How to make real AGI

To get a talking AI, you need to apply question-answer patterns to the forecasting algorithm. I invented this principle in 2015, one of the first in the world, so to speak. And this invention was not a shot in the dark, but the result of several years of thinking.

The prediction algorithm is simply a classifier. And the neural networks that are currently used are simply a continuous classifier, not a discrete one, as most people imagine.

The world exists according to rules, regardless of whether we know these rules or not. And these rules are innumerable. And classification-forecasting algorithms are innumerable.

Any forecasting algorithm can forecast, but not every one of them is effective for the set of rules of interest. We can say that any of the forecasting-classification algorithms has its own area of ​​effectiveness. And there is no universal monolithic and effective algorithm for all cases, so that it would be effective for everything. And a step to the left-right and there will already be a slightly different algorithm effective.

Everything written above also applies to neural networks, transformers or other types of neural networks.

And the solution for real AGI will be a complex of algorithms. A complex that will be sufficient for it to be able to continue inventing and embedding new pieces of forecasting algorithms.

And as a result, such an AGI will be able to do everything. Maybe not right away, but after a little thought and development. And of course it will not be able to do what is not feasible. Let's say a time machine that can be sent into the past is not feasible. But such an algorithm will be able to arrange flights to the stars, or eternal life.

In general, I have a vision of such a scheme. But no one who has the resources to calculate it believes me. They simply ignore all appeals to them. And everyone around me in every way, including physically, prevents me from implementing it on my own. This is the reality of Russian existence.