Recently, foreign media reported that ByteDance is accelerating its efforts to self-develop and manufacture AI chips to reduce dependence on expensive chips from NVIDIA. The report claims that ByteDance plans to collaborate with TSMC and expects to mass-produce two self-developed AI chips by 2026, having already reserved the production of hundreds of thousands of chips. More notably, the report pointed out that ByteDance aims for its self-developed chip clusters to achieve four times the computing performance of a single H100 chip at a similar cost. In response, ByteDance quickly addressed this report.

Recently, a research team from Shanghai Jiao Tong University and Harvard University introduced a novel model fine-tuning method — LoRA-Dash. This new approach claims to be more efficient than existing LoRA methods, particularly in the fine-tuning of specific tasks, achieving the same results with an 8 to 16 times reduction in the number of parameters. This is undoubtedly a major breakthrough for fine-tuning tasks that require substantial computational resources. In the context of the rapid development of large-scale language models, the demand for fine-tuning specific tasks is steadily increasing. However, fine-tuning often

Fluxgym is a new training tool aimed at AI enthusiasts, designed to lower the barriers to model training. It has extremely low hardware requirements, needing only 12GB of VRAM to conduct AI model training, significantly expanding the participant base. By combining a simple front-end with a powerful back-end, Fluxgym provides an easy-to-use interface, allowing users to initiate the training process in just three simple steps. Its core advantages lie in support for low-spec devices, a simplified user interface, and a smooth training process, enabling AI newcomers to get started quickly.

FLUX.1-dev-LoRA is an innovative image generation model that specializes in merging elements of reality with illustrations to create unique visual effects. Trained based on FLUX.1-dev, it can depict foreground characters in an illustrated style while maintaining realism in the background, making it suitable for generating creative images on social platforms. Users can generate images that meet their needs by providing scene descriptions and character action prompts. Official prompt examples demonstrate how to describe for specific art style fusion. Online experience and model download links are provided.

ByteDance's open-source FLUX Dev's Hyper SD Lora has revolutionized AI drawing speed with its efficient image generation capabilities, shortening the traditional, lengthy process to just 8 or 16 steps. This move not only provides a fast track for creative realization but also establishes a demonstration space on Hugging Face, allowing users to directly experience the drawing effects of the 8-step Lora. Although there is some degradation in image quality, it has minimal impact for most users. Comparison examples show that the degradation is not significant when not enlarged.

The LoRA model Half_illustration based on Flux.1 is an innovative image processing tool that can transform ordinary photos into artistic works that blend photography and illustration elements. Users can simply describe a scene or add specific instructions (such as 'in the style of TOK'), and the model can generate visually striking pieces that are both real and dreamlike. This tool supports a wide range of creative scenarios, including fashion photography in specific styles, artistic creations, advertising design, and more.

"The 'Online FLUX' by fal.ai offers a revolutionary experience for drawing enthusiasts, introducing ControlNet and LoRA features that are ready to use out of the box without complex configurations. ControlNet simplifies modifying characters' expressions or controlling scene objects; the LoRA model supports online training, allowing users to customize and share models, promoting creative diffusion. This platform integrates with third-party functionalities, making drawing more flexible and personalized, supporting various styles from cyberpunk to Chinese aesthetics. The usability of the Online FLUX is profound."

On August 7th, Meitu's Meitu Cloud Repair launched the newly upgraded 'Meitu Cloud Repair Pro Version', introducing new features such as AI batch coloring and AI batch retouching, aimed at providing a more comprehensive photo editing solution for the commercial photography industry. The AI workflow of Meitu Cloud Repair can help retouchers achieve AI automation from format conversion, retouching to delivery, significantly improving work efficiency.

Tencent's Hunyuan DiT model has announced the full open-sourcing of its training code, including LoRA and ControlNet plugins. LoRA is a technique for fine-tuning large language models that allows training models with specific characteristics using a small amount of data, without modifying the original model or increasing its size.

The Pentagon AI Task Force is set to launch a digital sandbox to help the defense community experiment with various AI tools. Task Force Lima will roll out a transition plan to advance the safe adoption of generative AI. The task force will educate personnel on how to use AI tools, emphasizing the importance of experimentation and evaluation. The task force aims to provide access to and education about AI technologies through the sandbox, fostering technological development.

Google Maps has introduced generative artificial intelligence, allowing users to search for specific types of atmospheres in urban locations. By analyzing map information with AI models, it provides suggestions that meet user requirements. Generative AI is currently being rolled out to local guides in specific regions of the United States, with plans to expand coverage after testing.

NASA is using AI technology to advance the Artemis moon mission with the VIPER rover exploring the lunar surface. SHERPA acts as the AI brain behind the VIPER mission, applying time-constrained planning to the mission. The VIPER mission is pushing the boundaries of space exploration and paving the way for humanity's return to the Moon.

Output has launched Co-Producer, enhancing human creativity in music production with AI technology. The Pack Generator is the first tool that creates royalty-free sample packs through text prompts. Unlike other AI music tools, it is powered by a blend of generative AI and audio samples from real musicians. The Pack Generator helps music producers quickly find new creative inspiration. Output plans to release more Co-Producer AI music creation tools in 2024.

Deep Exploration has released a large model of 67B, named DeepSeekLLM67B. This model is fully open-source and can be accessed at chat.deepseek.com. DeepSeekLLM67B has performed well on public evaluation leaderboards, exhibiting excellent reasoning, mathematical, and programming capabilities. DeepSeek has also open-sourced models in two sizes, 7B and 67B, and provided 9 model checkpoints from mid-training.

The Tied-Lora technology developed by NVIDIA aims to enhance the parameter efficiency of LoRA methods. By utilizing weight binding and selective training, it achieves an optimal balance between performance and trainable parameters. Experiments on different tasks and language models demonstrate that Tied-Lora can achieve comparable performance using only 13% of the parameters of the standard LoRA method. This technology significantly improves the parameter efficiency of the LoRA methods and achieves outstanding performance across various tasks.

🎬 Innovative AI is emerging in the Indian film industry, emphasizing ethical creation. 🤖 The Chief Data Officer of Zee Entertainment discusses the role of AI, highlighting the importance of human creativity. 📽️ Generative AI is transforming the way we approach intellectual property, with digital avatars' rights paving the way for the future.

The current mainstream alignment research directions are precisely designed to create better products. Simplified feedback models relying on human preferences only address surface-level issues. Preventing disaster requires serious involvement in ethical discussions regarding the application of artificial intelligence. Addressing technical issues requires attention to social context and governance structures. Public opinion is very important but cannot rely on those who profit from deployment.

Alibaba has launched the Quark Scanner King App based on AI large models, integrating various efficient scanning and editing functionalities. The Quark Scanner King App achieves precise and efficient text and image recognition through AI large models. This app meets users' needs for scanning, editing, and sharing in various scenarios, enhancing work and study efficiency.

Virtualitics is a startup that offers an enterprise AI-driven 3D data exploration platform, which has just announced the completion of a $37 million Series C funding round. Virtualitics' platform utilizes artificial intelligence and machine learning models to analyze data and presents the visualization results in 3D, helping users better understand complex data. Virtualitics plans to use this funding to expand its business, add more features, and enable users to analyze and understand complex numbers.