发自我的 iPhone

以下是转发的邮件：

发件人: deli wu <[email protected]>
日期: 2025年12月10日 GMT+13 下午1:22:17
收件人: deli wu <[email protected]>
主题: The Developed AI Model?

 The Developed AI Model? 
 
The thinking is so much expensive! For the following things, I really carefully was thinking about this issue. Finally I decided to make the issue. Because probably this issue really can change our world so, so fast indeed!! (I think that this is my responsibility for really loving our world-the whole earth!) 
 
Just three days ago I talked to an AI’s model about current very hot topics: The AI large model’s key core things. The answers should be 100 percent correct, But the answers are really surprised me! that’s why I’d like to share these with you, that’s so important! 
 
Surprised me but I was wondering if you could not be surprised? If then I would still need to study hard and continue to take my responsibility for our world as a common person, because this is everybody’ responsibility who lives on this earth. 
 
For the simple to understand, I will make a summary for you here, that I talked to the AI large model results that are as followings: 

1For current AI large models, are all not a cumulative memory model! 
 
2For model’s training stage, the model have memory and can create references; But for using stage, the models have no any memory and will forget the result of whatever processed model for next time using .
  
3Why are the AI large models doing things like that?  answers are: they can have memory and cumulative memory for using results for next time. But why the AI can’t do that for now? Because our people really worry the IA that can destroys our world at the cumulative memory stage. 
 
4If then, I asked very dare question for IA is:  
 
Could we first enable AI to generate cumulative memory, and then, after such a large AI model has solved the medicine research and development for treating various major diseases in the world, then we can design the large AI model back to its current AI state? Can you guess AI’s answer please?

(If you want, i can copy all the questions and answers from the ai model platform for you in Chinese) December 12, 2025