Despite the neural network architecture, LLMs don't operate like the human brain. They're just fancy curve fitting algorithms. They need to reason and understand things to help us discover new things. But how do you program understanding? How do human brains even understand?
Discussion
The human brain is far more complex than these generative models based on transformer architecture. I mean, at its core, it’s just doing basic matrix multiplication. The human brain is doing something way more extraordinary.
We don’t even know if that function lies in a general physical world, or the quantum realm.
