门槛进一步降低,使更多的创作者能够轻松上手。
5/15/2025 9:01:02 AM
AI在线