NLP sequence can completely simulate human brain intelligence

As shown in the figure, it is necessary to repeat the process in the figure continuously to realize the agent and have the memory ability
The memory can not be too long, so just change the memory end to one sum Summary input or keyword extraction input
To solve the problem of long memory , Control both the summary and the latest non summary as input
The most recent sentence will prevail , Past sentences for reference , Perform logic output
In this way, the memory of an agent and the logic of an agent are formed
, Then the input and output of the agent can be transformed into words to realize vision, hearing and smell
And all the perceptual and logical modeling
In this way, we can complete the first real artificial intelligence in human history
The basic component of this agent is sequence to sequence . You can use any sequence to sequence model

Of course, the smaller the basic sequence to sequence model, the less computational effort, the better .
From this moment, artificial intelligence has completely reached the advanced stage
Artificial intelligence , Will no longer be called artificial mental retardation
As long as you keep the fine-tuning state of the model, it will change constantly
Of course, you can fine tune the gradient accumulation, which will be in a certain period of time
Maintain previous memory transition and logic processing ability
Until the variable value is reached , Will make some changes
As far as the current big model is concerned, they will always be a part of this agent
Not smart at all , The reason is very simple. The training method is completely inconsistent with the operation process