阿里巴巴发布全新Qwen系列AI模型，支持多模态处理并适配移动设备
阿里推出其Qwen系列最新人工智能模型Qwen2.5-Omni-7B。该模型能够处理文本、图片、音频和视频，且效率高，可直接运行于手机和笔记本电脑上。现已于Hugging Face和GitHub公开，阿里巴巴预计其将被用于开发AI代理，例如为视障人士提供实时音频导航描述。
彭博社
📮投稿 ☘️频道 🌸聊天
[杂谈] 美服把把都是303回国服一看还是熟悉的味道
进来就是3c起手[s:ac:中枪]
分享一个纯净版 AI Chat UI
weijancc:
基于 NextChat 开源 Web 版本进行优化, NextChat 原版包含较多商业广告, 我对它进行了精简和界面美化, 效果预览:
快速上手:
- 访问在线演示: https://next-chat-web-lake.vercel.app
- 在左下角设置中配置自定义 API 接口即可开始使用。所有数据均在前端存储，不会泄露密钥.
docker 部署: 变量值修改为你使用的接口和密钥, 然后启动容器
docker run -d -p 3942:3000 \
-e OPENAI_API_KEY= \
-e BASE_URL= \
--name chatgpt-next-web \
--restart=unless-stopped \
cweijan/chatgpt-next-web
容器启动后, 访问http://localhost:3942 即可访问, 仅占 50m 左右内存
代码仓库: https://github.com/cweijan/NextChatWeb
自己部署 AI Chat 的意义:
- 我目前使用的 gpt-4o-mini, 通过 api 调用响应速度会比网页版快很多
- cursor chat 会计入 500 次快速对话, 因此在 cursor 内进行 chat 不划算
The 20 best Amazon Big Spring Sale deals under $30
Amazon’s Big Spring Sale is happening now through Monday, March 31st, with several dozen good-to-great deals available that don’t require an Amazon Prime membership. Our main roundup has you covered with more than 70 deals on Verge-favorite gear with quite a high price ceiling. But, for this post, we wanted to provide some low-cost alternatives […]
Mac 窗口失焦、无法操作的问题
Ayatox: 电脑型号：Macbook Pro(M1 Pro 芯片）
从升级 14.0 系统开始一直有一个问题，从一个 App 切换到另一个 App ，如果是先点击右键，就会失焦。
文本框无法输入，或者个别 App 切换到英文输入法。也无法复制，如果是文本编辑器，输入光标都没有。
访达的话，可以选中文件夹，无法重命名，左侧栏点哪里都没反应。
[解决的方法也简单] ：
左键点一下任意一个 App 就可以了。
也就是说，只要右键激活窗口，必然出现这个问题。很影响工作节奏！！！
哪位大神有解决方案？
从升级 14.0 系统开始一直有一个问题，从一个 App 切换到另一个 App ，如果是先点击右键，就会失焦。
文本框无法输入，或者个别 App 切换到英文输入法。也无法复制，如果是文本编辑器，输入光标都没有。
访达的话，可以选中文件夹，无法重命名，左侧栏点哪里都没反应。
[解决的方法也简单] ：
左键点一下任意一个 App 就可以了。
也就是说，只要右键激活窗口，必然出现这个问题。很影响工作节奏！！！
哪位大神有解决方案？
[讨论] 给大伙乐一乐抽到bug雾子反到被翻盘
前三局2：1我们优势，第4局抽到三镖爆炸牌感觉赢了[s:ac:blink]
和队友说了后明显大伙进攻都不积极了，然后我扔镖刮到墙炸死自己，近身混战崩死自己，对面源和大风车奥利莎反弹炸死我[s:ac:哭笑]
最后2：3输了[s:ac:哭]
和队友说了后明显大伙进攻都不积极了，然后我扔镖刮到墙炸死自己，近身混战崩死自己，对面源和大风车奥利莎反弹炸死我[s:ac:哭笑]
最后2：3输了[s:ac:哭]
Linux 6.15 Adds AMD Zen 5 SRSO Mitigation For KVM, Preps For Attack Vector Controls
While there is a lot of exciting new x86_64 CPU features coming with Linux 6.15, there is also some of the not so fun changes too: namely the "x86/bugs" pull request to bring the latest CPU security mitigation work to the mainline kernel.
The x86/bugs pull request has been merged for the Linux 6.15 kernel. This time around it has a new mitigation on the AMD side in enabling support for a new Speculative Return Stack Overflow (SRSO) mitigation for Zen 5 processors for that vulnerability dubbed "Inception". This new Zen 5 specific SRSO mitigation is for the Kernel-based Virtual Machine (KVM) and labeled as SRSO_MSR_FIX.
This new mitigation for Zen 5 Ryzen and EPYC processors is what began last year with Linux Prepares AMD "SRSO_USER_KERNEL_NO" Support For Zen 5 CPUs. With Linux 6.15 the mitigation is landing in refined form:
This doesn't change things outside the context of virtual machine (VM) use with Zen 5 processors.
Separately, the x86/bugs pull request has some preparatory patches to begin angling the Linux kernel to allow mitigating by attack vectors rather than controlling single vulnerabilities. That work isn't over the finish line with just some prep patches making it for Linux 6.15, but the topic is discussed further within Linux Attack Vector Controls Updated To More Easily Controlling CPU Security Mitigations.
More details within the x86/bugs pull request that has been merged to Linux 6.15 Git.
The x86/bugs pull request has been merged for the Linux 6.15 kernel. This time around it has a new mitigation on the AMD side in enabling support for a new Speculative Return Stack Overflow (SRSO) mitigation for Zen 5 processors for that vulnerability dubbed "Inception". This new Zen 5 specific SRSO mitigation is for the Kernel-based Virtual Machine (KVM) and labeled as SRSO_MSR_FIX.
This new mitigation for Zen 5 Ryzen and EPYC processors is what began last year with Linux Prepares AMD "SRSO_USER_KERNEL_NO" Support For Zen 5 CPUs. With Linux 6.15 the mitigation is landing in refined form:
"Add support for
CPUID Fn8000_0021_EAX[31] (SRSO_MSR_FIX). If this bit is 1, it indicates that software may use MSR BP_CFG[BpSpecReduce] to mitigate SRSO.
Enable BpSpecReduce to mitigate SRSO across guest/host boundaries.
Switch back to enabling the bit when virtualization is enabled and to clear the bit when virtualization is disabled because using a MSR slot would clear the bit when the guest is exited and any training the guest has done, would potentially influence the host kernel when execution enters the kernel and hasn't VMRUN the guest yet."
This doesn't change things outside the context of virtual machine (VM) use with Zen 5 processors.
Separately, the x86/bugs pull request has some preparatory patches to begin angling the Linux kernel to allow mitigating by attack vectors rather than controlling single vulnerabilities. That work isn't over the finish line with just some prep patches making it for Linux 6.15, but the topic is discussed further within Linux Attack Vector Controls Updated To More Easily Controlling CPU Security Mitigations.
More details within the x86/bugs pull request that has been merged to Linux 6.15 Git.
最佳调 prompt 实践是什么？
liyunlong41: 让 AI 去完成某项任务，prompt 需要不断调试来达到最优结果，那调试 prompt 最佳实践是什么？
是先选定模型，不断修改 prompt 来发掘出模型的能力，来达到最优结果，还是先定好具体的 prompt ，切换不同模型观察结果？ 各位有没有相关实践经验分享一下。
经常在让 AI 做一些任务时，很多时候不能达到自己的要求，下一步该怎么做，还是比较困惑。
是先选定模型，不断修改 prompt 来发掘出模型的能力，来达到最优结果，还是先定好具体的 prompt ，切换不同模型观察结果？ 各位有没有相关实践经验分享一下。
经常在让 AI 做一些任务时，很多时候不能达到自己的要求，下一步该怎么做，还是比较困惑。
