Author | Liu Baodan
Editor | Huang Yu
The rapid iteration of AI large model technology has raised higher requirements for AI governance, and Alibaba has begun to think more about how to effectively manage AI governance.
Recently, at the seminar on the development of large model technology and governance innovation, Qian Lei, Vice President of Alibaba Group, stated that in order to effectively govern large models, improve their controllability, and reduce application risks, Alibaba has comprehensively deployed security measures throughout the entire lifecycle of large model research and application.
First, enhance the intrinsic security capabilities of the model. Emphasize source control during the data collection and processing stages, conducting strict data screening and data security labeling; add security corpus during the model training and design stages; and conduct a comprehensive security assessment before the model goes live to ensure the qualification rate of various risk items in generated content.
Second, strengthen the external protective barrier capabilities. After the model is deployed and applied, further enhance the safety of the large model in application scenarios through external protective barriers, and verify through daily inspections whether the results meet expectations and whether reliability, trustworthiness, and usability meet Alibaba's requirements.
Alibaba also jointly released the &#34;Large Model Technology Development and Governance Practice Report&#34; with the China Electronic Technology Standardization Institute. This marks the third consecutive year that Alibaba has published research results related to AI governance, further deepening Alibaba's experience summary in the fields of technology research and governance.
Yuan Yuan, Director of Alibaba Research Institute, believes that the governance approach for AI large models is to analyze the three core development engines in the development of large models—computing power, algorithms, and data—and provide a development path suitable for China. Taking data as an example, under the premise of conforming to the characteristics of the real world, synthetic data can help models break through current capability limits and reduce more security concerns.
Since the end of 2022, when ChatGPT gained global attention, AI large models have become a track that technology giants are competing to invest in, with Alibaba being a core participant.
In terms of AI large models, Alibaba insists on the coordinated development of cloud and AI, evolving cloud from IT and DT infrastructure to artificial intelligence infrastructure. According to Qian Lei, Alibaba will continue to increase investment, first to enhance the capabilities of its self-developed foundational model Tongyi series, and second to reduce the cost of model application development by improving AI infrastructure and related support systems.
The API call price for &#34;Tongyi Qianwen&#34; has decreased by 97% over the past year. At the same time, Alibaba insists on full-size, full-modal open-source, promoting the transformation of AI capabilities into productivity across various industries. Currently, over 300,000 enterprises have accessed Alibaba's Tongyi Qianwen large model.
Data from the Hugging Face community shows that the number of derivative models developed based on the &#34;Tongyi&#34; model has exceeded 80,000, attracting over 6 million AI developers to participate, with downloads on global open-source platforms surpassing 100 million.
In Alibaba's view, while advancing cutting-edge technologies of large models to promote AI applications, it is also crucial to pay high attention to the safety of artificial intelligence and responsibly address the security issues brought by large models
However, AI large models are still in rapid iteration, and the key to solving safety and governance issues lies in balancing development and risk, much like changing parts on a highway, which requires high intelligence and expertise.
Greater challenges lie ahead. As AI large models truly enter people's lives, issues regarding artificial intelligence safety and governance will expose even more problems that all AI-related companies must face.
No technology has ever matched human intelligence like AI does, but this also means that once AI is misused, its destructiveness is equally concerning. The path of AI governance has just begun, and Alibaba has already attempted to provide its own answers.
Risk Warning and Disclaimer
The market has risks, and investment requires caution. This article does not constitute personal investment advice and does not take into account the specific investment goals, financial situation, or needs of individual users. Users should consider whether any opinions, views, or conclusions in this article align with their specific circumstances. Investing based on this is at one's own risk

BABA

89988

09988

Alibaba proposes a new path in AI governance. Vice President Qian Lei stated at a seminar that Alibaba has fully deployed security measures to enhance the controllability of large models and reduce application risks. The measures include improving the intrinsic security capabilities of the models and strengthening the peripheral protection capabilities. In addition, Alibaba, in collaboration with the China Electronics Standardization Institute, released the "Large Model Technology Development and Governance Practice Report," marking the third consecutive year of publishing related research findings. Alibaba will continue to increase investment to reduce the development costs of model applications

- Alibaba is enhancing its AI governance in response to rapid advancements in large model technology, focusing on safety and risk reduction throughout the model lifecycle.  
- Vice President Qian Lei highlighted measures such as improving internal security capabilities and strengthening external safeguards to ensure reliable model deployment.  
- The company emphasizes the importance of balancing development and risk management in AI, as it continues to invest in its AI infrastructure and expand its model's accessibility to over 300,000 enterprises.

Alibaba has proposed a governance path for AI