/sci/ - Science & Math

>>10506652
There will be no AI that mimics human intelligence if the most basic mechanism of the brain is not understood. The thing about great ideas is that they can be simplified. Before you understand it, some questions may just seem infinitely complicated.

Although I made simplifying assumptions, that does not mean systems emerged from this basic mechanism cannot grow extremely abstract and complex when observed on a larger scale.

The diagrams seemed simple because I organized them that way; I sorted related information together and put them in little bubbles for ease of understanding. Imagine how the uncountable neurons in just one of those little bubbles might be located separately in the biological brain, and all their connections that are fired accross the brain when activated. And that will be combined with the activation of another cluster, until eventually another collective activation is emerged. During this process, the property of the neurons and their connections also change accordingly. It's not easy to even imagine it visually. It only seemed simple because I wanted it to. And I should indeed consider advanced mechanisms that further optimize the structure, such as the brain's ability to utilize reward prediction error with the use of different types of neurons. Regardless, the design language itself is viable.