Do you think Artificial General Intelligence (AGI) is acheivable?

I have a hard time believing it's possible. The problem with current methods and training is that the models fit the data. This creates a system that is good at a very specific classification task.
As far as I am aware, we do not have a machine learning methods that can take what it has learned about, say, the game go, and apply that to chess. You have to retrain the model to handle chess.
Granted, we as humans around similar to a degree, but we can take the learning and apply it over.
As I see it, AGI would almost require a system that can rewrite is own coffee in order to adapt to a change in the environment and thus a change in requirements. Even hot swapping code modules as needed.
Linux has proven that you can hot swap kernel modules while it is running, so that part is not unfeasible. But a program being able to identify the weaknesses in its own programming seems like an impossible task.
The human brain is an incredibly complex organ. A lot of people like to compare a computer cup with the brain, but the more I find out about programming and the Brain, the more I realize that the similarities are slim. Intelligence on the level of human cognition is not something I think programmers and scientists will be able to achieve.
But what do you think? What additional information or insight do you have on this topic?

#agi
#artifical_intelligence
#artificialintelligence
#qoto
#whatdoyouthink