Additional thoughts on learning and intelligence
In previous posts I've looked at the idea that learning is the process of connecting responses to stimulus, and that the actual way that learning happens in animals is essentially a form of data compression. If we think of learning like this, then we should ask how the data compression works, what process or rules does it follow if any? The key to good data compression is to find patterns in the data and take advantage of that, and the world we live in comes ready made with rules that govern how it works, the physical laws of the universe. It would seem that it would be possible for our brains to take advantage of this to compress responses to stimuli efficiently, and that that process could have the beneficial side effect of making the prediction of physical processes easier. The learning process would not only be encoding individual responses, but also patterns in those responses and the rules that determine what shape those patterns take.
What would the role be of the intelligent parts of the brain if the brain was already naturally adapted to learning and predicting the physical world? Well, there's one part of the world that, at least to a first approximation, doesn't appear to follow simple physical laws, and that's animals. I'm not claiming that the laws of physics and chemistry don't govern animal behavior, but they're layered in such complex systems that it's not immediately obvious how to use those laws to make predictions. Making physical predictions about simple physical objects, or even relatively complex structures like trees or rivers is fairly straightforward. Simple applications of momentum, friction, gravity, etc. are enough to learn how most objects in the world will behave, but animal behavior is governed by a set of rules that's essentially indecipherable for a single individual. One area that makes learning and predicting this kind of behavior particularly difficult is that reactions can happen over huge time spans. If you push a rock it'll move and maybe keep moving for a bit, but if you push an animal the reaction is unpredictable and the reaction might change over minutes, hours, days or even longer. The intelligent parts of the brain are good at finding patterns in complex situations without a lot of repeated exposure, it may be that they're evolutionary advantageous for dealing with animal interactions.
We can imagine the brain of early humans, primates or even earlier animals trying to learn appropriate reactions to stimulus from both their inanimate and animal surroundings. A system that works well for learning one kind system is unlikely to be adapted to learning about the other. Instead it makes more sense to have two different kinds of systems interacting, with each specializing in a particular kind of problem. Learning might have been fine for simple animals that were trying to survive in a mostly inanimate environment, but as other animals started to become an important factor than having the ability to intelligently process those interactions would be much more useful. If intelligence is the ability to make comparisons, we could see how comparing predictions to observations or comparing different memories of similar situations would be useful in figuring out complex animal behavior. In essence the best reason for humans to become more intelligent in that they had to interact with other intelligent animals, including other intelligent humans.
What would the role be of the intelligent parts of the brain if the brain was already naturally adapted to learning and predicting the physical world? Well, there's one part of the world that, at least to a first approximation, doesn't appear to follow simple physical laws, and that's animals. I'm not claiming that the laws of physics and chemistry don't govern animal behavior, but they're layered in such complex systems that it's not immediately obvious how to use those laws to make predictions. Making physical predictions about simple physical objects, or even relatively complex structures like trees or rivers is fairly straightforward. Simple applications of momentum, friction, gravity, etc. are enough to learn how most objects in the world will behave, but animal behavior is governed by a set of rules that's essentially indecipherable for a single individual. One area that makes learning and predicting this kind of behavior particularly difficult is that reactions can happen over huge time spans. If you push a rock it'll move and maybe keep moving for a bit, but if you push an animal the reaction is unpredictable and the reaction might change over minutes, hours, days or even longer. The intelligent parts of the brain are good at finding patterns in complex situations without a lot of repeated exposure, it may be that they're evolutionary advantageous for dealing with animal interactions.
We can imagine the brain of early humans, primates or even earlier animals trying to learn appropriate reactions to stimulus from both their inanimate and animal surroundings. A system that works well for learning one kind system is unlikely to be adapted to learning about the other. Instead it makes more sense to have two different kinds of systems interacting, with each specializing in a particular kind of problem. Learning might have been fine for simple animals that were trying to survive in a mostly inanimate environment, but as other animals started to become an important factor than having the ability to intelligently process those interactions would be much more useful. If intelligence is the ability to make comparisons, we could see how comparing predictions to observations or comparing different memories of similar situations would be useful in figuring out complex animal behavior. In essence the best reason for humans to become more intelligent in that they had to interact with other intelligent animals, including other intelligent humans.