I’ve almost finished Hawkin’s book On Intelligence. I find the whole story fascinating. I sounds indeed simple and I do think simple should be the way.
However, so far, I’ve not found the answer to one of my interrogations. I get how the CLA strengthen the connections based on patterns to fit perceptions and be able to predict with accuracy the “state” of the world. However, what motivates it ? What makes me want to follow a path rather than another, not put my hand on a hot surface, catch the ball when it’s coming at me ? It is all thank to/because of that old reptilian brain that implements basics instincts and behaviours (looking for food, avoiding pain…). If so, does it mean that deep down, we’re only moved by pure instincts ?
First day on the forum and you’re asking the hard questions
Honestly these are questions we are all trying to figure out. We’ll let you now when we understand more, in the meantime, keep up with our research at https://numenta.com/papers/. Our latest work touches on a sensorimotor loop that we think exists in all cortical columns of cortex. But we are not talking much about action yet. We must first understand how the brain creates a sensorimotor model of the world before we can understand how actions are stored, persisted, and invoked.
My take on this is that much of the heavy lifting here is done in evolutionary older brain stem structures where the basic animal behaviour is shaped by the much smarter cortex. Call them instincts if it helps.
I have some prior posts that may help you get started on sorting this out:
Basically living organisms behave (=move) to acquire resources from the environment, so the more “free” resources or the more they “need them” the more “motivated” they become - ie “motivated” means here “willing ot invest resources to get resources”, since no living organism gets resources without investing resources (like time, muscle energy etc)…this is the efferent part of “motivation”…there is also an afferent part: Any organism must prioritize between goals it wants to achieve…achieving some goals give access to resources, like a lion killing a goat - this gives access to the goats meat, the resources. But killing the goat, achieving this goal, is what gave access. Another goal could be achieving offspring, to achieve this goal, the lion must engage a female lion target and spend resources on this. So you see there are acquring and spending of resources to achieve goals. If yo think about this, the concept of “motivation” can easily mean two non interrelatable things: Either investing resources to get more resources for safety and comfort reasons, or to spent resources for pleasure of some kind. So there is absolutely not only one motivational system (like in the belif-intention model you mention), there are at least two. The resource system will allways interfer with the pleasure system and it can overrule it…if you are hungry, food come before sex in most cases, and these systems probably operate in the back…like woman can become pregnant when they have a good period in their life…and if they are too skinny they cannot become pregnant…one can say it is between short term and long term survival…so pleasure for long term survival and pain for short term survival…but today in the human world we so often see that this original biological mechanism is out of function: Many people only need to care about pleasure, so our nervous systems spend lots of time to figure out what kind of pleasure we need next…new on netflix?..this is a third factor behind motivation: The kind of learning process we enter - we might hate repetitions (autisic people love it)…and we might be dependent on news (or fake news like click bait)…all in all, to sum up, theoretical neuroscience has not yet been able to use logical conceptual philosphical thinking to straight out these issues in a way that can make them programmable. Instead the discussion of the concepts used and the processes that HAS TO TAKE PLACE has not been investigated thorughly and professionally. Therefor so much AI and realAI ends up with poor conceptual models that cannot be transformed to usefull software. Good luck in a sparse stony ground of conceptual understanding.
I think that the Basal Ganglia is the piece of the brain you are looking for.
There is also the theory of Reinforcement Learning, which is a mathematical model of how we think the brain ought to be motivated. This theory, in conjunction with deep learning, has been used to solve the games of Chess and Go.
From my research, I find the process is much simpler. Our brains are a continual processing system and “motivation”, like numerous other states is an illusion.
We know that sensory input requires motion (Hawkins states this often although it has been a known fact for many years) but what most researchers miss is that every output also requires motion, physical or virtual. Even qualia is manifested in some act. I cannot “think” that I love someone without some motor action involved in that “thought”. This continual sensory input, cortical processing and motor output is a cycle of never ending, never stopping action and thus motivation is a perceived state which is actually a set of network dynamics at play. This is also why connectomics are just as important as neuron physiology. When you consider animals at there earliest evolutionay states, they are reactors to the environment they live in where sensory input and motor output go hand in hand to survive. In essence, our current brains haven’t really gone beyond this, we just have a much more complicated network that has evolved.
Motivaton is just sensory input giving us a desired output, physical or virtual. (Virtual is motor imagery)
Action causes sensation. But you can imagine sensation as well. Think about what it feels like to pet a cat. you know what fur feels like, and you did not need actions to activate that pattern. While there might be something like motor commands involved in imagining petting a cat, you don’t actually have to take action to recall sensations.
I think it’s probably similar for other conceptual thoughts. There may be some motor-like activator, but you obviously do not have to move your sensors to think about anything.
Yes but your imaginary sensory input involves motor output.You cannot “imagine” the feeling of petting your cat without moving your hand across the cat = motor imagery. Our nervous systems are explicable tied together through sensory-networks-motor-sensory-networks-motor-sensory-networks-motor … ad nauseam processing. The mistake every one makes is that they believe everything begins with sensory input. This is the IP model which is wrong.
Please see my link " Intelligence vs Consciousness" mentioned above.
I postulate that the links roughly parallel to the arcuate fasciculus couple the motor planning stages at an early motor planning level to the later sensory stages in the parietal lobe.
In the explanation, it describes a high-level loop that is the basic consciousness process.
This process is at a level before being gated to action.
So in a sense - the motor planning stages are driving the internal states in the brain.
This concept is inspired by the “The Global Neuronal Workspace Model” described here: