**: Flying Elephant Network.
With the rapid development of AI technology, large models have become a key driving force to drive AI forward. However, for enterprises, how to ensure data security and privacy protection has become an issue that cannot be ignored.
In this regard, the technical person in charge of a retail group said that the AI model they are experimenting with is still limited to a single store, and they dare not feed more data to the large model for security and permission considerations. This cautious attitude of enterprises is based on a deep understanding of data security and privacy protection.
In the process of promoting the application of AI large models, Orion Star deeply understands the importance of data security and privacy protection for enterprises. Some content should not be produced using a large public model such as ChatGPT, because once the secret is leaked, it will threaten the company's security.
For example, when users want to use AI models, they need to input their knowledge and data to the large models first. In this way, the user will get help from the big model, and the big model will get smarter and smarter. However, in the case of a large public model like ChatGPT, the user's own data will also be used to help competitors who also use this model.
In fact, more and more people in the industry have realized that since enterprise-level applications often involve professional scenarios and multi-person collaborative work, they pay more attention to collaborative office scenarios and the private nature of data. For those enterprises with particularly sensitive and high-value data, there is a greater need for customized or privatized AI models. As Wang Hengjiang of China Mobile pointed out, "A lot of our data must be privatized and personalized, and cannot be disclosed, which is certain, especially on the end side, there is a lot of privacy data in our mobile phones and personal home hardware." ”
For those companies that are engaged in engineering software, they are also absolutely reluctant to provide information to large models in the public domain. The same applies to companies such as semiconductors, as well as to private data in the medical field and automakers with large amounts of BOM data on auto parts. Because the data of these companies is highly valuable.
Therefore, in many fields, relying only on general large-scale models may not be able to achieve implementation, especially in enterprise security, finance, and government affairs. If these companies and organizations want to ride the AI bandwagon, they can only build private models.
Fu Sheng pointed out that Open AI has read almost all publicly published data on the Internet today. But no matter how vast the Internet is, it is also the tip of the iceberg of human knowledge system. Every company's documentation has its own competitiveness and characteristics. He emphasized, "If enterprises want to be competitive, they should use the privatization model, let the business data circulate internally, and let the experience that the enterprise has left in everyone's mind in the past become part of the overall decision-making intelligence." ”
In any case, the privatization of enterprise large models is an inevitable trend in the development of AI decision-making intelligence. With the increasing demand for data security and privacy protection and the potential risks of public models, enterprises will be more inclined to build private AI models. Only in this way can enterprises remain competitive and achieve sustainable development in the AI tide.