Alex(@renyu0722) 's Twitter Profileg
Alex

@renyu0722

旅行,创业,思考,分享。

ID:181022552

calendar_today21-08-2010 02:30:13

1,9K Tweets

150 Followers

2,4K Following

Jim Fan(@DrJimFan) 's Twitter Profile Photo

The upcoming Llama-3-400B+ will mark the watershed moment that the community gains open-weight access to a GPT-4-class model. It will change the calculus for many research efforts and grassroot startups. I pulled the numbers on Claude 3 Opus, GPT-4-2024-04-09, and Gemini.…

The upcoming Llama-3-400B+ will mark the watershed moment that the community gains open-weight access to a GPT-4-class model. It will change the calculus for many research efforts and grassroot startups. I pulled the numbers on Claude 3 Opus, GPT-4-2024-04-09, and Gemini.…
account_circle
Abel Art(@A_B_E_L_A_R_T) 's Twitter Profile Photo

Another little test with Metahuman for an upcoming project.
ViggleAI to be able to have a reference image and a more realistic rendering and Kaiber on the animated background.

account_circle
歸藏(guizang.ai)(@op7418) 's Twitter Profile Photo

Llama3 正式发布,开源模型的质量又将迎来再一次飞跃。

卧槽,最大的模型参数超过 400B,还在训练中,完成将会发布,还有多模态模型。

他们的测试结果看,70B全面超越Gemini Pro 1.5以及Claude3 Sonnet。

👇下面是我梳理的详细介绍:

1️⃣性能测试:

预训练和指导微调模型是目前 8B 和 70B…

Llama3 正式发布,开源模型的质量又将迎来再一次飞跃。 卧槽,最大的模型参数超过 400B,还在训练中,完成将会发布,还有多模态模型。 他们的测试结果看,70B全面超越Gemini Pro 1.5以及Claude3 Sonnet。 👇下面是我梳理的详细介绍: 1️⃣性能测试: 预训练和指导微调模型是目前 8B 和 70B…
account_circle
Gradio(@Gradio) 's Twitter Profile Photo

🎉BIG NEWS: Meta released 𝐋𝐥𝐚𝐦𝐚𝟑

Detailed 🧵:
🌟2 sizes: 8B for efficient deployment, 70B for large-scale apps
💡Base & Instruction-tuned variants available, ++ Llama Guard 2 safety model
🤯Expanded vocab size to 128K tokens; better encoding efficiency & multilingualism

🎉BIG NEWS: Meta released 𝐋𝐥𝐚𝐦𝐚𝟑 Detailed 🧵: 🌟2 sizes: 8B for efficient deployment, 70B for large-scale apps 💡Base & Instruction-tuned variants available, ++ Llama Guard 2 safety model 🤯Expanded vocab size to 128K tokens; better encoding efficiency & multilingualism
account_circle
Pierrick Chevallier | IA(@CharaspowerAI) 's Twitter Profile Photo

The team has just rolled out an early test of their 'rooms' system. For now, it's only open to those who have generated 10k or more images.

The features are :
•Text chat
•Voice chat
•Image creation system

Eventually, users will be able to create public or private…

account_circle
Javi Lopez ⛩️(@javilopen) 's Twitter Profile Photo

⚡ Inside Magnific Headquarters

In Spain, startups are not created in garages but in small, untidy offices. Come with me on this visit to where the magic happens 🦄

Have you ever seen a cooler office? Don't think so! It's dog-friendly! 🐶😂

account_circle
小互(@imxiaohu) 's Twitter Profile Photo

SD 3在线体验 ↓

1、直接体验,输入提示即可,速度很快:sdxlturbo.ai/stable-diffusi…

2、抱抱脸体验:huggingface.co/spaces/latentc…

需要API,申请SD 3 API点这里,很容易申请:platform.stability.ai/account/keys

3、这里有个开源方案,也可以自己部署一个:github.com/Doriandarko/sd…

account_circle
Heather Cooper(@HBCoop_) 's Twitter Profile Photo

🎇Retro-Futuristic Comic Art🎇

I'm working on a style using classic comic book art and modern digital illustration in prompts to create a 3D effect in vivid colors.

Here's several examples I animated with Leonardo.Ai Img2Motion:

account_circle
Amira Zairi(@Azed__design) 's Twitter Profile Photo

Chinese ink art RemixAI

💬Prompt:
Traditional Chinese ink, a [Subject +description], fine lines, fluid and expressive brush strokes, Monochrome, black and white on textured paper.

Model: Chenney
Ratio: 16:9

🧲استخدمت النموذج شيني لاني حسيت أعطاني نتيجة احسن ولانه اكثر…

Chinese ink art @getremixai 💬Prompt: Traditional Chinese ink, a [Subject +description], fine lines, fluid and expressive brush strokes, Monochrome, black and white on textured paper. Model: Chenney Ratio: 16:9 🧲استخدمت النموذج شيني لاني حسيت أعطاني نتيجة احسن ولانه اكثر…
account_circle
Tom Huang(@tuturetom) 's Twitter Profile Photo

红杉发布的 2024 AI 50 强榜单,GenAI 进入第二年:

1. AI 涉及领域增长:企业生产力工具、创意软件、基础设施
2. AI 推动企业降本增效:工作流程、体验革新、AI Agent,大公司正通过将AI融入其产品的方式而获益

23 年 60% 投资流向 LLM Infra,但是应用仍占据主导地位。

sequoiacap.com/article/ai-50-…

account_circle
-Zho-(@ZHOZHO672070) 's Twitter Profile Photo

文本直接生成 3D 全景!DreamScene360

使用文生图扩散模型、GPT-4V 提示词自优化机制和 Gaussian splatting 来实现高质量 360 度全景快速生成。并通过优化点云和 3D Gaussian,精确地重建场景中的不可见区域,提升场景几何一致性和视觉效果

dreamscene360.github.io

account_circle
indigo(@indigo11) 's Twitter Profile Photo

Ezra Klein Ezra Klein 最新一期 Podcast采访了 Anthropic CEO,其中透露了 Dario Amodei 对 AI 发展的速度和安全性的很多独到见解。感觉这些看法和上个月见到 Anthropic AI Safety 产品负责任人了解到的基本一致,看来他们公司的价值观对齐的非常好,全员 70% Doomer 😅

Anthropic 并不爱提 AGI…

Ezra Klein @ezraklein 最新一期 Podcast采访了 Anthropic CEO,其中透露了 Dario Amodei 对 AI 发展的速度和安全性的很多独到见解。感觉这些看法和上个月见到 Anthropic AI Safety 产品负责任人了解到的基本一致,看来他们公司的价值观对齐的非常好,全员 70% Doomer 😅 Anthropic 并不爱提 AGI…
account_circle
Bindu Reddy(@bindureddy) 's Twitter Profile Photo

The Path To AGI

OpenAI has claimed before that Sora is the path to AGI. Many, including me, have been skeptical.

I now see a path to how that can happen - Let's assume that the definition of AGI is an AI capable of most tasks humans can do. For AI to match human abilities,…

The Path To AGI OpenAI has claimed before that Sora is the path to AGI. Many, including me, have been skeptical. I now see a path to how that can happen - Let's assume that the definition of AGI is an AI capable of most tasks humans can do. For AI to match human abilities,…
account_circle
宝玉(@dotey) 's Twitter Profile Photo

福布斯采访 Notion 创始人:用“乐高”式创新挑战 Google Docs 和 Microsoft Office

Notion 是一款由人工智能驱动的生产力工具,正在向 Google Docs 和 Microsoft Office 发起挑战。该公司在 2024 年的《福布斯》AI 50 强榜单中脱颖而出,其流行程度的快速上升也得到了社交媒体上大量 Notion…

account_circle
ハルオミ(@HaruAiman) 's Twitter Profile Photo

100%AIで飲食店の広告いけるかも!

本来はカメラマンがお店で撮影する必要があるが、AIなら5分で作れる!!

Midjourney > Magnific AI > Runway > Topaz Video AI

account_circle
歸藏(guizang.ai)(@op7418) 's Twitter Profile Photo

莱森新教程,利用Magnific的风格转换功能和Tripoai,快速生成同一张图片不同风格的3D模型。

account_circle
Leo Xiang(@leeoxiang) 's Twitter Profile Photo

最近写爬虫爬取 Twitter 的数据,研究了一下世面上的大多数方案,大家感兴趣的话我分享一下我测试过的 数据爬虫产品。

account_circle
小互(@imxiaohu) 's Twitter Profile Photo

Parler-TTS :一个完全开源的的高质量TTS模型

它能够生成高质量且听起来非常自然的语音。还可以定制说话者的风格(如性别、音调、说话风格等)。

与其他TTS模型不同,Parler-TTS 完全开源发布,包括数据集、预处理、训练代码和权重。

只需一行代码即可安装。…

account_circle