很多人觉得美国的官方是在发神经,港机会有啥安全问题呢?实际上,如果你听说过挖掘机指数,以及供应链安全的话,就知道啥意思了1/n
Officials Chinese-made cranes at U.S. ports, including at several ports used by the military, could pose a security risk wsj.com/articles/penta… via @WSJ
首先振华港机的人这么说“Hailiang Song, ZPMC’s then-chairman, in a 2017 video on Microsoft’s website. In the video, then-President Qingfeng Huang added: “Through our main office in Shanghai, you can monitor all the cranes” 。对可以监控所有他们家卖出去港机的运行情况。2/n
This is a good question, many people confused finetune vs index. Fine tune is train the model, just make the model smarter, but the tuning process is complicate, expensive. Index, you can think this is some kind cheating, you provide data and slice data in chunks1/n#
Using openai embedding generate vectors, these vectors will store in your vector db. These are extra knowledge. When you ask openai , it will vectorize your questions, it will use search your vector DB , find the closest entry and context, then send this information to openAI 2/n
For example, you can ask openai summarize this closest context (search result from vector db), then return to user. It will behavior like openai have the latest knowledge. But this output like an chatgpt addon. The quality is not so nature, but the still acceptable. 3/n