I think this is related to the idea that intelligence is compression. But when we think of compression we immediately run into a conundrum: If something is compressible, it means the language used to express the piece of information is not optimal. An optimal description of a thing must be expressed in the most economical way possible. This can only be done if the right frame is used to express the thing in mind. In the right frame you can just “see” the answer, because any translation between your frame and the optimal frame represents a suboptimal routine that can be compressed away. Therefore there is no upper limit on intelligence in the sense that there is no computable way to get the shortest description of a thing. There is always the possibility of compressing an idea further and reaching greater levels of understanding.
In this sense, understanding a physical theory such as classical or quantum mechanics is something akin to being able to just “see” the answer without having to grind through the physical equations used to represent a phenomenon. In some sense this would be akin to developing an intuition behind the equations. But wait one moment. How is it that one can develop an intuition for such physical models? Why is it that it may be hard for some to just “see” the outcome of the equations that are used to express physical phenomena? My hunch is that the physical equations themselves are suboptimal ways for expressing the underlying physical reality in the way that our minds can quickly comprehend. What is really going on inside my mind is that I am building up a world model that tracks the phenomenon of reality and placing “checkpoints” that represent known points corresponding to the outputs of the physical equations. But the physical equations themselves are not a complete description of reality, but only represent these “checkpoints” that correspond to the physical quantities that we can measure at well defined locations. What is really going on I suspect then is just building up of a neural network that is able to predict such phenomena in detail and works surprisingly well for almost all domains of interest.
I think this is related to the idea that intelligence is compression. But when we think of compression we immediately run into a conundrum: If something is compressible, it means the language used to express the piece of information is not optimal. An optimal description of a thing must be expressed in the most economical way possible. This can only be done if the right frame is used to express the thing in mind. In the right frame you can just “see” the answer, because any translation between your frame and the optimal frame represents a suboptimal routine that can be compressed away. Therefore there is no upper limit on intelligence in the sense that there is no computable way to get the shortest description of a thing. There is always the possibility of compressing an idea further and reaching greater levels of understanding.
In this sense, understanding a physical theory such as classical or quantum mechanics is something akin to being able to just “see” the answer without having to grind through the physical equations used to represent a phenomenon. In some sense this would be akin to developing an intuition behind the equations. But wait one moment. How is it that one can develop an intuition for such physical models? Why is it that it may be hard for some to just “see” the outcome of the equations that are used to express physical phenomena? My hunch is that the physical equations themselves are suboptimal ways for expressing the underlying physical reality in the way that our minds can quickly comprehend. What is really going on inside my mind is that I am building up a world model that tracks the phenomenon of reality and placing “checkpoints” that represent known points corresponding to the outputs of the physical equations. But the physical equations themselves are not a complete description of reality, but only represent these “checkpoints” that correspond to the physical quantities that we can measure at well defined locations. What is really going on I suspect then is just building up of a neural network that is able to predict such phenomena in detail and works surprisingly well for almost all domains of interest.