Autonomous AI Completes Multi-Step Biological Experiment

A Chinese scientific team has reported the completion of a long-duration, AI-driven experiment involving 11,382 individual steps over a period of 203 hours. According to reports circulating on social media platforms like Reddit, the experiment was reportedly conducted in a fully automated biological laboratory.

While some unverified discussions suggest the process was managed by Gemini 2.5, the specific scientific utility of the 11,382-step sequence has not been independently confirmed in the provided summaries. The speed of the experiment highlights a growing tension in the scientific community regarding publication timelines; one commentator noted that a preprint from July 2025 only recently received peer-reviewed publication in Nature, suggesting that traditional publishing cycles may struggle to keep pace with the velocity of AI-driven research.

Infrastructure and Energy Scaling in China

As AI research moves toward autonomous laboratory environments, the energy requirements for the underlying computational power are driving significant infrastructure shifts. China has deployed what it describes as the world’s first 800-volt DC power system for a massive AI data center. This high-voltage system is designed to increase conversion efficiency and reduce the physical footprint of the facility.

This move toward an “electrostate” model—characterized by a modernized, high-capacity grid—is being viewed by some analysts as a foundational advantage for China’s AI ambitions. While U.S. efforts have focused on talent and investment restrictions, China has continued to expand its domestic ecosystem. This includes the deployment of large-scale AI science hubs and the use of specialized hardware, such as Baidu’s P800 Kunlun chip clusters, which are being utilized to train large-scale models.

Competitive Landscape in AI Development

The gap between Chinese and Western AI capabilities continues to fluctuate. While some reports suggest that companies like DeepSeek have narrowed the gap with U.S. models to within three months, other developments show significant divergence in specialized applications:

  • Hardware and Chips: Despite U.S. restrictions on talent and technology, Chinese firms are developing domestic alternatives. Huawei has introduced the CloudMatrix 384 server as an alternative to Nvidia’s infrastructure, and researchers have released QiMeng, an AI-based automated processor chip design system.
  • Specialized Models: Development is moving into highly specific sectors, including solar observation, maritime forecasting (OceanDS), and plant functional genomics (PlantGPT). In the medical field, researchers have developed an AI-based screening system for 25 ocular fundus diseases.
  • Efficiency and Cost: Some industry reports indicate significant cost advantages in the region; for example, MiniMax claims its M1 model costs approximately 200 times less to train than OpenAI’s GPT-4.

Global Market and Regulatory Divergence

The rapid expansion of AI infrastructure is facing different regulatory pressures in different jurisdictions. In the United States, political discussions have emerged regarding the requirement for data centers to fund their own energy demands and the potential creation of an “FDA-style” agency to vet frontier AI models before deployment.

Meanwhile, China is integrating AI across various sectors, from courtroom applications to large-scale industrial fermentation control. This expansion is supported by significant state-level investment, including a reported $295 billion plan to fund a nationwide AI buildout. As the scale of deployment grows, the competition between open-source models, such as Alibaba’s Qwen3, and proprietary systems remains a central focus for global technology markets.