How to build an ai like Neuro,or being better than her.

First step of building an strong ai,is to make a deep-learning network which can connect each other free.

How to solve it?

First,we should think about what's connecting.In my opinion,if you can know all its output,and your input to it is efficent for it,you are connected to it.

So, how to solve it?

Maybe we need a network which can process a input which is complex enough to carry as diverse as real life.Yes, it's llm.

This llm need to be able to process like:

{
    "model":"math",
    "info":{
        "example":"vector"
    }
}

or maybe more complex.Each model may be the same but it's not necessary to accept other's input.

So,maybe I can build a common model,but how does it remenber?

If it's always trained,can it remenber?Or how do I remenber?

And if it can accept long text?