What is hidden behind the elementary concept called "intelligence"? Some argue, like the mathematician Jeff Hawkins, that it is the ability to solve new problems by remembering the solution of previous problems
Israel Benjamin, Galileo
Direct link to this page: https://www.hayadan.org.il/memory260105.html
An old story pits a mathematician and a physicist against the challenge of boiling water in a pot on the kitchen counter. They both find the solution: move the pot from the counter to the gas stove and turn on the gas. After that, the two academics are required to boil water in a pot placed on the table.
The physicist moves the pot from the table to the gas stove and turns on the gas; The mathematician moves the pot from the table to the counter and declares "We have already solved this problem" (the Mahdrin also adds to the story an engineer, who simply sets the table on fire). This story has always been used to poke friendly jabs at mathematicians, but if you ask Jeff Hawkins, the mathematician in the story expresses the true basis of human intelligence: solving new problems by remembering how to solve previous problems.
A millionaire without a choice
Hawkins gained publicity thanks to his key contributions to the success of handheld computers and the "computerized digital assistant" (PDA - Personal Digital Assistant). He was the originator and chief creator of the first commercially successful PDA, under the name Palm Pilot. Among the other innovations included in this product was the first practical method of recognizing human handwriting: Hawkins dared to demand that people learn to draw letters in a new way, and rightly predicted that users would agree to this if in return they received the ability to write reliably and quickly.
In fact, Hawkins' real dream for the past twenty years has been to find a theory that would unify the many details known about the human brain into a comprehensive explanation of intelligence. Companies where he worked, such as Intel, refused to finance the projects he proposed - projects that focused on brain research with the aim of advancing the field of artificial intelligence.
MIT (the renowned Massachusetts Institute of Technology) wasn't interested either; The field of artificial intelligence at that time was mainly led by computer scientists who were not interested. Another blow was dealt to him by the University of Berkeley in California, which was not ready to approve a curriculum that would combine brain research with computer science.
These disappointments led Hawkins to the decision to earn enough money to finance the research he planned himself. This plan succeeded: in 2002, Hawkins founded the Redwood Neuroscience Institute, which aims to create biology-based models of memory and thinking. In other words: the institute works on theories that explain how the human brain works.
In October of this year, a book appeared that Hawkins wrote together with Sandra Blakeslee (Blakeslee), a journalist for the New York Times on science and especially neuroscience. The book is read "On Intelligence: How a New Understanding of the Brain Will Lead to the Creation of Truly Intelligent Machines" In the book, written for the general public, Hawkins summarizes his findings over the past twenty years, and explains a detailed theory about the functioning of the brain, the way it works, and the way we can imitate such abilities.
Different roles, one method
Neuroscientists have accumulated an enormous amount of information about what is happening in the different parts of the brain. Initially, the information was gathered from those unfortunate cases of damage to certain areas of the brain: the functions lost to each such victim helped to understand the functions of the different areas. Today it is possible to find out with tools such as functional MRI which areas of the brain are active when a certain cognitive task is performed.
In these and other ways, the many parts of the brain entrusted with processing information from the senses and activating muscles, as well as the performance of thinking tasks by dividing them into sub-tasks, each of which is directed to a different part of the brain, have been identified and mapped. This information, as enlightening and fascinating as it may be, yielded no more than hints to the central question of the essence of intelligence and the form in which the brain "produces" it.
Hawkins thought in other directions, influenced by an article by Vernon Mountcastle, one of the founders of modern neurology. This article pointed out the common structure of all parts of the "new cerebral cortex" (neocortex). It is customary to associate the new shell with awareness, thinking and the use of language, as well as with many other functions that we associate with human intelligence.
Although this shell has so many functions, most of which are focused on different areas of it, it is built entirely of six thin layers connected to each other. Hawkins came to the hypothesis that the different parts of the neocortex have more in common than they differ; Although each area has a defined role, all areas work in exactly the same way.
This hypothesis is supported by the long-known fact: functional impairments caused by brain damage (such as the damage of a stroke) disappear or become milder in some cases, when another area of the brain takes over the functions of the damaged area. Like a computer that can process words one minute, and the next minute perform a complex mathematical analysis, so - according to Hawkins - the same basic six-layer structure can be responsible for different tasks.
memory and prediction
What is the same general structure that is suitable for performing so many different cognitive tasks? Hawkins' next inspiration was a feature long known to psychologists and neurologists, and actually known to each of us: the brain's ability to quickly extract "interesting" details from the flood of information flooding our senses.
When we enter a familiar room, even without being aware that we are surveying the room, we can quickly find out if something has changed in it. When having a conversation in a noisy room, we manage to ignore the background noises and complete words we couldn't hear. These and many other phenomena illustrate our ability to think "top-down".
What is hidden behind the intelligence
This name expresses the opposite direction in relation to the direction we tend to attribute to the processing of sense data, which begins with the entry of a huge amount of details into our sense organs (for example receiving a large number of more or less bright points on the upper left side of the field of vision, surrounded by black points); The details are then filtered and analyzed to create a smaller amount of processed information (for example "a mostly white circle on a black background") from which the identification ("moon") is obtained. In top-down processing, what we expect to see dictates the processing of information at lower levels.
There is ample evidence for top-down processing. This evidence includes psychological experiments such as those showing that many people read a text without realizing that it contains spelling errors, and neurological findings - for example, it is estimated that a significant part of the information in the optic nerves flows in an unexpected direction, from the brain to the eyes.
Most of us are not aware of our ability to fill in the details that the senses are missing. This surprising ability is not among those that are expected to be mentioned when talking about intelligence. Hawkins hypothesizes that the reason for this is simply because this is the process that stands at the foundation of intelligent thinking. This process is a constant creation of expectations in the "higher" layers of the cerebral cortex, and comparing them to what is received from the "lower" layers to which the information from the senses arrives.
Intelligence, according to Hawkins, has a simple definition: the ability to create an internal model of the world and use this model to create expectations. In this way, intelligence contributes to increasing the chance of survival, and therefore the large investment in energy necessary for the existence of the brain is justified in evolutionary terms.
This idea, like most of Hawkins's ideas, is not new - see, for example, "creating the future" (this section, Galileo 72). Hawkins' unique contribution is in the synthesis of ideas that he claims constitute the full theoretical basis for understanding thinking. This model, unlike most models proposed in the past, has clear predictions, so it can be tested and disproved and meets the criteria of a scientific theory.
To create expectations or a plan, to compare the situation with what we expected, and to create updated expectations or corrective actions: this template is known to scientists as the basic method of research, and is known in the business world under different names such as "quality management" or simply "correct management". The very naturalness of such processes may suggest to us that they are deeply rooted in our thinking mechanisms.
creativity and intelligence
Creativity, as it is expressed in solutions to new problems, in writing, art or science, is surely one of the highest expressions of human intelligence. Hawkins sees creativity as part of the same model: creativity is prediction by analogy - dressing up already successful ideas in new clothes.
On the one hand, this simplistic description shifts the weight of the question to the creation of analogies - a process that is certainly not simple. On the other hand, it allows to suggest ways to improve creativity. For example, when trying to solve a problem, it is useful to break it down into its components and try to change the order of the components - perhaps by writing each one on a note, mixing the notes and examining the new order obtained. This "game" gives an opportunity to identify similarities to things we have already seen, and to create new analogies.
Is there any truth to these ideas? Some support can be seen in the fact that the connection between memory and intelligence is already known. Some of the tasks included in IQ tests are essentially memory tests. Other parts of IQ tests test the ability to predict ("find the next number in a series") or recognize the unusual.
If intelligence is truly based on retrieving details from memory, deriving predictions from them, and comparing the predictions with reality, these are exactly the tasks we would expect to see as predicting intelligence. The correlation between creativity and intelligence also corresponds to the hypothesis that both originate from the same neural mechanism (although it is necessary to explain here why this correlation disappears precisely in those with high intelligence).
In any case, Hawkins is proud that his new book includes proposals for more than ten experiments designed to test his theory. The research institute he founded strives to carry out these experiments, as well as to develop "smart" software - for example for processing visual information - built according to the principles of the new theory.
One of the important conclusions from this theory is that the prevailing approach to creating artificial intelligence will never succeed: it is impossible to build an intelligent computer with "a lot of brain and few senses". An intelligent computer will have to be connected to many and varied sensors, so that it can build expectations and confront them with reality.
These sensors do not have to be similar to the human sensory system: for example, a computer that would be directly connected to weather information (from meteorological stations, satellites, observation balloons, etc.), might be able to create quality predictions about tomorrow's weather with the same effortless naturalness that we Predicting the next syllable to be uttered by the person we are listening to.
https://www.hayadan.org.il/BuildaGate4/general2/data_card.php?Cat=~~~59002215~~~58&SiteName=hayadan
