Skip to content

Commit

Permalink
add 6-28 ai daily
Browse files Browse the repository at this point in the history
  • Loading branch information
Freelander committed Jun 28, 2024
1 parent 57a6994 commit ae60ff6
Show file tree
Hide file tree
Showing 2 changed files with 100 additions and 1 deletion.
3 changes: 2 additions & 1 deletion docs/.vitepress/utils/createSideBar.ts
Original file line number Diff line number Diff line change
Expand Up @@ -4,7 +4,8 @@ export function createSideBarZH() {
text: '2024年',
collapsed: true,
items: [
{ text: '6 月 27 日', link: '/posts/2024/6-26' },
{ text: '6 月 28 日', link: '/posts/2024/6-28' },
{ text: '6 月 27 日', link: '/posts/2024/6-27' },
{ text: '6 月 26 日', link: '/posts/2024/6-26' },
{ text: '6 月 25 日', link: '/posts/2024/6-25' },
]
Expand Down
98 changes: 98 additions & 0 deletions docs/posts/2024/6-28.md
Original file line number Diff line number Diff line change
@@ -0,0 +1,98 @@
---
title: 6 月 28 日
date: 2024-06-28
cover: https://oss.justin3go.com/blogs/fav0-001.jpg

---

每天花 1 分钟获取最新 AI 信息。

内容涵盖但不限于**前沿 AI 资讯****AI 工具****AI 绘画****开源项目****学习教程**等。

简报主要特点是描述精简,但对于重要信息,还是通过独立帖子进行详细介绍。

以下是 6 月 28 日的最新 AI 信息。

### 前沿资讯

**1、OpenAI 公布一个基于 GPT-4 的模型:CriticGPT。**

主要用于帮助他们识别 ChatGPT 生成代码中的错误,用 GPT-4 找出 GPT-4 的错误,有意思!

在其发布的一篇论文中,指出比没有帮助的训练者表现更好,能够发现更多问题。

详细介绍:https://openai.com/index/finding-gpt4s-mistakes-with-gpt-4/

![GRGQLiWaMAA8Fwe?format=png&name=orig](https://p.ipic.vip/6m9v4c.png)

**2、Google 开源 Gemma 2,对开发者开放 Gemini 1.5 Pro API。**

Google 向所有开发者开放 Gemini 1.5 Pro 上下文 200 万 Token 长度的 API 访问权限。

详细介绍:https://developers.googleblog.com/en/new-features-for-the-gemini-api-and-google-ai-studio/

并开源了 Gemma 2,有 9B 和 27B 版本,从给的测试结果来看,27B 版本比 Llama 3 70B 稍微差些。

详细介绍:https://blog.google/technology/developers/google-gemma-2/

模型下载:https://huggingface.co/collections/google/gemma-2-release-667d6600fd5220e7b967f315

![A chart showing Gemma 2 performance benchmarks](https://storage.googleapis.com/gweb-uniblog-publish-prod/images/Gemma_2_performance_final.width-1000.format-webp.webp)

**3、character.ai 上线角色通话功能,让 AI 聊天变得更真实了。**

你可以像打电话一样与 AI 角色交流。可用它练习语言、模拟面试、角色扮演、创作故事情节等等。

正如视频号上这位 AyishaMi 博主,通过 AI 帮她出谋划策,甚至代替她与其英国邻居沟通噪音问题。

感兴趣的可以去下载 Character.AI App 体验下。

下载地址:https://qr.page/g/JmEyeLWYQS

详细介绍:https://blog.character.ai/introducing-character-calls/

![Image](https://p.ipic.vip/4oxhbt.jpg)

### 前沿技术

**1、支持超过 7000 种语言的文本转语音模型 ToucanTTS。**

一个专为对话场景设计的语音生成模型,特别适用于大语言模型(LLM)助手的对话任务,以及对话式音频和视频介绍等应用。

支持中文和英文等 7000 多种语言的多语言语音合成。实现多人对话语音合成,并可克隆说话者音色。

GitHub:https://github.com/DigitalPhonetics/IMS-Toucan

详细介绍:https://toucantts.com/zh

![image-20240628181058981](https://p.ipic.vip/bq9ir8.png)

### 开源项目

**1、Pipecat:一个可用于构建语音和多模态对话的 AI 开源框架。**

你可以用它来创建私人教练、会议助手、儿童故事讲述玩具、客服机器人等 AI 语音助手。

GitHub:https://github.com/pipecat-ai/pipecat

Cerebrium 联合 Daily.co 基于 Pipecat 开发了一款实时语音 AI 机器人,实现与人类对话接近的 500 毫秒语音到语音响应时间。

在线体验:https://fastvoiceagent.cerebrium.ai/

如果你也打算构建一个实时语音 AI 代理,可以看下它们分享的教程。

教程地址:https://docs.cerebrium.ai/v4/examples/realtime-voice-agents

![Try out the demo here: ](https://www.daily.co/blog/content/images/2024/06/fastbot-ui.png)



**2、一键给图片添加水印的开源免费工具: Image Watermark Tool。**

无需将图片上传到服务器,可直接在本地设备上轻松为你的图片添加水印,极大保护你的隐私。

你还可以自由选择水印颜色、大小、透明度、倾斜角度等个性化自定义。

GitHub:https://github.com/unilei/image-watermark-tool

另外,项目还提供了 Vercel 一键部署,感兴趣的可以自己部署使用。

0 comments on commit ae60ff6

Please sign in to comment.