Problem Solving Mechanism of the Brain

[Black Rose]

I have not totally figured it out yet.

But I saw a video saying the basal ganglia is important for breaking big problems into smaller ones.

I found this article and this image (not by the same research teams)

Attentional Reinforcement Learning in the Brain - New Generation Computing

Recently, attention mechanisms have significantly boosted the performance of natural language processing using deep learning. An attention mechanism can select the information to be used, such as by conducting a dictionary lookup; this information is then used, for example, to select the next...

link.springer.com

-

 

[Old Things]

Not everything is based on the brain.

 

[Black Rose]

Not everything is based on the brain.

please summarize the video

does it say how problems get solved?

 

[Old Things]

Not everything is based on the brain.

please summarize the video

does it say how problems get solved?

Watch the video or not. It is about consciousness.

 

[EndogenousRebel]

If were talking about reinforcement learning we gotta talk about *the cingulate cortex

 

[Black Rose]

If were talking about reinforcement learning we gotta talk about *the cingulate cortex

The brain is symmetrical. So both sides would need to communicate cortically and subcortically. And being in a 3D world the cingulate helps with balance in performing actions. So learning has some measure of weight on the sides where imbalance occurs.

The amygdala is connected to the orbital frontal cortex linked to the medial front cortex, which is connected to the cingulate in the way that it learns good and bad values.

2D rendition of cingulate and thalamus

Any action needs a counter-action. This goes into the thalamus after the basal ganglia tell us if it was correct or not as a valued action. So learning is about a hierarchical series of reinforcement.

 

[Hadoblado]

I'm not sure what the OP article has to do with problem-solving. I couldn't find any mention of problem-solving in it. Or OT's article either for that matter. If you want to pursue this line of questioning it's probably best to scope your questions more explicitly.

Problem-solving will draw on different parts of the brain depending on what the problem entails. For example, a monkey trap will require the motor cortex while the Monty Hall problem will not. There are likely hundreds of regions involved depending on what level of resolution you're after.

I would start by asking the same question but with one very simple task, then branch out to other tasks to start building a model of problem-solving.

 

[Hadoblado]

I haven't read this, but it's highly cited for something only eight years old. From a quick glance this is probably around the level of 2nd year psychology? Maybe third. Definitely higher than first. I can't tell you from experience whether it's a good read but it seems like a strong start for what you're interested in.

Edit: It's an entire textbook on the cognition and neuroscience involved in problem-solving. It probably still requires more prior knowledge than you have, but it's a much stronger starting point than going directly to articles.

 

[scorpiomover]

So learning is about a hierarchical series of reinforcement.

Yes.

1) What this means, is that people don't learn overnight.

If you tell someone something once, they are unlikely to remember it. If you tell someone the same thing day after day, people tend to remember it.

If you show someone how to do something once, they are unlikely to remember it, and will probably screw it up. If you show someone how to do the same thing day after day, people tend to remember it and when they need to do it, do it competently and efficiently.

2) Also, because people don't learn overnight, the subconscious learns from the things you do repeatedly. So even the things you don't consciously choose to do repeatedly, your subconscious tends to learn and then repeat even when you're not thinking about them. So the subconscious learns from habits. So the habits you choose/have, determine your areas of competency.

3) It is hierarchical, because knowledge builds on other knowledge to make new knowledge. If you know how to make bricks, you can make a house out of bricks. In the same error, minute errors in earlier learning can lead someone to interpret every piece of future information in terms of what is consistent with that minute error, even though it is wrong. So when fixing an error in earlier thinking, you also have to think about what that means to all of their consequences and their consequences of consequences, or the brain may not realise that they need updating too and may keep using the old code.

There are probably other effects that I have not mentioned here.

 

[Black Rose]

I haven't read this, but it's highly cited for something only eight years old. From a quick glance this is probably around the level of 2nd year psychology? Maybe third. Definitely higher than first. I can't tell you from experience whether it's a good read but it seems like a strong start for what you're interested in.

Edit: It's an entire textbook on the cognition and neuroscience involved in problem-solving. It probably still requires more prior knowledge than you have, but it's a much stronger starting point than going directly to articles.

It references ACT-R

That was developed in the 1980s

Before then in 1970s the perception action cycle was the main model

In 1992, Siri was made

Anyway in the paper/book, I get that to solve problems you need to build representations of the parts and how they fit together. And if you try something you fail you must remember that as segments of the larger problem. So it is the chucking the parts into those that work and don't work and configuring them to a solution. Not only that but you need to do so in working memory before trying them in the real world.

So that is a complex thing to do inside a computer because computers before 2015 lacked perception. So any demonstration needed toy worlds with decreased abstracted details to work. Today we use probability to give a more realistic measure of problem solving.

In the case of the brain, it has the extraordinary capacity to understand complex detailed perceptions. What I am interested in, is the coordination of the memory system in the brain to get things working from scratch. I believe this has to do with the reptile brain. The core circuit that allows the development of the system.

The cortex is important because it has the architecture within humans to operate in 3D worlds. It allows metacognition and thinking before acting. Those features come from inhibition that the reptile brain plays a part in where we reflect on what might be rewarding and what will not. The antiphon I would use to describe this is that we reflect because inhibition makes us come up with solutions in the front brain by memories reconfigurable until some idea emerges of what needs to be done.

Thought then is a circulation of memories in new ways to build new representations never seen that might work. The hierarchy of representation can be built up by reflection (circulation) of memories facilitated by lizard brain inhibitions. A repertoire of solutions tested over time that have interspersed parts/segments between them.

I would start by asking the same question but with one very simple task, then branch out to other tasks to start building a model of problem-solving.

I think that to begin you need a core and that core does branch out, but then it comes back to itself to facilitate different levels of abstract learning.

Animals with brains learn by deconstructing the problem internally this way.

The simplest I think is the Honey bee.

 

[Black Rose]

I believe that a simple problem-solving algorithm exists.

I think it just needs to be scaled up as a recurrent network.

-

Problem-solving is a deconstruction internally and then implementation of some set of actions in levels of abstraction with a branched out recurrent network.

 

[Hadoblado]

I'm a little worried that you're setting yourself up for failure when you enter an area with already established beliefs?

Why is inhibition important?
Why is inhibition from the reptilian brain particularly important?
Why does a singular simple problem-solving algorithm exist?

Inhibition as part of the circuit is likely one of the last things you establish, especially for deep brain areas like the reptilian. The reptilian brain is probably the last place I would personally look overall because humans are uniquely good at problem-solving but are far from the only animals with a reptilian brain.

You might have good reasons for thinking these things, I just don't see them rn. IMO, before you try to innovate in a field, you need to understand what other people are saying and why, then establish your position by explaining something they can't.

 

[Black Rose]

I'm a little worried that you're setting yourself up for failure when you enter an area with already established beliefs?

everyone has preestablished beliefs

Why is inhibition important?

for self-control,

people who stop themselves from acting carelessly can make better decisions.

and the two main functions of brain cells are to inhibit or be excitatory.

with the top-down control of the front brain we stop ourselves when needed

Why is inhibition from the reptilian brain particularly important?

because of the reward system, (motivations)

emotions are required for any action to take place

bad reactions are inhibited and self-monitoring means we need this for any goal

Why does a singular simple problem-solving algorithm exist?

It could be because it expands from small to large.

A bee can learn some things and other animals have larger brains.

But overall, I think solving problems is about deconstruction.

You look at what needs to be done and break it into smaller parts.

A Honey bee can do this but the larger the problem the more parts need to be incorporated into the strategy.

Inhibition as part of the circuit is likely one of the last things you establish, especially for deep brain areas like the reptilian. The reptilian brain is probably the last place I would personally look overall because humans are uniquely good at problem-solving but are far from the only animals with a reptilian brain.

If you remove the reptile brain from a human then problem solving stops.

This part of the brain is doing something important.

Specifically, it is engaged in motivations.

Motivations are the basis of the stop-and-go process of learning.

If I were to try and solve a problem I would need to understand how to move in such a way as not to hurt myself and also gage what is most important (prioritize)

You might have good reasons for thinking these things, I just don't see them rn. IMO, before you try to innovate in a field, you need to understand what other people are saying and why, then establish your position by explaining something they can't.

who are these people, what are they saying, what is it they cannot explain?

basically, my hobby is to look at ways to create intelligence inside machines but my models lack some coherency that I am establishing by laying out my conceptual framework here. So far I gained new ideas by trying to explain what I know and how it might work.

The main idea being: brains are memory networks folding together in such a way as to learn what they can do by thinking before it does the action. multiple levels exist in thinking. Honey bees have learned to roll colored balls into a hole to get rewards for example. So it had to establish the consequences of its actions for itself. I think this could scale up as an agent-based model, it would plan what to do with the parts involved in any situation via the folding of memory.

I think Google Deepmind made such a system to play StarCraft II

Human thinking is more complex though because of metacognition.

Metacognition would work by some loop within the front brain.

What Is Intelligence? Where Does it Begin?​

kurzgesagt channel

 

[Hadoblado]

I'm a little worried that you're setting yourself up for failure when you enter an area with already established beliefs?

everyone has preestablished beliefs

Yes, but the grounding for those beliefs and the fluidity of accommodating new information is important. The strength of neuroscience over other areas in psychology is that it's overwhelmingly evidence-based. My entire degree I had my beliefs overturned daily. Even in my thesis I had results that broke my brain because they were the complete opposite of what was expected based on what we knew at the time - I still don't know wtf was going on. You can't just guess based on experience that is mostly tmk irrelevant and expect it to work out.

Why is inhibition important?

for self-control,

people who stop themselves from acting carelessly can make better decisions.

and the two main functions of brain cells are to inhibit or be excitatory.

with the top-down control of the front brain we stop ourselves when needed

Okay but that's kind of like placing brakes as the central system in how cars work before you've considered acceleration. I would start from input/output and top-down mapping the circuit many many steps before disentangling excitatory and inhibitory mechanisms.

Why is inhibition from the reptilian brain particularly important?

because of the reward system, (motivations)

emotions are required for any action to take place

bad reactions are inhibited and self-monitoring means we need this for any goal

Again this is like talking about the chemical composition of fuel when you're trying to figure out how cars work.

Why does a singular simple problem-solving algorithm exist?

It could be because it expands from small to large.

A bee can learn some things and other animals have larger brains.

But overall, I think solving problems is about deconstruction.

You look at what needs to be done and break it into smaller parts.

A Honey bee can do this but the larger the problem the more parts need to be incorporated into the strategy.

Deconstruction likely helps with improving the efficiency of limited computational power.

Inhibition as part of the circuit is likely one of the last things you establish, especially for deep brain areas like the reptilian. The reptilian brain is probably the last place I would personally look overall because humans are uniquely good at problem-solving but are far from the only animals with a reptilian brain.

If you remove the reptile brain from a human then problem solving stops.

This part of the brain is doing something important.

Specifically, it is engaged in motivations.

Motivations are the basis of the stop-and-go process of learning.

If I were to try and solve a problem I would need to understand how to move in such a way as not to hurt myself and also gage what is most important (prioritize)

If you remove the lungs problem-solving stops. If you want to solve this problem, you need to deconstruct further than the necessities.

You might have good reasons for thinking these things, I just don't see them rn. IMO, before you try to innovate in a field, you need to understand what other people are saying and why, then establish your position by explaining something they can't.

who are these people, what are they saying, what is it they cannot explain?

basically, my hobby is to look at ways to create intelligence inside machines but my models lack some coherency that I am establishing by laying out my conceptual framework here. So far I gained new ideas by trying to explain what I know and how it might work.

The main idea being: brains are memory networks folding together in such a way as to learn what they can do by thinking before it does the action. multiple levels exist in thinking. Honey bees have learned to roll colored balls into a hole to get rewards for example. So it had to establish the consequences of its actions for itself. I think this could scale up as an agent-based model, it would plan what to do with the parts involved in any situation via the folding of memory.

I think Google Deepmind made such a system to play StarCraft II

Human thinking is more complex though because of metacognition.

Metacognition would work by some loop within the front brain.

What Is Intelligence? Where Does it Begin?​

kurzgesagt channel

Whoever has things to say about the field. In this case, the people writing the textbooks and articles about their theories. Experts. Open your mind and explore ideas that aren't your own. Keep the pieces even if you don't use them and have them inform your evaluation of your own ideas. Change your mind then change it back. Be elastic but critical. //rant

Aaaanyway I think I'm gonna let you go with this because it feels like I'm telling you off when that's not my intention. I see you as disoriented and I want to support you with some metacognitive awareness but I don't think I'm helping. Good luck.

 

[Black Rose]

Yes, but the grounding for those beliefs and the fluidity of accommodating new information is important. The strength of neuroscience over other areas in psychology is that it's overwhelmingly evidence-based. My entire degree I had my beliefs overturned daily. Even in my thesis I had results that broke my brain because they were the complete opposite of what was expected based on what we knew at the time - I still don't know wtf was going on. You can't just guess based on experience that is mostly tmk irrelevant and expect it to work out.

I am glad you had the experience of going to a decent school during those years which are the most important to learning higher-level conceptions. I was not able to do so but I spent a good amount of time thinking about what intelligence is and how it could happen. I understand what it means to be flexible because I always came to some conclusion that I would later forget and have to come up with a new model.

Whoever has things to say about the field. In this case, the people writing the textbooks and articles about their theories. Experts. Open your mind and explore ideas that aren't your own. Keep the pieces even if you don't use them and have them inform your evaluation of your own ideas. Change your mind then change it back. Be elastic but critical.

One thing I found out was that there is this thing called the free energy principle by Fristin, he goes about it from the angle of energy physics. It was in 2012 that he came up with it. At that time I never knew anything about energy systems. The basis of it is to conserve energy when learning. Homeostatic learning. At that time I barely understood error correction and the cycle of perception with action. my reinforcement system did not have a reward function until last year.

2021 was when I understood what a loop was. I had in my mind what homeostasis was before but it was incomplete. I used loops in my computer programs but was unaware that loops are the main principle of intelligence. After completing my homeostatic program in 2023 which was basically a thalamus learning to regulate itself. I fully realized Loops are what help with learning and keep organisms stable. They even help think ahead. If the loop is the basis for everything then all that matters is how loops change for intelligent beings to adapt to their surroundings.

In my mind, I see that the brain has many dependencies. I get that the brain is changing itself but a core set of loops exist. They allow the brain to pay attention to itself. Thus it can change in an intentional way. Make choices. Of all options available what should I do. This would be the inbuilt mechanism of problem-solving, having a representational memory where we must change ourselves to find answers. Looking inward to create those options. Then confirm or reject them based on what we do in the external world. Not confirmation bias because we would have multiple options and multiple scenarios where they would be possible. By testing them we gain knowledge in some way changing the structures of loops as a memory consolidation without catastrophically forgetting but opening up new options with the new data. In the end, it is like entanglement and wave collapse.

(idea internal loop formation with external interactions making them realized)

-

Computers have recently become really good at simulating brains. I saw a mouse brain simulation and it has all the brain cells a mouse has mathematically accurate. My computer can do a trillion operations a second because it is 10 years old but I can imagine what a computer today would be able to do with a quadrillion operations a second with brain a simulation. I saw a PlayStation 5 commercial and it was in real-time on a 4K definition OLED TV. It could do simulations as well.

This video is 4 years old but it makes sense that computers are now powerful enough to simulate to some degree with accuracy what happens in the brain.

Since this project is about personalized medicine it would be nice to know what my brain is doing.

The Virtual Brain - Explainer Video [updated]​

 

[EndogenousRebel]

If you're trying to understand the brain as an information system, like a computer that's cool, but that is a high level abstraction.

Not as high level as Jung's cognitive functions like intuition, feeling, thinking, sensing, but it's still pretty up there.

These things represent actual physical events that occur in the nervous system.

The theories that you would make, in order to be taken seriously would have to in some way explain the underlying phenomena found in neurological studies, predictively. Maybe that is not what you're going for?

You are creating something that is by definition a post-hoc analysis, which isn't really an issue, except that in these terms, you have no hypothesis.

 

[Black Rose]

If you're trying to understand the brain as an information system, like a computer that's cool, but that is a high level abstraction.

Not as high level as Jung's cognitive functions like intuition, feeling, thinking, sensing, but it's still pretty up there.

These things represent actual physical events that occur in the nervous system.

The theories that you would make, in order to be taken seriously would have to in some way explain the underlying phenomena found in neurological studies, predictively. Maybe that is not what you're going for?

You are creating something that is by definition a post-hoc analysis, which isn't really an issue, except that in these terms, you have no hypothesis.

I'm not sure what level of abstraction I should be looking at.

Should it be mathematical?

What is the explanation other than intelligence?

Intelligence in some way requires attention, self-attention—and social interaction.

So a biological system is supposed to be intelligent in some way.

The best I got is the coordination problem.

Parts work together.

I could predict what the brain is doing but I am not a brain scientist.

All I can do is translate what I know about brain mechanisms into models.

Those models will not be atomically precise. (10^25 atoms)

It is the case that I am trying to create meta-intelligence.

Maybe not at the scale of a supercomputer (exaflops) but some basic principles.

People with supercomputers already understand, I do not.

-

need to go to the store, be back soon,

will show my idea for algorithmic intelligence.

 

[EndogenousRebel]

If you're trying to understand the brain as an information system, like a computer that's cool, but that is a high level abstraction.

Not as high level as Jung's cognitive functions like intuition, feeling, thinking, sensing, but it's still pretty up there.

These things represent actual physical events that occur in the nervous system.

The theories that you would make, in order to be taken seriously would have to in some way explain the underlying phenomena found in neurological studies, predictively. Maybe that is not what you're going for?

You are creating something that is by definition a post-hoc analysis, which isn't really an issue, except that in these terms, you have no hypothesis.

I'm not sure what level of abstraction I should be looking at.

Should it be mathematical?

What is the explanation other than intelligence?

Intelligence in some way requires attention, self-attention—and social interaction.

So a biological system is supposed to be intelligent in some way.

The best I got is the coordination problem.

Parts work together.

I could predict what the brain is doing but I am not a brain scientist.

All I can do is translate what I know about brain mechanisms into models.

Those models will not be atomically precise. (10^25 atoms)

It is the case that I am trying to create meta-intelligence.

Maybe not at the scale of a supercomputer (exaflops) but some basic principles.

People with supercomputers already understand, I do not.

-

need to go to the store, be back soon,

will show my idea for algorithmic intelligence.

Well, I would think in terms that would enlighten us. So this is the easy part where you ask questions that we don't have answers to, and then starting from a reasonable place that we do know, and from there explore and experiment with ideas until you have something that you can dialectically argue with.

I don't really have an answer to most of your questions. So I'll just use this as a blank check to look into stuff that I've wanted to do myself.

Whats up with cytoarchitecture?

So the information the brain is transmitting isn't going across a single medium like a wire that stays constant all the way through, such as in network computing.

Ostensibly, the information is travelling across the brain and being transmuted by the terrain it is traveling across.

Even within localized areas of the brain associated with same functions the cytoarchitecture can vary greatly.

A popular assumption would be that these structures are there for some evolutionary benefit, how far does that go? Can we generalize this for parts of the brain that have similar cytoarchitecture or is this something that is super context dependent?

Does the information of interest have to be within that brain region to be modified in such a way, or will the brain send information to the region where that information is without it having to go through every node in the brain, hence saying commute time?

Keeps me up at night.

 

[Black Rose]

I don't really have an answer to most of your questions. So I'll just use this as a blank check to look into stuff that I've wanted to do myself.

Whats up with cytoarchitecture?

I am not an expert but I read Jeff Hawkins first book in 2004 - He said any signal translated into motion because brains are for motion. I did not understand much other than the part where he said signals go up and down a hierarchy. So if the brain makes wrong predictions it corrects them.

This diagram is what I know now:

I need to read his second book but I tried to understand it in 2005.

 

[Black Rose]

No one has to take it seriously but this is how I think the brain is partitioned to solve problems, it is not in 3D.