The realization of an object from the background — MoK

BECOMING

We humans mirror and recapitulate
All of evolution while growing in our mother’s womb,
Racing through the stages in which life evolved.

During this nine months and even beyond that
We move from mindlessness to shadowy awareness
To consciousness of the world around us
Onto consciousness of the self
And then even to becoming conscious
Of consciousness itself.

For the first two and one-half years of life
The inexplicable holistic world
Is experienced less and less holistically
As the child discovers the
Bounds of discrete objects.

Can't comment on neurological development, but from how I understand what the option is, I would say that an option always requires another option for it to be an option. Only if I know that I can also take the left path, does the right path become an option. Otherwise, it's just a path. Or rather, the path, I should say.

I would create a primary single object from the two options taken as a whole. And you yourself offered a fork in a maze as a starting point. The fork is this single object. You realize it is a fork, a decision-point, the sudden blur of the way further. And then you

immediately realize that there are two options available for you, namely the left and right path. — MoK

So my answer would be that the fork always precedes the options. To understand an option, one first experiences a moment of unclarity. And then grasps the elements that make up this unclarity, like how a wave has ascending and descending parts.

As the single fork precedes the multiple options, you get this single object as a starting point. A single option is just a part of a fork, which you always extract later, already knowing there will be multiple options as it comes from recognizing the fork as being a fork.

And because options are always multiple by the very nature of a fork, it doesn't matter if the process of picking an option is always deterministic. They are still options because they are multiple.

I would like to invite Pierre-Normand here since he is very knowledgeable in AI. Hopefully, he can help us answer these questions or give us some insight. — MoK

I'd like to comment but I'm a bit unclear on the nature of the connection that you wish to make between the two issues that you are raising in your OP. There is the issue of reconciling the idea of there being a plurality of options available to an agent in a deterministic world, and the issue of the cognitive development (and the maturation of their visual system) of an infant whereby they come to discriminate two separate objects from the background and from each other. Can you make your understanding of this connection more explicit?

I'm sorry, but I can't follow you. Could you please write your opinion in plain English so I can understand what you're discussing?

Can't comment on neurological development, but from how I understand what the option is, I would say that an option always requires another option for it to be an option. Only if I know that I can also take the left path, does the right path become an option. Otherwise, it's just a path. Or rather, the path, I should say. — Zebeden

I say that you have only one option available when there is only one path available to you.

So my answer would be that the fork always precedes the options. To understand an option, one first experiences a moment of unclarity. — Zebeden

Could you please elaborate here?

I say that you have only one option available when there is only one path available to you. — MoK

Then, I would say that I have no options at all. I think that the only option is not an option, but rather a mere necessity.

Could you please elaborate here? — MoK

First, you experience a situation that requires decision-making. Once you're in such a situation, only then do you start examining options. Before that, everything was clear and certain (I was just going forward on this single path), and now I'm weighing my options at the crossroads, hence the uncertainty.

I'd like to comment but I'm a bit unclear on the nature of the connection that you wish to make between the two issues that you are raising in your OP. — Pierre-Normand

Ok, I will try to make things more clearer for you.

There is the issue of reconciling the idea of there being a plurality of options available to an agent in a deterministic world, — Pierre-Normand

First, I have to say that De Broglie–Bohm's interpretation of quantum mechanics is correct since it is paradox-free. The motion of particles in this theory is deterministic though. By deterministic I mean given the state of the system at a given point in time the state of the system at a later time is uniquely defined by the former state. So, the motions of particles in the brain are deterministic as well accepting De Broglie–Bohm's theory. What bothers me is that we for sure know that options are real. We also know for sure that the existence of options is due to neural processes in the brain. Neural processes are however deterministic so I am wondering how options can possibly result from neural processes in the brain. I think we can resolve the big problem in the philosophy of mind, the problem is that hard determinists claim that options cannot be real. Of course, the hard determinists cannot be right in this case since we can obviously distinguish between a situation in which there is only one object and another situation in which there are two objects. I studied neural networks in good depth in the past. My memory on neural networks is very rusty now but I would be happy to have your understanding of this topic if you can explain it in terms of neural networks as well. Can we train a neural network to realize between one and two objects and give outputs 1 and 2 respectively? If yes, what does happen at the neural level when it is trained to recognize two objects?

and the issue of the cognitive development (and the maturation of their visual system) of an infant whereby they come to discriminate two separate objects from the background and from each other. — Pierre-Normand

Please let's focus on one object first. If we accept the Hebbian theory is the correct theory for learning then we can explain how an infant realizes one object. Once the infant is presented with two objects, she/he can recognize each object very quickly since she/he already memorized the object. How she/he realizes that there are two separate objects is however very tricky and is the part that I don't understand well. I have seen that smartphones can recognize faces when I try to take a photo. I however don't think they can recognize that there are two or more faces though.

Can you make your understanding of this connection more explicit? — Pierre-Normand

I tried to elaborate the best I could. Please let me know what you think and ask questions if you have any.

First, you experience a situation that requires decision-making. Once you're in such a situation, only then do you start examining options. Before that, everything was clear and certain (I was just going forward on this single path), and now I'm weighing my options at the crossroads, hence the uncertainty. — Zebeden

I am interested to know what happens at the neural level when we realize that there are two paths.

So I am wondering how can deterministic processes lead to the realization of options. — MoK

This is trivially illustrated with the most simple code.

Take a step.
Count the ways forward (don't include the way you came)
If 0, it's a dead end. Only option is to turn around.
If 1, continue that one way
else there are multiple options.

It's that easy. The realization of multiple options is as simple as counting, and there is even multiple options with case 1 since a good maze following program might conclude it to not be productive to follow the current path to its unseen end.

Almost all computer programs are fully deterministic and are great models to simplify what might otherwise be a complex subject.

What you need to worry about is not the realization of options, but how determinism always results in the same choice given the exact same initial state. So our program might be crude and uses the right-hand rule, in which case it doesn't even count options, it just takes the first rightmost valid path and doesn't even notice if there are other options. A better program would be more optimal than that, but then complexity is required, and it still does the same thing given the same initial state.

So realization of options is one thing, but no matter how many options there are, only one choice can be ultimately be made, even if determinism is not the case. You can follow a choice in the maze, and if it dead ends, you go back and take the other way, which is 'doing otherwise'. Even the right-hand robot can do otherwise in that sense.


As for the infant process of neural development, that's an insanely complex issue that likely requires a doctorate in the right field to discuss the current view of how all that works. It seems irrelevant to the topic of determinism and options.

First, I have to say that De Broglie–Bohm's interpretation of quantum mechanics is correct since it is paradox-free. — MoK

All the interpretations are paradox free. None of them has been falsified (else they'd not be valid interpretations), and some of them posit fundamental randomness, but several don't.

I don't like Bohmian mechanics because it requires FTL causality and even retro-causality, forbidden by the principle of locality, but that principle is denied by that interpretation. That makes it valid, but it doesn't make me willing to accept it.

This realization is due to neural processes in the brain. — MoK

Not quite. That realisation is neural processes in the brain. It is not seperate from yet caused by those neural processes.

And a babe's brain is pre-wired to recognise faces and areola.

Please let's focus on one object first. If we accept the Hebbian theory is the correct theory for learning then we can explain how an infant realizes one object. Once the infant is presented with two objects, she/he can recognize each object very quickly since she/he already memorized the object. How she/he realizes that there are two separate objects is however very tricky and is the part that I don't understand well. I have seen that smartphones can recognize faces when I try to take a photo. I however don't think they can recognize that there are two or more faces though. — MoK

Hebbian mechanisms contribute to explaining how object recognition skills (and reconstructive episodic memory) can be enabled by associative learning. Being presented (and/or internally evoking) a subset of a set of normally co-occurring stimuli yields the activation of the currently missing stimuli from the set. Hence, for instance, the co-occurring thoughts of "red" and "fruit" might evoke in you the thought of an apple or tomato since the stimuli normally provided by by this subset evokes the missing elements (including their names) of the co-occurring stimuli (or represented features) of familiar objects.

In the case of the artificial neural networks that undergird large language models like ChatGPT, the mechanism is different but has some functional commonalities. As GPT-4o once put it: "The analogy to Hebbian learning is apt in a sense. Hebbian learning is often summarized as "cells that fire together, wire together." In the context of a neural network, this can be likened to the way connections (weights) between neurons (nodes) are strengthened when certain patterns of activation occur together frequently. In transformers, this is achieved through the iterative training process where the model learns to associate certain tokens with their contexts over many examples."

I assume what you are driving at when you ponder over the ability to distinguish qualitatively similar objects in the visual field is the way in which those objects are proxies for alternative affordances for action, as your initial example of two alternative paths in a maze suggests. You may be suggesting (correct me if I'm wrong) that those two "objects" are being discriminated as signifying or indicating alternative opportunities for action and you wonder how this is possible in view of the fact that, in a deterministic universe, only one of those possibilities will be realized. Is that your worry? I think @Banno and @noAxioms both proposed compatibilist responses to your worry, but maybe you have incompatibilist intuitions that make you inclined to endorse something like Frankfurt's principle of alternate possibilities. Might that be the case?

This is trivially illustrated with the most simple code. — noAxioms

I agree that one can write code to help a robot count the number of unmoving dots in its visual field. But I don't think a person can write code to help a robot count the number of objects or moving dots.

As for the infant process of neural development, that's an insanely complex issue that likely requires a doctorate in the right field to discuss the current view of how all that works. — noAxioms

I searched the internet to death but I didn't find anything useful.

It seems irrelevant to the topic of determinism and options. — noAxioms

It is relevant.

All the interpretations are paradox free. — noAxioms

Copenhagen interpretation for example suffers from the Schrodinger's cat paradox. It cannot explain John Wheeler's delayed choice experiment. etc. Anyway, I am not interested in going to a debate on quantum mechanics in this thread since it is off-topic. All I wanted to say is that for this thread the motion of particles in a brain is deterministic.

Not quite. That realisation is neural processes in the brain. It is not seperate from yet caused by those neural processes. — Banno

We have a slight difference here. I am a substance dualist and it seems to me that you are a physicalist. But please let's focus on the topic of the thread and put this difference in view aside.

And a babe's brain is pre-wired to recognise faces and areola. — Banno

Do you have any argument or know any study to support this claim? I am asking how an infant can distinguish between one object or two objects. I would be interested to know how an infant's brain is pre-wired then. So saying that an infant's brain is just pre-wired does not help to have a better understanding of what is happening in her/his brain when she/he realizes one object or two objects.

I assume what you are driving at when you ponder over the ability to distinguish qualitatively similar objects in the visual field is the way in which those objects are proxies for alternative affordances for action, as your initial example of two alternative paths in a maze suggests. You may be suggesting (correct me if I'm wrong) that those two "objects" are being discriminated as signifying or indicating alternative opportunities for action and you wonder how this is possible in view of the fact that, in a deterministic universe, only one of those possibilities will be realized. Is that your worry? — Pierre-Normand

Yes. I am wondering how we can realize two objects which look the same as a result of neural processes in the brain accepting that the neural processes are deterministic.

I think @Banno and @noAxioms both proposed compatibilist responses to your worry — Pierre-Normand

@noAxioms suggests that we are counting objects. I don't think that is the case when we are presented with two objects. We immediately realize two objects as a result of neural processes in the brain. We however need to count when we are presented with many objects.

@Banno suggests that an infant's brain is pre-wired. That could be true. But that does not answer how an infant could possibly realize two objects since it does not address how the brain is pre-wired.

but maybe you have incompatibilist intuitions that make you inclined to endorse something like Frankfurt's principle of alternate possibilities. Might that be the case? — Pierre-Normand

Yes. We are morally responsible if we could do otherwise. That means that we at least have two options to choose from. The options are however mental objects, like to steal or not to steal, which are slightly harder to discuss but I think that we are dealing with the same category when we realize two objects in our visual field or when we realize two mental objects. So I think we can resolve all the discussions related to the reality of options if we can understand how the brain can distinguish two objects in its visual field first.

I think Banno and @noAxioms both proposed compatibilist responses to your worry, — Pierre-Normand

A compatibilist says that free will and determinism are compatible with each other, but I would need both words more precisely defined were I to agree with that.

noAxioms suggests that we are counting objects. — MoK

I was showing the counting of options, not objects.

I agree that one can write code to help a robot count the number of unmoving dots in its visual field. — MoK

You are complicating a simple matter. I made no mention of the fairly complex task of interpreting a visual field. The average maze runner doesn't even have a visual field at all, but some do.
All I am doing is showing the utterly trivial task of counting options, which is a task easily performed by a determinsitic entity, answering your seeming inability to realize this when you state "So I am wondering how can deterministic processes lead to the realization of options".

The solution is to count the options (in the maze example, paths away from current location) and if there is more than one, options have been realized. If there is but one, it isn't optional. The means by which these options are counted is a needless complication that is besides the point.

But I don't think a person can write code to help a robot count the number of objects or moving dots.

I wrote code that did exactly that. It would look at a bin of parts and decide on the next one to pick up, and would determine the angle at which to best do that. This was 45 years ago when this sort of thing was still considered innovative.

Copenhagen interpretation for example suffers from the Schrodinger's cat paradox.

Nonsense. Just because you don't know how it explains a scenario doesn't mean it doesn't explain it. Copenhagen was developed as an epistemological interpretation which means the observer outside the box doesn't know (wave function describing state) the cat state and the observer inside has a more collapsed wave function state. Super easy.
Sure, off topic, so I'll leave off the delayed-choice thingy.
But your assertion that Bohmian mechanics is the only valid interpretation (a deterministic one) is on topic, and thus the falsification of the other interpretations is very much on topic.

Again, I counted six kinds of determinism, and some of those are almost certainly the case and some of them are almost certainly not the case. Bohmian mechanics was number 2.

We are morally responsible if we could do otherwise. That means that we at least have two options to choose from. — MoK

Moral responsibility is far more complicated than that, as illustrated by counterexamples, but the core is correct. There being more than one course of action available, and it is very hard to come up with an example where that is not the case. I am in a maze, but find myself embedded in the concrete walls instead of the paths between. I have no options, and thus am not responsible for anything I do there.

The options are however mental objects, like to steal or not to steal

Stealing and not stealing are physical actions, not mental objects. Bearing moral responsibility for one's mental objects is a rare thing, but they did it to Jimmy Carter, about a moral person as they come.

The fallacy seems to be in the assertion that determinism somehow takes away choice, which of course is nonsense since we'd not have evolved large (and very expensive) brains if not to make better choices. I cannot think of a single way that a choice can be made better by a non-deterministic process than by a similar but deterministic process. I invite such an example, but a deterministic algorithm implemented on a non-deterministic information processor is still a deterministic process.

I was showing the counting of options, not objects. — noAxioms

We don't count options if a few are presented to us. We just realize the number of options right away as a result of neural processes in the brain. I am interested in understanding what is happening in the brain when we are performing such a simple task.

You are complicating a simple matter. I made no mention of the fairly complex task of interpreting a visual field. The average maze runner doesn't even have a visual field at all, but some do.
All I am doing is showing the utterly trivial task of counting options, which is a task easily performed by a determinsitic entity, answering your seeming inability to realize this when you state "So I am wondering how can deterministic processes lead to the realization of options".

The solution is to count the options (in the maze example, paths away from current location) and if there is more than one, options have been realized. If there is but one, it isn't optional. The means by which these options are counted is a needless complication that is besides the point. — noAxioms

No, you consider the existence of options granted and then offer a code that is supposed to work and counts options. Thanks, but that is not what I am looking for.

Stealing and not stealing are physical actions, not mental objects. Bearing moral responsibility for one's mental objects is a rare thing, but they did it to Jimmy Carter, about a moral person as they come. — noAxioms

I am talking about available options to a thief before committing the crime.

What does happen at the neural level when the infant realizes the object, and distinguishes it from the background? — MoK

I imagine it entails pattern recognition: seeing the same image pattern against a relatively constant background. Artificial neural networks learn patterns, and they are considerably simpler that biological neural networks because they lack neuroplasticity (the growing of new neurons and synapses).

So I am wondering how can deterministic processes lead to the realization of options. — MoK

Options that are before us lead us to mentally deliberate to develop a choice. If we could wind the clock back, could we actually have made a different choice? Clearly, if determinism is true, then we could not. But if determinism is false- why think our deliberation would have led to a different outcome? The same mental factors would have been in place.

It seems to me you are talking about the Hard Problem of Consciousness. Photons hit retina; signals go to the brain; pattern recognition shows that there are two possible paths; stored information of past encounters with similar patterns are triggered; on and on and on. But, unlike the Roomba, I realize I have options.

I imagine it entails pattern recognition: seeing the same image pattern against a relatively constant background. Artificial neural networks learn patterns, and they are considerably simpler that biological neural networks because they lack neuroplasticity (the growing of new neurons and synapses). — Relativist

I did an extensive search and I found many methods for object recognition. Here, you can find two main methods, namely CNN, and YOLO. Granted that objects are recognized I am interested to know methods for counting objects. I did an extensive search on the net and got lost since it seems that the literature is very very rich on this topic! The current focus of research is to find the best method for counting the very high dense number of objects where objects could overlap for example. Here is a review article that discusses the CNN method for crowd counting. I am interested in a simple neural network that can count a limited number of isolated objects though. I will continue the search and let you know if I find anything useful.

Options that are before us lead us to mentally deliberate to develop a choice. If we could wind the clock back, could we actually have made a different choice? Clearly, if determinism is true, then we could not. But if determinism is false- why think our deliberation would have led to a different outcome? The same mental factors would have been in place. — Relativist

I am not interested in discussing the decision here. I am interested in understanding how we realize two objects so swiftly. If I show you two objects, you without any counting realize that there are two objects in your vision field. The same applies when you are in a maze. You realize that there are two paths available to you without counting as well. The mechanism is completely deterministic though. Two objects, two paths in a maze, etc. we are dealing with the same topic, and although the mechanism is fully deterministic we could recognize two options. So that part of the puzzle is solved for me.

It seems to me you are talking about the Hard Problem of Consciousness. — Patterner

No, here I am interested in understanding how we realize objects/options in our vision fields. Please read the previous post if you are interested.

@Relativist Ok, after a long search, I found an interesting thesis that deals with a neural network that can count. I read up to chapter 3. It is late now and time to go to bed! :wink:

It looks interesting. I'll read it when I get a chance. The bibliography also lists some references that may also be helpful.

Yes. I am wondering how we can realize two objects which look the same as a result of neural processes in the brain accepting that the neural processes are deterministic. — MoK

I think you may be using the word "realize" meaning "resolve" (as used in photography, for instance, to characterise the ability to distinguish between closely adjacent objects.) Interestingly, the polysemy of the word "resolve", that can be used to characterise an ability to visually discriminate or characterise the firmness in one's intention to pursue a determinate course of action suggests that they are semantically related, with the first one being a metaphorical extension of the second.

I know you're talking about mental processing of visual data, but that's far more complex than anybody here is qualified to answer, so I am instead picking statements that seem to be falsified by a simple, understandable model.

No, you consider the existence of options granted — MoK

We were considering a fork in the path of a maze. Are they not a pair of options?
Sure, one cannot choose to first go down both. Of the options, only one can be chosen, and once done, choosing otherwise cannot be done without some sort of retrocausality. They show this in time travel fictions where you go back to correct some choice that had unforeseen bad consequences.

I guess I don't know what you consider to be options.

I am talking about available options to a thief before committing the crime.

So you do grant the existence of multiple options before choosing one of them. What part of the maze example then is different than the crime example?

But, unlike the Roomba, I realize I have options. — Patterner

A Roomba wouldn't work if it didn't realize options. If there are two paths to choose from, it needs to know that. If it always picked the left path, there would be vast swaths of floor never visited. It needs awareness of alternative places to go.

What fundamentally do you do that a Roomba doesn't? If you mean it is not remote controlled, I'll agree. It makes its own choices. The RC car on the other hand is remote controlled and has no free will of its own. That's a fundamental distinction between the Roomba and the RC car, but I ask about the Roomba and you, because I suspect you're the RC car, a puppet of another.

But, unlike the Roomba, I realize I have options.
— Patterner
A Roomba wouldn't work if it didn't realize options. — noAxioms

How about wording it this way:
A Roomba wouldn't work if it didn't realize it has options.


I'm afraid you've lost me, regarding the puppet.

Yes, the bibliography also lists the references which could be useful. I will go through them after I finish the thesis.

I am unsure whether we first realize two objects and then distinguish/resolve them from each other upon further investigations or first distinguish/resolve them from each other and then count them and realize that there are two objects. The counting convolutional neural network works based on later.

I am unsure whether we first realize two objects and then distinguish/resolve them from each other upon further investigations or first distinguish/resolve them from each other and then count them and realize that there are two objects. The counting convolutional neural network works based on later. — MoK

We can indeed perceive a set of distinct objects as falling under the concept of a number without there being the need to engage in a sequential counting procedure. Direct pattern recognition plays a role in our recognising pairs, trios, quadruples, quintuples of objects, etc., just like we recognise numbers of dots on the faces of a die without counting them each time. We perceive them as distinctive Gestalten. But I'm more interested in the connection that you are making between recognising objects that are actually present visually to us and the prima facie unrelated topic of facing open (not yet actual) alternatives for future actions in a deterministic world.

We can indeed perceive a set of distinct objects as falling under the concept of a number without there being the need to engage in a sequential counting procedure. Direct pattern recognition plays a role in our recognising pairs, trios, quadruples, quintuples of objects, etc., just like we recognise numbers of dots on the faces of a die without counting them each time. We perceive them as distinctive Gestalten. — Pierre-Normand

I would think there's a limit to this. We might recognize the number of dots on a die because of the specific arrangements that we've seen so many times. Would we do as well with five or six randomly arranged objects? Or ten or fifteen?

I know you're talking about mental processing of visual data, but that's far more complex than anybody here is qualified to answer, so I am instead picking statements that seem to be falsified by a simple, understandable model. — noAxioms

I found this useful thesis about counting objects by a convolutional neural network.

We were considering a fork in the path of a maze. Are they not a pair of options? — noAxioms

Sure they are.

Sure, one cannot choose to first go down both. Of the options, only one can be chosen, and once done, choosing otherwise cannot be done without some sort of retrocausality. They show this in time travel fictions where you go back to correct some choice that had unforeseen bad consequences. — noAxioms

The point is that both paths are real and accessible, as we can recognize them. However, the process of recognizing paths is deterministic. This is something that hard determinists deny. The decision is a separate topic though. I don't think that the decision results from the brain's neural process. The decision is due to the mind. That is true since any deterministic system halts when you present it with options. A deterministic system always goes from one state to another unique state. If a deterministic system reaches a situation where there are two states available for it it cannot choose between two states therefore it halts. When we are walking in a maze, our conscious mind is aware of different situations always. If there is one path available then we simply proceed. If we reach a fork we realize the options available to us, namely the left and right path. That is when the conscious mind comes into play, realizes the paths in its experience, and chooses one of the paths. The subconscious mind then becomes aware of the decision and acts accordingly.

I guess I don't know what you consider to be options. — noAxioms

By options, I mean a set of things that are real and accessible and we can choose from.

So you do grant the existence of multiple options before choosing one of them. What part of the maze example then is different than the crime example? — noAxioms

In the example of the maze, the options are presented to the person's visual fields. In the case of rubbery the options are mental objects.