WONG IN-HOON TALKS ABOUT OPEN AND CLOSED SOURCE AI MODELS: THE TWO WILL COEXIST FOR A LONG TIME, AND CLOSED SOURCE MODELS ARE NOT NATURAL AND SAFER

On 26 July, in an interview with Bloomberg podcasts, In-hoon Ying Weida Wong stated that open models and closed source models would coexist in the future for a long time, serving different needs. Closed-source models are often more user-friendly and less costly, but enterprises need to build their own AI based on open-source models if they need to control data, professional capabilities, service levels or national sovereignty-related capabilities. According to Huang In-hoon, open models are essential for security, cybersecurity, innovation, start-ups and technological sovereignty. Closed-source models are not naturally safer by virtue of their proprietary properties and may still be subject to escape, theft or internal leakage; if the model's shield or sandbox is not well designed, they may also cause attacks on other corporate systems. AI The industry cannot rely on a single defensive node, but should build a large-scale distributed self-defence capability. It would be wrong to interpret AI simply as an open or closed source, and to argue that open models are necessarily unsafe is equally ill-founded. Hwang In-hoon, citing as an example the security incident at Hugging Face, understood that when the proprietary model was unable to assist in the sorting of the problem, Hugging Face used the open model GM 5.2 to identify the holes and the location of the invasion and quickly complete the repairs. In his view, the event demonstrated the ability of open models to provide the necessary self-defence capabilities and the importance of AI technological diversity。