Spend 1 minute every day to get the latest AI information.

The content includes but is not limited to cutting-edge AI news, AI tools, AI painting, open-source projects, and learning tutorials.

The main feature of the brief is its concise description, but important information is still introduced in detail through independent posts.

Here is the latest AI information for July 3.

Cutting-edge News

1. Meta AI has launched a new, leading, and fast text-to-3D asset generation process: Meta 3D Gen.

It can create high-quality 3D models and textures in less than a minute, supporting physically-based rendering and exceeding industry benchmarks in visual quality and prompt fidelity under complex text prompts.

Detailed introduction: https://ai.meta.com/research/publications/meta-3d-gen/

2. Coze has ended its free strategy!

Started charging, with prices set quite high, likely only subscribed by users with specific needs. The domestic version of Coze (扣子) is still free for now.

Official website: https://www.coze.com/

3. Suno has launched an iOS client.

Supports music creation and playback features, and allows you to create your own playlists. It needs to be downloaded from the US App Store.

Download link: https://apps.apple.com/us/app/suno-make-and-explore-music/id6480136315

Direct searches may yield many knock-off apps, so look for the official logo or use the above link to download directly.

Cutting-edge Technology

1. Powerful RAG technology is here! Microsoft has open-sourced GraphRAG.

Unlike traditional RAG methods, it provides significant QA performance improvements by using knowledge graphs, establishing community hierarchies, generating community summaries, and leveraging these structures when performing RAG tasks.

Detailed introduction: https://microsoft.github.io/graphrag/

GitHub: https://github.com/microsoft/graphrag

In simple terms, it is a retrieval-augmented generation technique based on knowledge graphs, effectively improving the inference ability of large language models in handling datasets.

AI Tools

1. Another open-source TTS model discovered: Fish Speech.

Trained for 300,000 hours, it is proficient in Chinese, Japanese, and English, with rich voice performance. The Chinese effect is particularly good, and it can handle mixed Chinese and English.

Detailed introduction: https://t.zsxq.com/ZUveg

Learning Books

1. A tutorial on large-scale pre-trained language models on GitHub: "Fundamentals of Large Models".

The tutorial is based on courses from Stanford University and Li Hongyi, supplemented by contributions from the open-source community and the latest research progress. It aims to provide readers with in-depth theoretical knowledge and practical methods.

Detailed introduction: https://t.zsxq.com/z3ULt

See the directory image below for more details. Those interested in learning about the basics of large models can check it out.