Spend just 1 minute daily to get comprehensive updates on AI technology developments, industry trends, and market insights.

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

Stay tuned to AI Daily for the latest in AI trends. For important information, detailed posts will be made separately.

Here is the latest AI information for July 17.

Cutting-edge News

1. Claude has launched an Android client.

Similar to the iOS and Web versions, it supports features such as vision, multilingual processing, and advanced reasoning.

Download link: https://play.google.com/store/apps/details?id=com.anthropic.claude

2. Former OpenAI co-founder Andrej Karpathy announced the establishment of Eureka Labs, a new AI + education platform.

The first course LLM101n has already been released on GitHub, providing a step-by-step guide to creating LLMs. It currently only has a directory but has already garnered 18.3k stars.

Official website: https://eurekalabs.ai/

3. Mistral released a math model called MathΣtral.

A specific 7B model designed for mathematical reasoning and scientific discovery, with a 32k context length, open-sourced under the Apache 2.0 license and available for commercial use.

In multiple benchmarks, it outperforms similar models such as DeepSeek Math 7B, QWen2 7B, and Llama3 8B.

Detailed introduction: https://mistral.ai/news/mathstral/

Model download: https://huggingface.co/mistralai/mathstral-7B-v0.1

Cutting-edge Technology

1. Another small language model that can run on mobile phones, SmolLM.

Available in 135M, 360M, and 1.7B parameter sizes, it outperforms other similar models in multiple benchmarks.

Detailed introduction: https://huggingface.co/blog/smollm

2. Alibaba releases the latest audio language model Qwen-Audio.

It can accept various audio signal inputs and respond with audio analysis and direct text according to voice commands. It offers two different audio interaction modes:

• Voice chat: Allows free voice interaction with Qwen2-Audio without text input.
• Audio analysis: Allows providing audio and text instructions for audio analysis during interaction.

Currently, only the paper is released, but the model download and online experience will be available soon.

Detailed introduction: https://github.com/QwenLM/Qwen2-Audio/blob/main/README_CN.md

Open-source Projects

1. A local model adaptation open-source project based on GraphRAG: GraphRAG-Ollama-UI.

Adapts Microsoft's GraphRAG locally, integrates with Ollama, and provides a visual interactive interface.

Supports file management such as upload, edit, and delete, real-time visual knowledge graph, supports using local models, and offers one-click deployment with Docker.

GitHub: https://github.com/severian42/GraphRAG-Ollama-UI

Learning Tutorials

1. Implementing Stable Diffusion from scratch.

Found a project on GitHub where the author systematically explains the working principles of Stable Diffusion and diffusion models and the mathematics behind them.

GitHub: https://github.com/juraam/stable-diffusion-from-scratch

The project also provides a series of steps to help understand this information, culminating in the training of a diffusion model.