From birth, the brain is equipped with a lot knowledge and the capacity to learn. Nature, known as genetics, provides the infrastructure for the brain to learn, but the content of what is learned depends on nurture such as culture, and interactions with others. Evolution is a slow process of adaptation, but evolution provided the capacity to learn. Learning enables quick adaptation to unpredictable conditions. Learning is how the external world becomes represented in an internal model. Updating the model when needed. New experiences change how the brain organizes itself. Synapses constantly change, reflecting what is learned. Learning can be accelerated or inhibited depending on the context. Learners need focused attention, active engagement, error feedback, and cycles of consolidation. Learning is a discovery and updating process which depends on how a culture treats curiosity and opportunities to learn.
The brain is far more detailed than the blueprints to build it. It would be impossible to code all information into the brain, so learning needs to supplement genes. Even the most simple of life’s creatures that have a brain, learn by habituation and association. Habituation learning is adaption to stimulus. Association learning is predictions based on prior information discoveries.
Even with limitations such as blindness or other brain impairments, individuals are capable of developing normal capacities, and using them with great dexterity. Brain dynamic of recycling means to reorient functions without genetic modification. Learning and education recycle functions. The brain appears to need room for more complex thoughts. Some functions become impaired.
As some environment information is the same throughout generations, evolution makes them predictably. Alternately, evolution makes some parameters change rapidly to adjust to volatile environmental aspects.
Babies are born with considerable knowledge inherited by evolutionary process. Nature and nurture are not opposites. Each rely on the other. Learning takes place within innate constraints. Learning does not start from nothing as learning uses many prior assumptions. The a priori hypotheses are used in obtaining meaning, and seeing what works best given the environment. Even from an early age, humans are capable of computing many abstract ideas and can access abstract institutions which enable higher learning.
Memory is a reconstruction. Memory is based on contact between two neurons. The more the related neurons fire together, the more they are wired together. Memory vanishes without retesting of knowledge. Long-term memory is based on testing the material, rather than just studying it.
The brain needs more than just intellectual stimulation, it takes appropriate nutrition, oxygenation, and physical exercise. Brain development requires exposing to various stimulus to make it flexible, otherwise the brain won’t develop the circuits. During childhood, the brain is overhauling its organization quickly, by either creating or eliminating synapses. This quick change also explains a large reason for childhood sensitivity periods.
What is seen are the projections that the brain has made meaningful from the flow of data. Learning uses previously missed information to change the internal model. Knowing what to learn to update the model.
New observations update thoughts in a probabilistic manner. A gradual rejection of false hypotheses, and maintenance of more rigorous hypotheses. Considers a myriad of ways to express the internal model, then utilizing that which incorporates the most data of the external world. The best fit for the state of external world.
Sometimes learning can get stuck. No options to do better seem to exist. Changes seem counterproductive, as they increase errors. Although better outcomes are possible, they are too far to be understood.
Convolutional neural networks learn faster and better because they generalize information. What was learned can be applies elsewhere.
Humans learn from each other. Even a single experience, a single trial, can bring about new understandings. Trying to learn more and more abstract rules, so that as many observations fit into the rule. While the brain creates a lot of meaning from very little data, machines need a lot of data to make some meaning. For computers, learning is difficult because there are so much data and possibilities to explore. Hard to select what to focus on. Artificial systems have a hard time learning abstract concepts, are not data-efficient, lack social learning, and lack composition.
Pillars of Learning:
To extract as much information from the environment, evolution created functions that facilitated learning. Stability requires all four functions. The functions are attention, active engagement, error feedback, and consolidation.
Attention amplifies focus. Attention is how the brain selects information, amplifies it, channels it, and deepens its processing. Decides when, what, and how to attend to information. Paying attention, also means choosing what to ignore. Directing attention means to choose, filter, and select. Without attention, students cannot perceive the teachers lesson, therefor cannot learn. Attention can be misdirected, which inhibits learning.
Active engagement encourages curiosity and experimentation. Active exploration of the world. Passive organisms learn little or nothing. To learn, the brain needs to form hypothetical ideas of the outside world, and then then test them. Passive or distracted students do not benefit from lessons, because their brains are not updating their models of the world. Only by actively following the course is information learned. Teachers aid in pedagogical progression, to guide student learning. Students do not learn much without guidance. But do need a structured learning environment with strategies for active engagement.
Error feedback corrects predictions of the world. Learning from mistakes is a popular form of learning. Every error is an opportunity to learn. Error reduction through feedback. Feedback that explains how to improve. Discovering errors enables correcting errors. Quality and accuracy of feedback influence speed of learning. Without a surprise, there is no learning. Prediction error is needed to learn. Error feedback is not punishment. Many children are punished or stigmatized for errors, and learn not to be curious to reduce errors. Errors should be corrected rather than punished.
Consolidation makes learned behaviors automatic, and involves sleep. Consolidation frees up mental energy for other purposes. Automation reduces the mental strain of an activity, allowing the mental bandwidth to be used elsewhere. Sleep is not inactivity, or just waste disposal. Brain remains active during sleep. Sleep goes over what was learned during the day, and gradually transfers it into an efficient compartment in memory. Sleep quality and quantity depends on how much was learned, as the more learned means more sleep is needed. During sleep, new information is not absorbed. Sleep makes discoveries more abstract and general.
Most of the information is about childhood learning, because childhood is a time of major brain development. The focus on childhood learning leaves out implications for adult learning. What does learning mean for adults? Childhood learning implications might not relate well to adult learning.
Error punishment during school is a major inhibitor of learning for children. But even adults are punished for errors, and there can be a lot of social sanctions against learning. Non-childhood learning inhibitors are missing, but that does incentivize considering how cultures can facilitate or inhibit learning.
Artificial Intelligence or machine learning, is explained in the book, but as a contrast to human learning. Highlighting how humans learn by expressing the limitations of machines. Machine learning is a feature of the book, but is not prominent.