At the CCS2024 cybersecurity event series held in Chengdu, Baidu Vice President Chen Yang shared how large model technology can revolutionize software development. He pointed out that the powerful understanding, generation, logic and memory capabilities of large models can significantly improve development efficiency and enhance software security. Baidu's independently developed intelligent coding tool "Wenxin Quick Code" has been widely used internally, providing strong support for Baidu's 30% code contribution.
At the CCS2024 cybersecurity event series held in Chengdu, Baidu Vice President Chen Yang delivered a keynote speech and shared how large model technology promotes changes in the field of software development. Chen Yang pointed out that the combination of large model understanding, generation, logic and memory capabilities with software development can significantly improve development efficiency and enhance software security.
Baidu has been exploring the field of intelligent coding for two years. The intelligent coding efficiency improvement tool "Wenxin Quick Code" developed by Baidu has been widely used within the company, significantly improving the R&D efficiency of engineers. At present, 30% of the codes generated by Baidu every day are produced by Wenxin Kuaicode, the overall adoption rate has reached 46%, and the efficiency improvement of engineers has reached 12%. This not only increases the number of code submissions and the speed of business iterations, It also enables engineers to focus on more valuable and creative work.
Chen Yang emphasized that while Wenxin Kuaimao improves research and development efficiency, it also attaches great importance to project quality. This tool incorporates quality and security considerations into the entire development process, including requirement clarification, security design, module sorting, code explanation, security vulnerability scanning and repair, etc., ensuring the security and quality of the code. In Baidu's internal practice, Wenxin Quick Code's security vulnerability scanning accuracy has exceeded 95%, and 83% of the scanning vulnerabilities have been repaired.
Although large model technology has shown significant value and practical effects in the field of intelligent R&D, it still faces difficulties and challenges such as the application of private domain knowledge, differences in industry quality standards, cooperation with engineering culture, and value measurement during the implementation process. Through Wenxin Kuaima's internal and external experience, Baidu has summarized a set of standardized implementation processes and best practices to promote the transformation of corporate engineering culture and the construction of a data-driven value closed loop.
All in all, the successful application of Baidu's "Wenxin Quick Code" has provided valuable experience for the implementation of large model technology in the field of software development, and also provided best practices for other companies to learn from. In the future, as technology continues to mature and applications continue to expand, large model technology will play a greater role in the field of software development.