Search AI Products and News

Bloomberg, citing sources, reports that Ant Group has made a significant breakthrough in AI, reducing AI model training costs by 20% using chips from Alibaba and Huawei. This injects new momentum into China's drive for technological self-reliance. The report further indicates that Ant Group's internal tests show these Chinese-made AI chips rival Nvidia's offerings in performance. Widespread validation and adoption of this could significantly alter the global AI landscape.

Ant Group's Ling team recently published a preprint on arXiv titled "Every FLOP Matters: Scaling a 300-billion parameter Mixture-of-Experts LING model without high-end GPUs," detailing two novel large language models: Ling-Lite and Ling-Plus. These models incorporate several innovations enabling efficient training on low-performance hardware, significantly reducing training costs.

HPC-AI Tech recently announced Open-Sora 2.0, a groundbreaking video AI system achieving commercial-grade quality at approximately one-tenth the typical training cost. This represents a potential paradigm shift in the resource-intensive field of video AI, mirroring efficiency gains seen in language models. While existing high-quality video generation systems like Movie Gen and Step-Video-T2V can cost millions of dollars to train, Open-Sora 2.0 offers a significantly more affordable alternative.

Recently, Musk officially launched the brand new chatbot Grok 3 in a live stream and disclosed the astonishing cost of training the model. Grok 3 is currently available to Premium + subscribers and has shown outstanding performance in various evaluation areas, even surpassing other competitors in the market, such as Gemini, DeepSeek, and ChatGPT. During the live stream, Musk mentioned that the training process for Grok 3 consumed a total of 200,000 NVIDIA GPUs, a figure that is truly shocking.

In today's rapid development of Large Language Models (LLM), the training and inference costs of models have increasingly become a focus of research and application. Recently, the Tencent Hunyuan team released an important study that delves into the 'Scaling Laws' of low-bit floating point quantization training, which refers to the principles governing the scale of floating point quantization training. The core of this research lies in exploring how to significantly reduce computational and storage costs without sacrificing performance by lowering the precision of the model.

According to the CEO of Anthropic, the current cost of training AI models is as high as $10 billion, and it could rise to $100 billion or even $1 trillion in the next three years. This forecast has sparked concerns about whether the AI bubble is about to burst.Image Source Note: The image was generated by AI, provided by the image authorization service provider Midjourney Dario Amodei points out that as AI models continue to evolve, hardware demand will also grow exponentially, becoming a major

Shanghai AI Lab has released the large model training toolbox XTuner, which supports various hardware adaptations, significantly lowering the training cost barrier. XTuner is compatible with multiple open-source large models and can execute tasks such as incremental pre-training and instruction fine-tuning. XTuner balances usability and configurability with a built-in standardized process for one-click training.

Hugging Face has launched a new AI service called 'Training Cluster as a Service' that allows companies to train AI models without needing extensive computing resources. Users can utilize a pricing calculator to configure the required model based on parameters, features, training data volume, and speed, with a wide range of pricing. Despite the availability of this service, top AI models like GPT-4 or Google's Gemini remain significantly expensive and powerful. The costs of training AI may increase with advancements in processor technology.