The memory-prediction framework is built around the idea that the neocortex implements a uniform algorithm throughout its entire structure, meaning that information in the visual, auditory, sensory, and even motor areas for example are all processed in the same way, even though they are hooked up to different systems. The manner in which the nervous system processes this information allows the brain to learn by mapping its structure to the structure apparent in the external world *(notice that this is consistent with the [[Santiago Theory of Cognition]], with the Santiago theory providing an explanation of how this structural mapping produces an experience of a world for the organism)*. The creation of such a cognitive structure occurs as [[The brain processes information hierarchically]]: basic sensory data are processed at low levels of the hierarchy according to basic features (such as line orientations or colors), which are then processed into progressively more abstract features that are known as [[Invariant Representations]] of the thing in the world, because these more abstract representations will be activated despite variable presentations of the actual thing that they are representing. Since the brain learns information by virtue of its neural connections, memories of information are stored in the network of neural connections the brain possesses. Memories are formed more strongly by the amount of experience a brain has with the stimuli that produce the memory, as neural connections become more strong and more detail can be processed with repeat exposure. [[Retiring the Language of Thought Hypothesis|This is exactly why I argue that serial-computer-based theories of cognition don't work]]. One of the key features of the brain's having a hierarchical structure is that as a remembered stimulus is repeatedly encountered, the neural connections that form the memory are formed again at lower levels of the brain's hierarchy. This means that as familiarity is gained with some stimulus, the brain is able to token the relevant memories more quickly because they become located nearer the bottom-up pathways that information takes through the brain. This is how [[Reflexes]] are formed. The brain is full of [[Recurrent Networks]], which account for how [[The brain processes information hierarchically]]. The strongest connections in the brain's network are from the bottom-up, but there are more numerous connections from the top-down which allow other regions of the cortex to contribute to the information being processed in a given brain region. This is the mechanism that allows the brain to make predictions: say you are about to eat some sour candy. Your brain sees the colorful packaging (often neon) and processes that the packaging is colorful and has the name of some sour candy on it. Regions of the brain familiar with color and language are able to make the prediction that the candy in the bag is sour, and as you are aware that your intention is to eat the sour candy, so is the neural connections throughout your brain, thus making the prediction that the tongue will encounter some sour candy very soon. Sour candy is such a powerful stimulus that it is able to produce a conditioned response, but throughout your brain, neural network regions are making the top-down prediction that the taste-buds will soon encounter some sour taste. You eat some of the candy, but much to your surprise, it is actually rice candy, and tastes relatively bland. There is a felt-sense of surprise and confusion because the predictions being made throughout your brain was that you would encounter a sour taste! When it did not happen, the predictions were violated and it was raised to your attention that something was not right! This densely connected, astronomically large network of neurons in the brain is not communicating in simple feedforward networks; connections throughout all aspects of the brain in addition to the storage of memories in the neural networks themselves is what allows the brain to have such a rich interaction between remembered experience and predicting the future. Such is the heart of Jeff Hawkins's memory-prediction framework, presented in [[On Intelligence - Jeff Hawkins]]. --- #Neuro #Psychology/Cognition #Psychology/Learning *June 7, 2021* #2021/6